# SMART TOOLS FOR CARING: NANOTECHNOLOGY MEETS MEDICAL CHALLENGES

EDITED BY : Giada Genchi and Gianni Ciofani PUBLISHED IN : Frontiers in Bioengineering and Biotechnology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-806-6 DOI 10.3389/978-2-88945-806-6

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# SMART TOOLS FOR CARING: NANOTECHNOLOGY MEETS MEDICAL CHALLENGES

Topic Editors:

Giada Genchi, Istituto Italiano di Tecnologia, Italy Gianni Ciofani, Istituto Italiano di Tecnologia and Politecnico di Torino, Italy

Image by Elisa Mele, Giada Genchi and Gianni Ciofani

Citation: Genchi, G., Ciofani, G., eds. (2019). Smart Tools for Caring: Nanotechnology Meets Medical Challenges. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-806-6

# Table of Contents

*05 Editorial: Smart Tools for Caring: Nanotechnology Meets Medical Challenges*

Giada G. Genchi and Gianni Ciofani


Anna Denchai, Daniele Tartarini and Elisa Mele

*52 Borrowing From Nature: Biopolymers and Biocomposites as Smart Wound Care Materials*

Giulia Suarato, Rosalia Bertorelli and Athanassia Athanassiou

*63 One-Step Synthesis of Hexagonal Boron Nitrides, Their Crystallinity and Biodegradation*

Özlem Şen, Melis Emanet and Mustafa Çulha

*72 The ABC Guide to Fluorescent Toolsets for the Development of Future Biomaterials*

Ferdinandus and Satoshi Arai

*79 Bright Dots and Smart Optical Microscopy to Probe Intracellular Events in Single Cells*

Hideaki Fujita, Chongxia Zhong, Satoshi Arai and Madoka Suzuki


Meghan Robinson, Ian Fraser, Emily McKee, Kali Scheck, Lillian Chang and Stephanie M. Willerth

# Editorial: Smart Tools for Caring: Nanotechnology Meets Medical Challenges

Giada G. Genchi <sup>1</sup> \* and Gianni Ciofani 1,2 \*

*<sup>1</sup> Smart Bio-Interfaces, Istituto Italiano di Tecnologia, Pontedera, Italy, <sup>2</sup> Department of Mechanical and Aerospace Engineering, Politecnico di Torino, Turin, Italy*

Keywords: smart materials, remote stimulation, drug delivery, tissue engineering, clinical translation

### **Editorial on the Research Topic**

### **Smart Tools for Caring: Nanotechnology Meets Medical Challenges**

This Research Topic of Frontiers in Bioengineering and Biotechnology-Nanotechnology stems from the international workshop "Smart tools for caring: Nanotechnology meets medical challenges" held in Pontedera (Italy) on March 2, 2018 with the goal of promoting a multidisciplinary discussion on the most innovative techniques and materials with strong therapeutic potential in human healthcare. The issue presents 14 selected, peer-reviewed contributions (1 Brief Research Report, 7 Mini-Reviews, 1 Opinion, 3 Original Researches, 3 Reviews) from different research fields, ranging from materials science, to biotechnology, physics, and engineering.

#### Edited and reviwed by:

*Sandra Camarero-Espinosa, Maastricht University, Netherlands*

\*Correspondence:

*Giada G. Genchi giada.genchi@iit.it Gianni Ciofani gianni.ciofani@iit.it*

#### Specialty section:

*This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology*

Received: *07 January 2019* Accepted: *17 January 2019* Published: *30 January 2019*

#### Citation:

*Genchi GG and Ciofani G (2019) Editorial: Smart Tools for Caring: Nanotechnology Meets Medical Challenges. Front. Bioeng. Biotechnol. 7:11. doi: 10.3389/fbioe.2019.00011*

In agreement with the scopes of the workshop, this issue reviews general topics of crucial importance in the design of materials for interaction with biological systems, most notably mechanical properties, surface features (including roughness, porosity, and patterning), and their spatiotemporal variation (Cimmino et al.; Palamà et al.). The synthesis and fabrication of innovative materials are presented with focus on the modulation of cell responses in events like cell adhesion, proliferation, and differentiation (Bandiera et al.; Denchai et al.; ¸Sen et al.; Suarato et al.). In this concern, instrumental techniques for characterization of bio-/non bio-interactions (Arai and Ferdinandus; Fujita et al.; Tonda-Turo et al.), light-based methods for remote stimulation of materials and their anatomical targets (Bossio et al.; Polino et al.), and complex approaches for diagnosis and treatment of pathological conditions through genetic manipulations of somatic cells, nanomaterials, and robotic devices (Cervadoro et al.; Robinson et al.; Soto and Chrostowski) are presented.

More in detail, the importance of material properties like stiffness is discussed with relevance to the fabrication of three-dimensional constructs of glioblastoma multiforme, with the final goal of elaborating reliable therapies against this aggressive tumor of the central nervous system (Palamà et al.). Along with stiffness, other biochemical and biophysical signals like ligand exposure and topography of the materials are thoroughly reviewed, and the significance of their time- and space-variant patterns in response to external stimulations (light, temperature, electricity, or even enzymes and cells) is discussed in relation to the elaboration of specific protocols for exposure of cells to dynamic bioinstructive signals (Cimmino et al.).

The importance of the use of polypeptides (for instance collagen, fibroin, and keratin), polysaccharides (like hyaluronic acid, chitosan, and alginate) or even their mixtures of natural origin (such as in mycelia with controlled composition based on the fungal feeding substrate) is stressed as an environmental sustainable strategy for the preparation of biodegradable materials, especially in the context of wound healing (Suarato et al.).

**5**

In this Research Topic, the fabrication of novel materials with controlled architecture is addressed with electrospinning and wet chemistry techniques. Corroborated by computational models that predict and correlate surface properties with biological responses, electrospinning is reviewed as an effective technique for micro- and nanopatterning of materials to impart biologically relevant cues (under the form of fiber orientation, roughness, and density) to tune cell adhesion, proliferation, and differentiation (Denchai et al.). Electrospinning is also used as a technique for the fabrication of hybrid scaffolds with thermoresponsive hydrogels (based on recombinant elastinlike polypeptides) in view of drug loading and release in elastolytic conditions, such as those associated to wound healing (Bandiera et al.). One step synthesis of hexagonal boron nitride nanomaterials (hBNs) from different boron sources and in the absence of catalysts is also described for possible application in cancer treatment. In particular, evidences on hBN stability and degradation under both oxidative and hydrolytic conditions are reported, supporting the use of colemanite as a boron source (¸Sen et al.).

Microscopy techniques for investigating biological events with intracellular compartment resolution are discussed: basic indications of the choice and application of fluorescent probes are presented with relevance to photothermal and photodynamic therapeutic approaches (Arai and Ferdinandus). Moreover, methods based on probes (genetically encoded fluorophores, chemical fluorescent molecules, and other nanomaterials) are reviewed along with purely optical, probe-less methods (Fujita et al.). Among the techniques for characterization of bio-/non biointeractions, quartz crystal microbalance is described as a tool for dynamic investigation of chemical, mechanical, and electrical properties of surfaces interacting with cells and of biological response to biomedical devices (Tonda-Turo et al.).

The activation of smart materials on their target sites by non-invasive approaches is here discussed with a focus on phototransduction methods based on photovoltaic devices or photocatalytic nanoparticles. Photovoltaic devices are presented for stimulation of electrogenic cells in the retina and for promotion of non-electrogenic cell proliferation, such as during angiogenesis and wound healing (Polino et al.). Photocatalytic nanoparticles based on poly(3-hexylthiophene) with semiconductive properties are exploited for modulation of calcium influxes in HEK-293 cells through the generation of reactive oxygen species under visible light irradiation, denoting potential for cell behavior tuning with high spatio-temporal resolution (Bossio et al.).

This Research Topic also includes a concise review of innovative materials with theranostic potentiality for modulation of cell response in terms of inflammation control, immune system targeting, and treatment of atherosclerosis (Cervadoro et al.). An innovative approach for genetic reprogramming of somatic cells (astrocytes) into neurons with a non-viral method (based on a fusion protein containing a transcription factor and a protein transduction domain, in association to SMAD, Notch, and histone deacetylase inhibitors) is presented for possible treatment of central nervous system injuries (Robinson et al.). The need of considering emerging clinical translation and potential commercial use of micro- and nanorobotic devices is also thoroughly reviewed: even in this case, the discussion stresses on fine spatiotemporal control of activities such as delivery of drugs, imaging agents or cells, and biopsy collection. The discussion is completed by considerations on experimental methodology reproducibility and standardization requisites, on intellectual property protection, on device fabrication and on the implementation of innovative devices in current therapeutic protocols for the elaboration of concrete translational medicine approaches (Soto and Chrostowski).

To conclude, we hope this Research Topic will represent a useful reference of easy consultation for those readers interested to gather concise information on the state of the art, as well as to understand the necessary directions of multidisciplinary research on innovative materials and related topics of high relevance to human healthcare for the immediate future.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### ACKNOWLEDGMENTS

This research is partially supported by the Compagnia di San Paolo Starting Grant Number 55\_AI16GC01.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Genchi and Ciofani. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Microenvironmental Rigidity of 3D Scaffolds and Influence on Glioblastoma Cells: A Biomaterial Design Perspective

Ilaria Elena Palamà<sup>1</sup> \*, Stefania D'Amone<sup>1</sup> and Barbara Cortese<sup>2</sup> \*

<sup>1</sup> Nanotechnology Institute, CNR-Nanotechnology Institute, Lecce, Italy, <sup>2</sup> Nanotechnology Institute, CNR-Nanotechnology Institute, University La Sapienza, Rome, Italy

Keywords: 3D scaffolds, glioblastoma, stiffness, neurospheres, microenvironment

### INTRODUCTION

### Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Luca Ceseracciu, Fondazione Istituto Italiano di Technologia, Italy Giulia Suarato, Fondazione Istituto Italiano di Technologia, Italy Piergiorgio Gentile, Newcastle University, United Kingdom

#### \*Correspondence:

Ilaria Elena Palamà ilaria.palama@nanotec.cnr.it Barbara Cortese barbara.cortese@nanotec.cnr.it

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 01 August 2018 Accepted: 03 September 2018 Published: 24 September 2018

#### Citation:

Palamà IE, D'Amone S and Cortese B (2018) Microenvironmental Rigidity of 3D Scaffolds and Influence on Glioblastoma Cells: A Biomaterial Design Perspective. Front. Bioeng. Biotechnol. 6:131. doi: 10.3389/fbioe.2018.00131 Glioblastoma (GBM), or grade IV glioma, is an extremely aggressive tumor that infiltrates through the brain leaving the patient with a median survival time from 12 to 15 months (Ostrom et al., 2013). Individual aspects of the microenvironment features play a critical role on GBM cell dynamics and treatment resistance (Bellail et al., 2004; Zamecnik, 2005; Calabrese et al., 2007). Because of GBM's aggressive and invasive behavior, inhibition of GBM migration is envisaged as an important therapeutic objective (Bravo-Cordero et al., 2012; Wells et al., 2013). However, current models fail to account for the complex brain microenvironment. The demand of preclinical models that can faithfully mimic the clinical scenario may bridge the discrepancy between preclinical and clinical data and aid to develop treatments that are more effective.

The extracellular matrix (ECM) of the GBM microenvironment is constitutively composed of the polysaccharide hyaluronic acid (HA), and in a distinctive minor degree of tenascin-C, collagen IV and V, fibronectin, and laminin (Giese and Westphal, 1996; Rape et al., 2014). Also, typically with high glioma grade, the HA's cellular receptor CD44 is overexpressed, suggesting that CD44 enriched cells invade more efficiently the brain parenchyma (Bellail et al., 2004). GBM malignancy is furthermore promoted through interactions with the other aforementioned ECM components through different biochemical pathways (Sarkar et al., 2006; Lathia et al., 2012), which trigger an increase of the concentration of the non-cellular components (Bellail et al., 2004; Lathia et al., 2012). This increased density of the tumor ECM consequently increases the mechanical stiffness of the microenvironment (Ananthanarayanan et al., 2011; Wiranowska and Rojiani, 2011; Pathak and Kumar, 2012; Pedron and Harley, 2013; Kim and Kumar, 2014; Umesh et al., 2014; Heffernan et al., 2015).

In our opinion, to gain further insights into tumor invasiveness, heterogeneity and treatment resistance, we have to look more deeply at how cell behavior is influenced by matrix stiffness (a process known as mechanotaxis or durotaxis) (Lo et al., 2000; Cortese et al., 2009; Palamà et al., 2012, 2016).

### WHAT IS THE INFLUENCE OF THE 3D SCAFFOLD MECHANICAL PROPERTIES?

Notably, 2D platforms do not adequately mimic the in vivo tumor environment. Recent work has focused on 3D scaffolds and matrix influence on cells with different materials and cells, as reported in **Table 1**. However, inconsistencies on how 3D scaffold stiffness affect cell proliferation and influence drug delivery and treatment resistance have been reported in literature (Wang et al., 2014, 2016; Heffernan et al., 2015; Pedron et al., 2015; Lv et al., 2016; Palamà et al., 2017). In order to tune the mechanical properties of different materials, the most common methods are (1) altering the crosslink density and (2) changing the base polymer concentration which both influence different parameters, such as the ECM architecture, stiffness, pore size, diffusion of soluble factors of the scaffolds, and ligand density. A 3D culture platform that aims to mimic the native GBM microenvironment has also the additional requirement of containing HA. Pure HA lacks mechanical strength and the ability to promote cell adhesion due to its anionic properties (Wang et al., 2012). Moreover, it does not allow control over mechanical stiffness. These downsides have been partly overcome by using synthetic ECM polymers (Lutolf and Hubbell, 2005; Seliktar, 2012). One semi-synthetic material, predominantly used to independently tune the stiffness of the scaffold, is HA-based hydrogel functionalized to favor cell adhesion. For example, Ananthanarayanan studied HA gels of varying stiffness embedded with GBM spheroids and corroborated that their invasive capacity and morphological patterns were similar to what was seen in vivo in human brain slices, in opposition to glioma cells cultured in 2D and 3D collagen contexts (Ananthanarayanan et al., 2011). Differences were theorized to be related to the variation of expression of CD44. This was confirmed by Harley and co-workers, who identified CD44 as a key driver of glioma malignancy with cells encapsulated in gelatin and PEG-based hydrogels grafted with a HA hydrogel network (Pedron et al., 2013). Analogous observations were made by Erikson using porous chitosan–hyaluronic acid scaffolds of different stiffness, obtained varying the chitosan content. With a higher polymer content, the pore walls were thicker, with reduced interconnections between pores as well as the pore size (Erickson et al., 2018). Stiffness was shown to influence the morphology of the cell aggregates as well as the expression of


drug resistance, hypoxia, and invasion-related genes (Mih et al., 2011; Zustiak et al., 2016; Erickson et al., 2018).

### TYPE OF CELLS USED IN THE IN VITRO MODELS

A further critical key parameter is the choice of the cells used (Zustiak et al., 2014). Typically, commercially available human tumor cell lines are used, but they neglect predicting clinical outcomes due to different genetic aberrations. Glioblastoma stem-like cells (GSC) can mimic the tumor of origin being tumor cells with stem cell properties (Saha et al., 2008). However, they require isolating stem cells from each tumor patient and expansion to an adequate number within a clinically acceptable time. Aggregated cultures would recapitulate better the GBM microenvironment, allowing cell–cell contacts and collective migration. Non-adherent cultures lack the cell–matrix interactions present in the tumor stroma, whereas complex spherical cancer models (i.e. non-adherent cancer cell linederived spheroids, or spheroids derived from primary tumor dissociation) can promote cell–cell interactions. The use of patient-derived cells cultured as neurospheres is a significant advance respect to glioma cell lines (Rao et al., 2013; Cha et al., 2016) however they do not accurately reproduce the original tumor composition due to heterogeneity loss and lack of an adhesive matrix. A solution could be represented by GBM organoids (Hubert et al., 2016), although these require months for generation, thus becoming useless in aid of patient treatment and not necessarily being an improvement to the patient outcomes (Oh et al., 2014), whereas neurosphere cultures can be established within only few weeks.

### INFLUENCE OF COMPOUNDING STIMULI ON THE SCAFFOLDS

Diverging results have also been implicated with the matrix metalloprotease (MMP) secretion. For example, an increase of MMP-9 production in hyaluronic acid-based hydrogels with increasing stiffness was reported by Pedron's group while Wang and colleagues described an opposite behavior (Pedron et al., 2013, 2015; Wang et al., 2014). The reason, in our opinion, is related to the difficulty to discriminate the role of stiffness or of the biochemical stimuli on cell invasion. In fact, only a few works report a selected degree of decoupling of the mechanical properties, porosity, and/or biochemical cues. The interference of other compounding stimuli in the design of functional cell culture substrates should be minimized if not isolated. Changes in the ligand density and the pore size of the matrix may obstruct migration of cells and alter solute diffusion (Shu et al., 2002). For example, Cha used a different molecular weight of

### REFERENCES

Ananthanarayanan, B., Kim, Y., and Kumar, S. (2011). Elucidating the mechanobiology of malignant brain tumors using a brain matrix-mimetic HA to simply coat the collagen I fibers without modifications and crosslinking (Cha et al., 2016). Using a higher molecular weight and 3D structure of the polymer may have induced different cell responses. Rao reported a high degree of thiolation, which may have altered the bioactivity of the substrate (Rao et al., 2013). Kumar and co-workers managed to decouple the effects by assembling hydrogel networks of collagen I and agarose and increasing the stiffness by increasing the concentration of non-adhesive agarose while keeping collagen I levels constant (Ulrich et al., 2010). However, increasing the concentration of agarose may result in smaller pores that restricted migration on stiffer hydrogels. Moreover, the presence of the agarose interfering with collagen fiber deformation and bundling may have thereby restricted local ability of tumor cells to stiffen their microenvironment (Kilian and Mrksich, 2012; Rape et al., 2015).

In conclusion there is still no existing artificial GBM microenvironment which can replace an in vivo model. It is essential to ask if it is worth to increase the complexity of the ECM microenvironment and to define which parameters are at least required to achieve a physiologically relevant model ex vivo. We think that tuning matrix stiffness will be pivotal at both a preclinical and clinical level, to move forward this field of investigation of cell behavior during tumorigenesis thereby providing an important tool to target and investigate the more effective therapy at different stages of cancer progression. To unravel the tumor invasiveness and to demonstrate their clinical value, a fully comprehensive analysis approach has to be achieved. This invites further study and highlights the importance of conducting parallel measurements using spheroid cell lines in highly multistructured conditions as well as comparisons with patient outcomes.

### AUTHOR CONTRIBUTIONS

IP, SD, and BC wrote the manuscript. All authors reviewed the manuscript and have given approval to the final version of the manuscript.

### FUNDING

FIRS project Tecnopolo di nanotecnologia e fotonica per la medicina di precisione (grant no: CIPE n. 78/2017 B83B17000010001).

### ACKNOWLEDGMENTS

This study is supported by the Italian Association for Cancer Research (AIRC) through the grant MFAG n. 16803 and partially by the FIRS project Tecnopolo di nanotecnologia e fotonica per la medicina di precisione.

hyaluronic acid hydrogel platform. Biomaterials 32, 7913–7923. doi: 10.1016/j.biomaterials.2011.07.005

Bellail, A. C., Hunter, S. B., Brat, D. J., Tan, C., and Van Meir, E. G. (2004). Microregional extracellular matrix heterogeneity in brain modulates glioma cell invasion. Int. J. Biochem. Cell Biol. 36, 1046–1069. doi: 10.1016/j.biocel.2004.01.013


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Palamà, D'Amone and Cortese. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Spatio-Temporal Control of Cell Adhesion: Toward Programmable Platforms to Manipulate Cell Functions and Fate

#### Chiara Cimmino1,2, Lucia Rossano1,2, Paolo Antonio Netti 1,2 and Maurizio Ventre1,2 \*

<sup>1</sup> Department of Chemical, Materials and Industrial Production Engineering, University of Naples Federico II, Naples, Italy, <sup>2</sup> Center for Advanced Biomaterials for Healthcare@CRIB, Fondazione Istituto Italiano di Tecnologia, Naples, Italy

### Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

### Reviewed by:

Debora Berti, Università degli Studi di Firenze, Italy Sandra Camarero-Espinosa, Maastricht University, Netherlands Abbas Shafiee, Queensland University of Technology, Australia

> \*Correspondence: Maurizio Ventre maventre@unina.it

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 06 October 2018 Accepted: 21 November 2018 Published: 04 December 2018

#### Citation:

Cimmino C, Rossano L, Netti PA and Ventre M (2018) Spatio-Temporal Control of Cell Adhesion: Toward Programmable Platforms to Manipulate Cell Functions and Fate. Front. Bioeng. Biotechnol. 6:190. doi: 10.3389/fbioe.2018.00190 Biophysical and biochemical signals of material surfaces potently regulate cell functions and fate. In particular, micro- and nano-scale patterns of adhesion signals can finely elicit and affect a plethora of signaling pathways ultimately affecting gene expression, in a process known as mechanotransduction. Our fundamental understanding of cell-material signals interaction and reaction is based on static culturing platforms, i.e., substrates exhibiting signals whose configuration is time-invariant. However, cells in-vivo are exposed to arrays of biophysical and biochemical signals that change in time and space and the way cells integrate these might eventually dictate their behavior. Advancements in fabrication technologies and materials engineering, have recently enabled the development of culturing platforms able to display patterns of biochemical and biophysical signals whose features change in time and space in response to external stimuli and according to selected programmes. These dynamic devices proved to be particularly helpful in shedding light on how cells adapt to a dynamic microenvironment or integrate spatio-temporal variations of signals. In this work, we present the most relevant findings in the context of dynamic platforms for controlling cell functions and fate in vitro. We place emphasis on the technological aspects concerning the fabrication of platforms displaying micro- and nano-scale dynamic signals and on the physical-chemical stimuli necessary to actuate the spatio-temporal changes of the signal patterns. In particular, we illustrate strategies to encode material surfaces with dynamic ligands and patterns thereof, topographic relieves and mechanical properties. Additionally, we present the most effective, yet cytocompatible methods to actuate the spatio-temporal changes of the signals. We focus on cell reaction and response to dynamic changes of signal presentation. Finally, potential applications of this new generation of culturing systems for in vitro and in vivo applications, including regenerative medicine and cell conditioning are presented.

Keywords: dynamic platforms, cell adhesion, switchable signals, ligands, topographic, patterns, substrate stiffness

## INTRODUCTION

A major part of our understanding of cell biology is occupied by the effects of soluble signals (drugs, small molecules, peptides, proteins, growth factors) on cell behavior. In fact, several routes have been developed to formulate well-defined media that enabled studying cell response to specific soluble signals systematically (Yao and Asayama, 2017). There are comparatively less studies aimed at investigating the effects of insoluble matricellular signals, in the form of topographic relieves, patterns of biochemical moieties, or mechanical signals, on cell functions. These proved to dramatically affect cell behavior (Ventre et al., 2012). In different biomedical and clinical applications, cells are in close contact with material surfaces that can present one or more signals simultaneously. For instance, prosthetic devices or synthetic scaffolds invariably expose surfaces to the biological environment and the biochemical/biophysical characteristics of the surfaces can have a great impact on the implant performance in vivo (Williams and Bhatia, 2014; Rasouli et al., 2018). Therefore, achieving a sound knowledge on the role of material properties on cell functions would provide valuable elements to engineer devices with improved functions. This requires implementing design concepts and fabrication technologies that enable reproducing certain features of the extracellular matrix (ECM) that most effectively affect cell functions and fate. Advancements in materials engineering, functionalization methods and most importantly micro- and nano-fabrication technologies provided researchers with "artificial" alternatives to conventional rigid plates or glass, which more closely mimic the native microenvironment (Leijten and Khademhosseini, 2016). The integration of micro- and nano-engineered platforms with cell cultures not only allowed to elicit specific cellular reactions, thus controlling their functions and fates, but also enabled understanding cell-signal interactions. In fact, micro- and nanoengineered platforms display signals whose spatial arrangement may be targeted to the whole cell, subcellular compartments, cluster of receptors or even individual receptors, thus enabling to achieve a fine-tuning of a broad spectrum of signaling pathways (Dalby et al., 2014; Donnelly et al., 2018). In most of the cases, the signals displayed by materials are static in nature, i.e., once embossed on the culturing platform they cannot be changed in time and space. The native ECM is far from being a static repository of signals, as it constantly changes in time and space in response to or as a part of growth, aging, disease, injuries. For instance, temporal variations of the ECM, including changes in the microarchitecture and stiffness, play an important role in regulating different biological processes in vivo including differentiation and morphogenesis, but also the progression of pathologies (Lu et al., 2012; Handorf et al., 2015).

Cell biologists usually relied on reductionist approaches to study cell-signal interactions in vitro seeking systems aimed at reducing the complexity of interactions or at eliciting specific cell responses to investigate cell-signal interplay. These systems were instrumental to shape our understanding on the mechanisms underlying cell recognition and reaction to signals, but in most of the cases they are not able to capture specific aspects as multi signal stimulation or dynamic changes. This calls for novel platforms able to more closely mimic the ECM both in terms of signal display and dynamic changes of these signals.

Most of our knowledge on cell-material recognition and response to biochemical/biophysical signals arises from studies performed in two-dimensions (2D). Although most cells live in a three-dimension (3D) context in vivo, it is still debated whether adhesion in 3D is identical to a 2D environment or follows different routes (Harunaga and Yamada, 2011; Doyle and Yamada, 2016). This notwithstanding, we prefer to focus our attention on 2D systems as the development of dynamic materials is recent and mostly concerns planar surfaces. Examples of dynamic 3D environments have been already presented which showed enlightening results (Khetan et al., 2013; Das et al., 2016; Brown et al., 2018). However, achieving a precise control on the adhesive processes in 3D at a subcellular level is still challenging. Furthermore, 2D platforms should not be necessarily considered as pure artifacts. Specifically engineered materials proved to be very effective in controlling even complex aspects of cell behavior including stem cell self-renewal, targeted differentiation and morphogenesis (Nikkhah et al., 2012; Ventre et al., 2012; Ventre and Netti, 2016b). These examples have an intrinsic usefulness as the outcomes could be exploited in applications such as drug screening and discovery, cell-based therapies, and tissue engineering.

Increasing the complexity of material substrate to control cell functions and fate in vitro with the introduction of dynamically changing signals would better mimic a natural context thus enabling the possibility to guide and stimulate cells with improved effectiveness.

In this review we first illustrate the basic mechanism of cell ECM or material interactions focusing on cell adhesion processes to provide basic guidelines to engineer bioactive platforms to control cell behavior. We also discuss notable examples of cell interaction with "static" platforms to provide insights into cell's reactions and responses to specific signal arrangements, being more details on this aspect reported elsewhere (Bettinger et al., 2009; Ventre et al., 2012; Yao et al., 2013). The central part of the article reviews strategies and technologies to encode dynamic signals on material platforms. In particular, this work focuses on dynamic changes of ligands and their spatial patterns, micro- and submicro-scale topographies and material stiffness. Furthermore, emphasis is given to response of cells to the spatio-temporal changes of signal display. Finally, we will address limits of the current platforms and technologies suggesting possible ways to improve their performances thus creating systems that can affect cell functions in a more thorough and consistent manner.

### THE PROCESS OF CELL ADHESION AND CELL RESPONSE TO MATERIAL SIGNALS

Cells interact with the culturing microenvironment, including material surfaces, through an array of receptors that enable perceiving different chemical/physical cues such as roughness, hydrophobicity, ligand density and distribution, stiffness, and charge. The receptors involved in such a process of recognition are located on the cell membrane and they include immunoglobulin super family of cell adhesion molecules (IgCAMs), cadherins, integrins, selectins, and proteoglycans, as well as non-integrin collagen, laminin, and elastin receptors (Hinek, 1996; Campbell and Humphries, 2011; Humphries et al., 2015; Di Cio and Gautrot, 2016). In particular, the heterodimeric, transmembrane integrin receptors not only are responsible for initiating and maintaining stable adhesion, but take part in the formation of multiprotein signaling hubs that trigger biochemical events that greatly influence cellular behavior (Humphries et al., 2015). While non-specific interactions of cells with materials dominate early adhesion events, stable cell adhesions form when surfaces display ligands to which integrin receptors bind specifically (Cohen et al., 2004; Parsons et al., 2010). Integrins are constituted by α and β subunits. So far, 18 α and 8 β subunits have been identified that can combine to form 24 different receptors with different binding properties (Campbell and Humphries, 2011). Through different combinations of α and β subunits it is possible to determine the affinity of the heterodimer with specific ligands. Several aminoacidic sequences have been recognized to interact with integrins specifically. For example, IKLLI, LGTIPG, LRE, LRGDN, PDGSR, RGD, YIGSR, YKVAV in laminin, DGEA GFOGER and RGD in collagen I and KQAGDV, REDV, RGD, and PHSRN in fibronectin (FN). When sufficiently close to ligands, integrin dimers can acquire an activated form by undergoing to a conformational change that enhances the affinity to ligands and allows the interaction with proteins and signaling molecules from the cytoplasmic side (Calderwood, 2004). After activations, integrins cluster together in discrete locations of the membrane facing the adhesive surface. According to the availability of ligands and stability of the integrin-ligand complex, cluster may mature in focal complexes up to micrometric multiprotein entities named focal adhesions (FAs) (**Figure 1**). Proteins from the cytoplasmic side may stabilize the protein gathering and/or may act as a bridge with the actin cytoskeleton. Integrins, once activated, bind to actin fibers through cytoplasmic proteins such as α–actinin, talin, and vinculin. Cytoskeleton-generated contractile forces promote FA maturation through conformational change of mechanosensitive proteins that modify their functions. For example, vinculin undergoes to a conformational change upon force application, which improves the affinity toward other proteins involved in FA stabilization (Dumbauld et al., 2013). Also, talin possesses cryptic binding sites for vinculin that unfold and become available upon stretching (del Rio et al., 2009). Other examples of mechanosensitive proteins constituting the adhesion plaques are paxillin and p130cas (Janoštiak et al., 2014). Therefore, FAs constitute the gate through which the cytoskeleton interacts with the extracellular environment, allowing the perception of different signals, such as biochemical (ligands density and their spatial distribution) and biophysical (topography and mechanical characteristics), and the reaction to these is accompanied by alterations in FA features, cytoskeleton assembly and cell-generated forces (Geiger et al., 2009).

Moreover, molecules related to FAs are important signaling proteins such as kinases [focal adhesion kinase (FAK), Src] and GTPase (Rho, Rac) that alter their activities when subjected to the mechanical forces exerted by the actomyosin machinery. For example, contractile forces can induce the phosphorylation of additional sites of the Src family kinases substrate p130CAS, which can eventually trigger signal transduction pathways (Sawada et al., 2006). The conversion of mechanical stimuli (either externally applied or cell-generated) into intracellular biochemical events is usually referred to as mechanotransduction (Eyckmans et al., 2011). These mechanical signals are also perceived by the nucleus, as the actin cytoskeleton is connected directly to the nuclear membrane (Isermann and Lammerding, 2013). The nucleus reacts to the forces generated by actin by changing its shape, modifying the assembly of chromatin as well as the accessibility of enzymes and transcription factors, eventually altering gene expression (Li et al., 2011; Gupta et al., 2012; Jain et al., 2013). In this scenario elements affecting cell adhesion (ligand availability and patterning or the transmission of mechanical forces) through biochemical/biophysical signals embossed on material surfaces may enable the activation of intracellular biochemical events ultimately affecting cell behavior in a process that we refer to as the material-cytoskeleton crosstalk (Ventre et al., 2012).

A series of strategies has been developed to control biochemical/biophysical properties of the culturing materials that mostly affect FA formation and growth. The first examples are those concerning functionalization with ligands. Several proteins or peptides involved in cell adhesion process may be linked to the material surfaces by covalent binding or by physical interactions. Arginylglycylaspartic acid (RGD) is among the most studied and used ligand to control cell adhesion (Hersel et al., 2003; VandeVondele et al., 2003). Early studies aimed at identifying the optimal density of the ligands that enable cell adhesion and spreading. In this regard, Massia and Hubbell observed that fibroblasts spreading on glass surfaces required a minimum RGD density of 1.0 × 10−<sup>15</sup> mol/cm<sup>2</sup> (corresponding to an interligand spacing of ∼440 nm), while a density of 1.0 × 10−<sup>14</sup> mol/cm<sup>2</sup> (interligand spacing of ∼140 nm), was necessary to promote the formation of focal contacts and stress fibers (Massia and Hubbell, 1991). Therefore, precise spatial positioning of adhesive spots may have a profound impact on the process of cell adhesion. Micro- and nano-fabrication techniques have enabled the realization of cell adhesion substrates allowing to modulate the dimension, positioning and spacing of the adhesive zones on a micro- and submicro-scale. For example, block copolymer micelle nanolithography allows to control spacing and density of individual ligands. Through this method, it was proven that adhesion formation occurs in a specific range of interligand densities, i.e., from 58 to 73 nm. Ligand densities surpassing this range cause an excessive adhesion and a decrease of cellular mobility, while lower ligand densities do not allow cell spreading (Arnold et al., 2004). Exerting a tight control on individual FA localization and shape requires achieving a sharp adhesion mismatch on the surface, i.e., zones conducive for FA formation and growth juxtaposed to others not allowing integrin engagement. This prevents FAs to grow and acquire shapes uncontrollably. Surfaces exhibiting micro- or nano-scale relieves in the form of pillars or pits can reproduce such a scenario effectively. In fact, the formation of FAs strongly depends on the geometry of the topography. Proteins and ligands can adsorb

thus generating a focal complex and then a focal adhesion. (B) The multi-layered structure of a mature focal adhesion comprises proteins with different functions: signaling (integrins, FAK, and paxillin), force transduction (vinculin and talin) and linkers to the cytoskeleton (Vasp, zyxin, and α-actinin). Reprinted with permission from Ventre and Netti (2016b). Copyright 2016 American Chemical Society.

non-uniformly or might not be readily available to cells thus impairing FA dynamics (Lord et al., 2010). There is still a lack of systematic studies that rigorously indicate which combinations of pattern height, spacing and size promote or impair cell adhesion. However, experimental evidences suggest that characteristic dimensions that most effectively interfere with FA dynamics exist, which are suggestive of a possible common mechanism of cellular response to topographies. Examples of insufficient cell adhesion have been reported when cells are seeded on nanoprotrusions with a feature height above 80 nm, as this dimension hampers integrin clustering (Lim et al., 2005b; Biggs et al., 2010). The same situation of insufficient adhesion occurs when cells are grown on pillars whose diameters are <70 nm and the interpillar spacing is >300 nm. FAs formation takes place when integrins are able to bridge the gap between neighboring features or when the distance between these is ∼70 nm (Kim et al., 2005; Lee et al., 2008; Sjöström et al., 2009). Other studies showed that the pitch and size of nanopits are also fundamental parameters in governing cell adhesion. For instance, FAs growth is confined in the interpit area, thus limiting the growth on arrays of nanopits of 100 nm in depth, 120 nm in diameter and 300 nm in spacing arranged in ordered square or hexagonal lattices (Biggs et al., 2007; Dalby et al., 2007).

The geometry of the topographic model, besides influencing the formation of adhesions, strongly affects FAs orientation. In fact, while pillars and pits arrays limit FAs elongation, microand nano-gratings guide FAs growth in a specific direction. Directional growth of FAs strongly affects cytoskeleton assembly, as stress fibers are oriented in the same direction of the underlying pattern. In this way the entire cell body assumes an elongated morphology and migrates along the pattern direction. This phenomenon is known as "contact guidance" (Teixeira, 2003; Ventre et al., 2014). This, however, depends on the size and spacing of the topographic features as too wide or too close ridges do not exert an effective confinement on FAs, thus impairing their guiding effect. For example, osteoblasts cultivated on nanogratings with ridge spacing <75 nm allowed integrin clustering over the grooves that caused the cells to respond as on flat substrates (Lamers et al., 2010). Conversely, a change in osteoblast response was observed when lateral spacing exceeded 75 nm. Cells were elongated and coaligned to the pattern direction with ridges and grooves of 80 nm in width or on patterns with 50 nm ridges and 100 nm grooves. Instead, no significant orientation was observed on the pattern with 100 nm ridges and 50 nm grooves (Lamers et al., 2010). In order to visualize the dynamic interplay between FA formation/maturation and nanogrooved patterns, Natale et al. transfected preosteoblasts with plasmids characterized by fluorescent actin and paxillin (Natale et al., 2014). Actin fibers were observed to bundle and align in directions, not necessarily parallel to the pattern direction. However, fibers oriented in directions other than that of the pattern ended with dashed adhesions that rapidly collapsed under the effect of actin-generated forces. In contrast, the FAs that grew along the direction of the pattern appeared more stable and promoted the formation of actin bundles parallel to the ridges.

The stiffness of the culturing substrate profoundly affects FA morphology and stability (Pelham and Wang, 1997). Usually, cells on rigid materials (plastics, highly crosslinked elastomers, stiff hydrogels) express long and wide FAs connected to well-defined actin bundles. Conversely, cells on compliant substrates (scarcely crosslinked elastomers or soft hydrogels) possess punctuate and highly dynamic FAs and a destabilized cytoskeleton (Ventre and Netti, 2016b). The molecular mechanisms underpinning cell recognition of stiffness are not thoroughly clear as some evidence suggests that different stiffnesses lead to differences in anchoring densities of proteins thus interfering with sensing, whereas other authors report that cells react to changes in stiffness independently from ligand anchoring (Trappmann et al., 2012; Wen et al., 2014). However, the extent of contractility generated on either soft or stiff substrates differently activates mechanotransduction events at FA or nuclear level, ultimately affecting cell functions and fate (Engler et al., 2006; Anderson et al., 2016).

Surface engineering has benefitted from these observations because the material characteristics may be modified in order to influence the dynamic formation and maturation of the FAs hence affecting the activity of the signaling proteins or altering the structure and contractility of the cytoskeleton. Today, concepts as ligand density and patterning, topography and roughness, or material stiffness are regarded as crucial players in affecting various aspects of cell behavior ad are increasingly taken into consideration when designing experiments aimed at capturing certain features of the ECM. However, as the response to soluble drugs or factors depends on their dose and temporal delivery, it is clear that the static presentation of biochemical/biophysical material signals does not fully mimic the dynamic nature of the native microenvironment and might lead to abnormal or non-physiologic reactions. This calls for novel culturing systems conceived to display biochemical/biophysical material signals according to predefined spatio-temporal programmes.

### GENERAL CONCEPTS ON DYNAMIC SUBSTRATE ENGINEERING

Synthetic materials with properties that are static in time are often inadequate for mimicking natural cell environments, where temporal changes in chemical and physical properties are important in a wide variety of contexts including development, differentiation, morphogenesis, diseases progression, wound healing, and homeostasis (Frantz et al., 2010; Lu et al., 2011; Bonnans et al., 2014; Kular et al., 2014). To assess the role of the dynamic changes of the biochemical/biophysical microenvironment on cell behavior, the development of platforms enabling a fine control on the spatio-temporal presentation of adhesion sites is key central. Recent developments in materials engineering have led to a variety of approaches in which material properties can be dynamically and reversibly modified in response to user-directed stimuli such as light, temperature, pH, or other external fields (Roy et al., 2010; Stuart et al., 2010). Concerning the fabrication of responsive biomaterials for dynamic signal presentation, different chemical/physical characteristics of the substrate, as well as those of the external stimulus, must fulfill specific requirements. Responsive materials may contain moieties responsible for endowing the materials with the dynamic behavior (Liu and Urban, 2010). These moieties do not have to harm cells whether inactive or activated by an external stimulus. Furthermore, the latter does not have to hamper cell viability by itself or does not have to cause leaching of toxic substances from the material. For instance, temperature or pH responsive materials should perform the intended transformations in a limited range of temperature or pH values (Watanabe and Okada, 1967; Kruse et al., 2017). Similarly, light can only be used at specific wavelengths and intensities (Masuma et al., 2013). Additionally, the choice of employing a particular stimulus may be dictated on whether a uniform or local material actuation is necessary. For instance, temperature and electric fields can be suitable for activating substrates uniformly, whereas laser beam can act locally. This can enable dynamic exposure to signals at a cellular or subcellular level. However, additional devices, such as UV masks or electrodes, can be employed to force the external field to act locally rather than uniformly. Alternatively, the responsive moieties could be spatially patterned thus causing the culturing platforms to be responsive in selected regions only (vide infra). Literature is sometimes ambiguous on the term static and dynamic (referred to signal presentation). We here refer to dynamic presentation when signals and more specifically adhesive signals significantly change configuration and or location in a way that cells recognize such a change and modify their behavior accordingly. The extent of change in signal display depends on the specific context that has to be analyzed as changes may occur on different length or time scales (receptor-whole cell level; seconds to weeks).

In what follows we present consolidated strategies along with the recent developments for engineering synthetic substrates displaying signals dynamically. More specifically, we will focus our attention to those signals directly affecting cell adhesion events at different length scales: from FA to whole cell level. We will therefore emphasize the role of dynamic display of biochemical (ligands), topographic and mechanical signals on cell adhesion, spreading, migration, and differentiation.

### DYNAMIC DISPLAY OF LIGANDS

Self-assembled monolayers (SAMs) are one among the first and perhaps most used and studied platforms to enable signal display dynamically. SAMs can be defined as orderly assemblies of molecules formed by the adsorption of an active surfactant on a solid surface (Ulman, 1996). The molecular structure of the individual building blocks that form SAMs can be divided in three parts: a head that binds to a substrate (generally a metal); a central part acting as spacer, whose chemical properties determine the interactions with the surrounding molecules, ultimately defining the packing and stability of the monolayer, which is usually achieved by van der Waals attractive forces between adjacent chains or by the introduction of specific intermolecular interactions; a tail group which can be inert, bioactive or susceptible to functionalization (Ulman, 1996; Srisombat et al., 2011). The composition of SAM-based platforms can be predetermined and the chemical properties can be considered uniform over relatively large areas. Furthermore, SAMs are prone to additional functionalization or patterning after they are formed (Mrksich, 2009). These are fundamental requirements to fabricate artificial platforms to study cell adhesion events or cell response to specific arrangements of ligands in a systematic manner. Examples of SAMs used as cell culture substrates are thiol or silane head groups derivatives on gold or silicon, respectively (Schreiber, 2004; Onclin et al., 2005). Oligo(ethylene glycol) SAMs were used as non-fouling protein surfaces and consequently cell repellent substrates. Alternatively, a number of biomolecules, including DNA, RGD, Lamininderived Peptide PA22-2, FN, Vascular Endothelial Growth Factor, have been used to bioactivate SAM surfaces on active groups present on the molecules tail (Bamdad, 1998; Liu et al., 2007; Mendes, 2008; Jans et al., 2009; Mrksich, 2009; Afara et al., 2012).

Another method to fabricate cell culturing substrates is polymer casting. Surfaces of cast polymer are compatible with functionalization techniques such as chemical grafting or topographic embossing (Vendra et al., 2011). Phase separation of polymers is a fast and reliable method that enables controlling the morphology and dimensions of micro- and nano- domains in the polymer, which affect surface topography and protein adsorption thus impacting on cell adhesion and eventually cell fate (Lim et al., 2005a; Krishnamoorthy et al., 2006; Frith et al., 2012).

The synthesis of polymer brushes, i.e., thin coatings made of polymer chains covalently attached to a substrate, represents a versatile method to fabricate controlled environments for cell cultures (Chen et al., 2017). Chains can be attached directly to reacting substrates (grafting-to) or might be allowed to polymerize from surface anchored initiators (grafting-from). In both cases, the density of polymers should be sufficiently high to force polymer chains to be in a stretched form, preventing chain collapse in a random coil form. Of the two methods, graftingfrom produces high density brushes as preformed polymers in the grafting-to method can be displaced apart owing to steric hindrance (Advincula, 2004). Several parameters related to polymer brush processing including, chain architecture and length, density, functionalization, all proved to significantly affect cell adhesion and behavior (Krishnamoorthy et al., 2014). The development of new synthetic routes allowed to tune these properties, thus leading to the fabrication of culturing platforms with tailored signal display (Chen et al., 2017; Feng and Huang, 2018).

Besides solid substrates, the functionalization of hydrogel surfaces has been successfully used to control local chemical/physical properties of the culturing microenvironment (DeForest and Anseth, 2012; Guvendiren and Burdick, 2013a). Both synthetic and natural hydrogels emerged as versatile and promising system for cell culture since they mimic important aspects of the native ECM such as mechanical properties, viscoelasticity, porosity, along with their ability to sequester proteins and growth factors (Caliari and Burdick, 2016). From a practical standpoint, macromolecules constituting hydrogels can be extensively modified (in pre- or post-processing) thus allowing to tailor the chemical, physical properties of the network in an orthogonal manner and to endow the substrate with complex functions (Kharkar et al., 2013; Ventre and Netti, 2016a). Hydrogels find their natural application field as 3D cell culturing systems. However, they are also useful to study certain cell behaviors in 2D such as mechanosensing and mechanotransduction that require substrate stiffness to be finely tuned (Thiele et al., 2014).

### Electrically Controlled Presentation of Ligands

Although SAMs were originally intended as static surfaces, the incorporation of moieties in the polymer chains that induce conformation changes, bond breaking or react with other species under external stimuli, endows SAM-based surfaces with dynamic behavior. In a seminal work by Jiang et al. electrical desorption of SAMs was used to fabricate dynamic surfaces to study bovine capillary endothelial cell migration (Jiang et al., 2003). Micro-printed patters of protein repellent alkanethiols [HS(CH2)11(OCH2OCH2)3OH and HS(CH2)17CH3] on gold confined cells in specifically designed rectangles. The application of a cathodic voltage pulse (−1.2 V) caused the alkanethiols to desorb. In the absence of a protein repellent layer, soluble FN in the medium readily adsorbed on the bare gold enabling the cells to move from the initial rectangle onto the newly available adhesive regions (**Figure 2**). Authors from the same research group exploited this strategy to investigate the effect of cell shape on migration and in particular on the initial direction of displacement (Jiang et al., 2005). The authors printed via microcontact printing (µCP) adhesive islets of different shapes (either symmetric circles, rectangles, squares or asymmetric teardrops, wide drops, narrow drops, triangles). By analyzing the Golgi apparatus, centromere and cell centroid positioning the authors noticed that cells had a marked tendency to displace toward the blunt end of the drop-like shapes. The authors suggested that the shape asymmetry was sufficient to bias the direction of motion. The combination of µCP and electrically switchable SAMs, makes this method a robust platform that enables studying cell dynamics without possible complications arising from cell-cell contacts. However, once SAMs desorb, adhesion patterns cannot be restored. An attempt to integrate electric signal switch and reversibility was proposed by Yeo et al. who fabricated hexadecanethiol-based SAMs on gold incorporating O-silyl hydroquinone moiety functionalized with RGD (Yeo et al., 2003). The application of an electric potential to the substrate causes the oxidation of the O-silyl hydroquinone group yielding the corresponding benzoquinone with the hydrolysis of the silyl ether and the release of the RGD (**Figure 3A**). To prove the validity of the proposed scheme, the authors micropatterned SAMs in the form of 220µm discs of hexadecanethiol on gold and the remaining regions were filled with a monolayer of RGD electroactive alkanethiolate as described above (**Figure 3B**). The system was coated with FN, which favored homogeneous adhesion of Swiss 3T3 (**Figure 3C**). The application of an electric potential of 550 mV released the RGD, thus promoting cell detachment. Only the cells on the non-electroactive patterned discs remained in place (**Figure 3D**). RGD functionalized cyclopentadiene moiety was then supplemented to the culture medium; this reacted with the benzoquinone tails by means of a Diels-Alder reaction. Microscopic observations revealed cells that invaded the newly formed RGD activated regions (**Figure 3E**). The authors showed an on-off-on pattern of signal display; this, however, required additional steps involving the introduction of specifically synthesized chemicals. A further development of this scheme involved the use of either electroactive RGD tagged quinone ester or the conventional O-silyl hydroquinone (Yeo and Mrksich, 2006). The former releases the RGD under reductive potentials, as opposed to the latter that is sensitive to oxidative potentials (**Figures 4A,B**). To prove selective release of fibroblasts from the substrate the authors micropatterned circular SAMs on gold with either of the two electroactive alkanethiolates all of them displaying RGD. Swiss 3T3 fibroblasts adhered on the micropatterned regions only (**Figure 4C**). Applying either a negative or positive potential caused selective detachment of cells from the electroactive quinone ester or Osilyl hydroquinone moieties, respectively (**Figure 4D–F**). These studies demonstrate the effectiveness of using small voltages to trigger dynamic surfaces. Furthermore, the electric potentials applied are compatible to cell culturing conditions.

A possible drawback of electric desorption of SAMs consists in the impossibility to change the electric potential locally thus achieving a spatial control on the desorption, unless networks of electrodes are assembled together (Yoon and Mofrad, 2011). A possible solution to this issue was proposed by Ng et al. who extended the concept of electrically switching surfaces (Ng et al., 2012). Rather than controlling chemical reactions with electric potentials the authors fabricated charged hexa(ethylene glycol) based SAMs on a silicon electrode. Charged moieties (either sulfonate or ammonium) were conjugated at the distal end of the SAM chains that were juxtaposed to chains terminating with GRGDS adhesive peptide. When the system was subjected to an electric potential of the same polarity as the charged moiety, then polymeric chains projected out masking the neighboring ligands. Conversely, reversing polarity caused the chain to fold back, leaving exposed and accessible the ligands. The authors also created patterns of cationic and anionic regions and were able to revert cell-adhesive/cell-repellent zones dynamically and reversibly.

### Photo-Controlled Presentation of Ligands

The use of light as a possible trigger for dynamic surfaces has attracted considerable interests since it in principle

allows to directly implement spatial patterns of "switched onoff " signals by controlling the exposure of the irradiating light with (for example) a photomask. This is a substantial advantage as opposed to electrical controlled surfaces for which case embossing adhesive patterns of signals could require the design of complex networks of electrodes or additional manipulations of the substrate with, for instance µCP.

et al. (2003). Copyright 2003 American Chemical Society.

Using UV irradiation of a conventional microscope, Nakanishi et al. were able to emboss microscale patterns on alkylsiloxane based SAMs on glass, displaying a photocleavable 2-nitrobenzyl tail group (Nakanishi et al., 2006). Coating the photocleavable SAM with bovine serum albumin (BSA) prevented cell adhesion. Conversely, exposure to a 365 nm UV radiation caused tail cleavage, BSA release and exposure of OH groups. Fibronectin solubilized in the medium adsorbed on the photocleaved regions enabling adhesion. UV exposure can be performed directly, thus activating a wide area or through a photomask. In this case, microscale features could be formed enabling to study cell dynamics at a single or even at a subcellular level. In particular, HEK293 cells on a square array of 6 µm<sup>2</sup> adhesive islets formed FAs on the patterned islets only, whereas they were able to stretch and spread across the non-adhesive areas. These data confirm that photomasking enables achieving sufficient spatial resolution to study cell adhesion at a FA level dynamically (Nakanishi et al., 2006). The same group exploited a similar technique to study cell migration and membrane extension on the dynamic surfaces (Nakanishi et al., 2007). In this case, Pluronic 108 was used as cell repellent backfill. NIH3T3 cells were first seeded on 25 × 25 µm<sup>2</sup> squares and then UV light was directed in order to form either a 25 or 5µm wide stripe protruding out the edge of the original square. In the case

of square patterns, cells spread over the entire new bioadhesive area, displaying actin bundles with FA oriented at various angles at their termini. Conversely, NIH3T3 cells displayed long protrusion along the narrow stripe with coaligned FAs and actin bundles. Photoactivated desorption of cell repellent compounds possesses the great advantage of allowing a greater level of spatial control of the adhesive properties of the substrate. Cellular and subcellular resolution can be achieved using conventional apparatuses. However, the methods illustrated so far still rely on the non-specific adsorption of adhesion moieties

cell adhesion. (C) Swiss 3T3 fibroblasts adhere to circular patterns containing either electroactive O-silyl hydroquinone or quinone ester. (D) After application of 650 mV potential, cells detached only from regions presenting the electroactive O-silyl hydroquinone. (E) The application of −650 mV potential caused cell release of cells from the regions presenting the electroactive quinone ester only. (F) The subsequent application of a potential of −650 mV (D) or 650 mV (E) results in an additional release of cells. Reprinted (adapted) with permission from Yeo and Mrksich (2006). Copyright 2006 American Chemical Society.

from the medium which could be susceptible to cell-mediated remodeling.

In order to control cell-material adhesion at the molecular level directly and irrespective of influences arising from material properties Petersen et al. exploited a photosensitive ligand caging strategy (Petersen et al., 2008). The authors synthesized a cyclo(-Arg-Gly-Asp-D-Phe-Val) c-(RGDfK), which shows improved affinity for the αVβ<sup>3</sup> integrins, protected with a photolabile 3-(4,5-dimethoxy-2-nitrophenyl)-2-butyl ester attached on the Asp residue of the ligand. A tetra(ethylene glycol) linker was anchored to a silica surface from the one side, whereas the caged ligand was on the tail side. 3T3 fibroblasts were seeded on the substrate before and after irradiation with UV at 351 nm. The authors found Cimmino et al. Spatio-Temporal Control of Cell Adhesion

more than 10-fold increase in adhered cells on substrates exhibiting the ligand in the active form. Furthermore, the authors used photomasks to activate the substrate on selected regions and, as expected, cells adhered on the exposed ligands predominantly. However, a degree of non-specific adhesion was observed on non-irradiated areas especially at longer culturing times.

Light irradiation may also drive conformation changes of sensitive moieties. Displacement caused by the change in molecular conformation can be exploited to alter the chemical-physical properties of material surfaces. Spiropyran or azobenzene are two molecules that were extensively used for this specific purpose (Klajn, 2014; Fedele et al., 2018; Landry et al., 2018). In the context of cell adhesion, Higuchi et al. coated glass plates with a copolymer of nitrobenzospiropyran and methyl methacrylate (Higuchi et al., 2004). This copolymer can be subjected to the reversible transformation from a hydrophobic spiro conformation to a hydrophilic zwitterionic merocyanine isomer by means of irradiation of UV light. The authors first demonstrated the reversibility of the transition (which required at least 24 h in a dark environment) and then proved KUSA-1 cells detachment after a 4-min irradiation with light. Additionally, the authors also reported fibrinogen and platelet detachment. Edahiro et al. synthesized a photoresponsive cell culturing surface composed of poly(N-isopropylacrylamide) with nitrospiropyran as actuating element attached as side chain on the supporting polymer (Edahiro et al., 2005) (**Figure 5**). The authors showed that CHO-K1 cells persisted on the UV irradiated regions of the material, whereas a low temperature washing detached cells. The system proved to be reversible as irradiation with visible light at 400– 440 nm released the cells immobilized on the preceding UV exposure. While this system is sufficiently versatile enabling spatial patterns of cell adhesion sites, it requires additional steps of low temperature cell washings to achieve a complete reversibility.

The light induced cis-trans isomerization of azobenzene derivatives has been extensively used into biopolymers to change their structure in a controlled manner (Zhang et al., 2017). Exploiting this peculiar characteristic Auernheimer et al. were able to change distance and orientation of RGD ligands thus affecting cell attachment on poly(methyl methacrylate) (Auernheimer et al., 2005). The authors used 4- [(4-aminophenyl)azo]benzocarbonyl photoswitch as actuating element of spacers with different lengths containing an acrylamide anchor and a cyclic RGD tail (**Figure 6A**). Improved adhesion of MC3T3-E1 preosteoblasts were observed for all the spacers used when the azobenzene was in the trans configuration (i.e., after irradiation at 450 nm). Conversely, a general decrease in adhesion was observed when the azobenzene was in the cis state, which caused spacer shortening and decreased accessibility to the ligand (**Figure 6B**).

Similar to the approach of Auernheimer et al., Liu et al. designed and fabricated SAMs an Au containing alkanethiols with azobenzene groups and functionalized with GRGDS (Liu et al., 2009). Changing the conformation of the azocompound through light irradiation caused the ligand to be either exposed

nitrospiropyran chromophore sidechains. Below, photoisomerization scheme of the nitrospiropyran chromophore. Reprinted with permission from Edahiro et al. (2005). Copyright 2005 American Chemical Society.

on the surface (in the trans configuration) or hidden in the SAM (in the cis configuration). The authors demonstrated that the SAM reversibly enabled/impaired adhesion of NIH3T3 in a time-frame of few hours. Also, in this case, changes in the orientation and conformation occurring at the molecular level are sufficient to alter cell adhesion. Adhering cells could be detached by supplementing the culture medium with soluble GRGDS at pH 8.0. In this condition, the soluble ligand competes with the bound one. After detachment azomolecules could be switched to the cis isomer thus hiding the ligand and reducing cell attachment consequently. After irradiating with visible light, azomolecules reverted to the trans form, which once again enabled cell adhesion. In this case reversibility is subjected to additional steps therefore this method cannot be considered as fully reversible. RGD ligand burying into a polyelectrolyte multilayer (PEM) was exploited to reversibly modulate NIH3T3 cells adhesion with UV light irradiation (Goulet-Hanssens et al., 2012). Differently from solid substrates, PEMs enable modifying the stiffness according to the processing conditions such as layer numbers and pH.

Changes in azomolecule conformation were also exploited to modulate adhesion on SAMs through a host-guest approach (Gong et al., 2011). Gong et al. synthesized α-cyclodextrins on alkanesilane molecules to form SAMs on SiO2. This template enabled the formation of inclusion complexes with azobenzene-GRGDS via host-guest recognition. Stable

irradiated for 3 h at 450 nm (dark gray column) and overnight at 366 nm (light gray column), respectively. Arrow indicates comparable level of cell adhesion with respect to the uncoated material (white column). Adapted with permission from Auernheimer et al. (2005). Copyright 2005 American Chemical Society.

complexes were formed when the azomolecule was in the trans configuration allowing ligand display. In this setup HeLa cells effectively adhered and spread on the substrates. However, upon irradiation with UV at 365 nm for 10 min (a condition not harming the cells) caused an extensive cell detachment.

### Temperature Controlled Presentation of Ligands

The ability of certain polymers to significantly change conformation and packing with temperature was exploited to modulate polymer-protein interactions and hence cell adhesion to the polymer layer. One of the most effective temperature responsive polymeric systems for cell cultures is the poly(N-isopropylacrylamide) (PNIPAM) showing a lower critical solution temperature (LCST) around 32◦C, a temperature that does not alter cell viability significantly. Okano's group pioneered the use of PNIPAM to revert the adhesion of confluent cell layers enabling their detachment from culturing plates (Yamada et al., 1990). This represents a crucial step for harvesting cell layers for cell sheet engineering processes. Briefly, electron beam grafted PNIPAM onto conventional PS dishes allows cell adhesion and proliferation at temperatures above the LCST. In this case, PNIPAM is stabilized by hydrophobic interactions that exclude water from the polymer network (**Figure 7A**). Fibronectin firmly adheres to the hydrophobic network promoting cell adhesion. When cooled below the LCST, interchain H-bonds form allowing network swelling with the inflow of water. This conformational change causes the desorption of FN and subsequent detachment of the cells. The system can be reversibly made hydrophobic by re-heating the network above the LCST (**Figure 7B**). Such an elegant method proved to be effective in detaching large confluent cell sheets without applying mechanical stresses possibly harmful for the cells (Tang and Okano, 2014). This enables the manipulation of macroscopic layers for cell sheet engineering applications (**Figure 7C**). The same group demonstrated a similar mechanism of action by conjugating RGD on temperature responsive P(NIPAMco-2-carboxyisoprpylacrylamide) (Ebara et al., 2004). Cell adhesion, spreading and proliferation were observed above the LCST. Below this threshold, however, network swelling caused mechanical disruption of RGD-receptor couplings along with shielding RGD sequences from integrin engagement. Generally, changes in temperature cannot be focused in selected part of the samples, therefore local actuation cannot be achieved straightaway. However, conjugating signal patterning and temperature responsive polymers partly solved this issue (Williams et al., 2011).

### Enzyme and Mechanical Control of Ligand Presentation

Cells contribute with external stimuli and insults to reconfigure ECM structure dynamically. Common strategies cells pursue to remodel the environment include the secretion of enzymes cleaving ECM proteins and the application of mechanical forces (Janmey and Miller, 2011; Bonnans et al., 2014). Enzymes are characterized by a high selectivity and work in mild, physiologic conditions, making them ideal candidates to trigger the activity of responsive biomaterials. Using the approach of exposing adhesion sites with proteolytic enzymes, Todd et al. exploited the cleaving activity of chymotrypsin to convert a cell repellent polyethylene glycol (PEG)-based polymer film into a bioactive support (Todd et al., 2007). Briefly, a PEGacrylamide film was spin coated onto an epoxy-functionalized glass. RGD was attached onto the film and capped with a large 9-fluorenylmethoxycarbonylphenylalanine group. The exposure of the film to chymotrypsin caused the enzyme to selectively remove the capping group making the RGD ligand available for

integrin engagement and cell adhesion. This work demonstrated the feasibility of exploiting the biological activity of molecules to activate dynamic platforms, paving the way toward the development of systems activated on-demand by cell secreted molecules. However, while this system can be conveniently classified as dynamic, it is non-reversible since once unprotected by the enzyme activity the ligands cannot be capped reversibly.

More recently, Roberts et al. exploited enzyme switchable substrates to investigate and manage the delicate adhesion/cytoskeleton balance regulating quiescence or differentiation of human multipotent mesenchymal stromal cells (hMSCs) (Roberts et al., 2016). Glass substrates were functionalized with multifunctional chains constituted by a PEG linker, an RGD peptide, a dialanine cleavable site and either a fluorenylmethyloxycarbonyl blocking group or a PEG adhesion reducing moiety at the distal end. The introduction of elastase cleaves the dialanine site exposing the RGD that allows the transition from a "low" to a "high" adhesion state. hMSCs on platforms in a "low" configuration express β<sup>1</sup> rich FAs and negligible bone morphogenetic protein (BMP) receptor colocalization. Conversely, hMSCs on enzyme activated substrates expressed β<sup>5</sup> rich FAs and increased BMP signaling. Altogether, these data suggest that low (but not null) adhesion prevents metabolome activation for osteogenesis, whereas high adhesion increases BMP signaling through increased levels of intracellular tension. This work demonstrated how delicate the adhesion/tension balance is and subtle perturbations can result in differentiation of stemness maintenance. Only advanced dynamic platforms as the one here described can provide a reliable testing platform to unravel the complex mechanisms underpinning fate decisions.

Several natural biomolecules change their affinity with other molecules or functions when subjected to mechanical stretching (Sawada et al., 2006; del Rio et al., 2009; Singh et al., 2010; Janoštiak et al., 2014). Exploiting such a biomimetic route Bacharouche et al. fabricated an elastomer-based device in polydimethylsiloxane (PDMS) whose adhesion properties changed dynamically and reversibly according to the device deformation (Bacharouche et al., 2013). The authors grafted in sequence PEG chains under elastomer stretching and then RGD was conjugated to chains. In a relaxed state the ligand resulted embedded in the PEG brushes. When the elastomer was subjected to stretching, RGD became exposed and accessible. Human osteoprogenitor cells firmly adhered to the PDMS in the stretched form and expressed talin rich and well-defined FAs. In the relaxed state cells became round showing a diffuse talin staining all along the cortical region. This suggests that integrin disengagement from the concealing ligands destabilize FAs thus leading to cell detachment. Surface deformation is also responsible for the redistribution of material elements. For example, uniaxial deformation causes material elongation along the stretching direction and a contraction in the orthogonal direction. Following this observation Deng et al. investigated how cells reacted to continuously and reversibly variable interligand spacings (Deng et al., 2017). They transferred a quasi-hexagonal pattern of RGD functionalized gold nanoparticles spaced 35 nm apart (a spacing permissive for adhesion formation) on a highly stretchable poly(N-acryloyl glycinamide) hydrogel. Macroscopic stretching causing a ligand separation above 70 nm impaired the formation of stable adhesions in MC3T3 cells even if the spacing in the orthogonal direction was below 70 nm. Reversing stretching-relaxation forced the REF 72 cells to first acquire a polarized shape and a highly motile behavior and then (after relaxation) to display a spread shape and more stationary behavior. Therefore, not only cells show an extraordinary sensitivity to nanoscale spacing of ligands, but also they finely discriminate spatial variations in the distribution.

### Cell-Mediated Remodeling of Ligand

The platforms illustrated so far enable a dynamic display of signals, but do not allow a cell mediated, spatio-temporal redistribution of these. This becomes a critical issue for investigating biologic phenomena such as cell remodeling of ECM or for assessing the role of receptor clustering in signal transduction. To address these issues, platforms based on lipid mono or bilayers proved to be particularly effective. While these may represent the ideal systems to study the dynamics of cell-cell adhesions, various chemical strategies have been developed to functionalize lipidic membranes with matricellular cues (Gooding et al., 2014; Koçer and Jonkheijm, 2018). Molecules incorporated in the layer are endowed with a lateral mobility, whose dynamics is affected by the chemical composition of the layer. Concerning lipid bilayers, those usually employed as culturing substrates are made of 1,2-dicoleoyl-snglycero-3-phosphocholine (DOPC) that possess a low gel-fluid transition temperature, indicating a higher lateral molecular diffusion. Conversely, membranes of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) are more thermally stable with less diffusible molecules (Lindblom et al., 2006). Culturing substrates based on bilayers are conveniently formed on hydrophilic surfaces (supported lipid bilayers, SLBs). SLBs provide a good level of spatial control of ligand positioning when physical constraints are introduced on the nanostructured support on which the membranes are formed (Groves, 1997). In this case, nanometric relieves of the substrate locally impair lateral mobility of molecules, thus enabling clustering. Following this approach Yu et al. investigated the early adhesion events and lateral clustering of integrins (Yu et al., 2011). RGD functionalized DOPC membranes were formed on a glass substrate containing 5 × 100 nm metal wires spaced by 0.5–4µm gaps. Thanks to lateral diffusion and capture provided by the nano-barriers, the authors observed integrin recruitment and the formation of submicron clusters in which adhesion proteins such as talin, paxillin, and FAK were formed in a contraction-independent manner. This preceded actin filament assembly at cluster sites, for which, myosin contraction produced even longer clusters. Unraveling such a complex dynamics required the versatility and fluidity of SLB as opposed to conventional ligand-immobilized substrates. Along this line, Vafaei et al. have recently compared the biological activity of FN and Collagen type I covalently anchored to DOPC SLBs with that of proteins non-specifically adsorbed onto SiO<sup>2</sup> substrates (Vafaei et al., 2017). The authors found an increased flexibility of the proteins and more efficient cell adhesion and proliferation, providing further evidence that SLBs may represent a more biomimetic microenvironment to study certain biological processes, respect to synthetic platforms not allowing ligand displacement.

SLBs have also been employed to address the role of ligand clustering on stem cell fate decision. Koçer and Jonkhejim correlated hMSCs spreading with receptor clustering and cell differentiation (Koçer and Jonkheijm, 2017). To modulate lateral mobility either DOPC or DPPC membranes functionalized with RGD were used. hMSCs showed higher adhesion, higher level of expression of osteogenic differentiation markers and calcium deposits on DOPC membranes suggesting a positive role of ligand clustering and integrin activation in dictating stem cell fate decisions. Additionally, SLBs proved to be very stable as receptor recruitment and differentiation occurred on very different timescales (minutes vs. days) and yet they allowed cell culture for up to 10 days.

### DYNAMIC TOPOGRAPHIES

Static topographic relieves in the form of micro/nano pits, protrusions or channels dramatically affect various aspects of cell behavior including, migration, proliferation and differentiation (Dalby et al., 2014; Ventre and Netti, 2016b). Several technologies have been developed to perform systematic studies of response to topographic signals of a sufficiently large number of cells. Among these, replica molding, nanoimprint lithography, block copolymer micelle nanolithography exhibit a reasonable balance between features fidelity and resolution and large area patterning (Ventre et al., 2018). Extending the studies in a dynamic framework requires the development of stimuli responsive materials, along with manipulating/actuating strategies that must be both effective and cell compatible. The examples of surfaces displaying adhesion signals dynamically all share the common trait that the switches act on a molecular level (capping, conformation changes, desorption). For the successful implementation of dynamic topographies, i.e., relieves changing shape and height, the coordinated motion or matter removal on a submicro- or micro-scale level is necessary. Specific chemical strategies need to be developed for allowing this type of material transformation as well as the triggering stimulus must be effective on a reasonable time scale.

## Mechanical Control of Topographic Patterns

An early example of dynamically changing topographic pattern was proposed by Zhu et al. who fabricated a thin silicalike layer on a PDMS elastomer with radiofrequency oxygen plasma (Zhu et al., 2005). The silica layer was rendered protein repellent via chemical vapor deposition of (tridecafluoro-1,1,2,2 tetrahydrooctyl)-1-trichlorosilane and incubation with Pluronic F108. Cyclic stretching of the elastomer caused the thin brittle layer to form parallel arrays of cracks (having width ranging in the 0.1–3µm interval) on which C2C12 myoblasts adhered and elongated. Cells subjected to cyclic opening-closing of the cracks responded with sequential changes in elongation/retraction. A similar approach was reported by Lam et al. who fabricated patterns of periodic microscale waves by compression-induced buckling of a brittle thin film on the surface of a PDMS elastomeric substrate (Lam et al., 2008). C2C12 cells adhered and retraced the contours of the wavy pattern when the elastomer was compressed, whereas cells displayed a random orientation when the PDMS strain was released. These are elegant examples illustrating the fabrication of dynamic substrates in a simple and cost-effective manner. However, with these strategies it is not possible to achieve a fine control on the features of the topographic patterns. Also, the geometry of the pattern is limited to straight channels and cannot be changed readily.

Guvendiren and Burdick further elaborated this concept by developing a method to spatially control the geometrical features of the strain responsive topographies (Guvendiren and Burdick, 2013b). Exposing to ultraviolet/ozone uniaxially or biaxially stretched PDMS sheets resulting in the formation of a stiffer outer skin. Selective release of the strain produced a wave pattern perpendicular to the strain direction. Release of both directions of strains in the biaxially stretched sheets resulted in a labyrinthine pattern. Also, local masking irradiation enabled sitespecific patterning. hMSCs cultivated on the dynamic patterns responded to the switching by altering shape, orientation and nuclear area. Although very effective in the short culturing periods, the effectiveness of the dynamic topographies decreased in time as cell proliferation and cell-cell contacts overruled the topographic guidance.

### Light Responsive Topographies

Material degradation induced by light irradiation enables embossing topographic structures with a higher spatial control of the features with respect to the methods described above. Kirschner and Anseth synthesized a photodegradable PEG based hydrogel by coupling nitrobenzyl-based photodegradable acrylate to PEG-bis-amine and PEG macromer (Kirschner and Anseth, 2013). Irradiating the gel with UV light at 365 nm caused photolabile bond breaking and gel erosion. Using photomasks, the authors were able to direct erosion giving rise to topographic patterns whose depth was proportional to the irradiation time. Furthermore, by directing the laser beam of a multiphoton microscope multiple patterns were embossed on the same gel. Dynamic patterning in presence of cells was demonstrated by cultivating hMSCs on the top of a FN coated photolabile hydrogel. Microscale channels were imprinted on the gel which caused the cells to reorient, eventually increasing their aspect ratio. Afterwards, a grating of regular squares was produced by drawing lines orthogonal to the primary pattern. This restored a symmetric condition that induced cells to acquire a round shape. This platform proved to be highly versatile as it enabled writing multiple patterns with different features on the same substrate. Additionally, the patterning procedure along with the hydrogel degradation products does not harm cell. However, once sculpted hydrogel surfaces cannot acquire the initial shape; this may pose severe limitation on the reversibility of the system.

Thanks to their ability to change conformation upon light irradiation, azopolymers have been used in various biological applications (Beharry and Woolley, 2011; Wei et al., 2015; Goulet-Hanssens et al., 2016). One of the first examples of using azopolymers as cell culturing substrates was provided by Baac et al. (2004). Laser holography was used to imprint undulated nanoscale patterns on the commercially available light responsive azopolymer poly[(methylmethacrylate)-co-(Disperse Red 1 acrylate)]. Azopolymeric patterned films proved to be biocompatible enabling the attachment of different cell types. This paved the way to use patterned light to modify surface topography of films dynamically. An early attempt to inscribe a topographic pattern on an azopolymeric cell culturing substrate while preserving cell viability was reported by Barillé et al. (2011). The authors reported pattern inscription via irradiation with a two-laser beam interferometric pattern or through molecular self-organization induced by a single beam. PC12 cells reacted to the flat-grooved transition by elongating and aligning along the pattern direction and extending neurites. Although preserving cell viability, an influence of the liquid medium on the process of pattern inscription was reported. The issue of azopolymer stability in biologic media was addressed by Rocha et al. who noticed material reorganization in aqueous environments that varied according to the polarity and stiffness of the azopolymer (Rocha et al., 2014).

More recently Rianna et al. investigated the feasibility of employing Poly(Disperse Red 1 methacrylate) (pDR1m) as a suitable substrate to emboss topographic patterns reversibly for cell culture experiments (Rianna et al., 2015). Preliminary tests to assess pattern stability under conditions comparable to those experienced during cell culture, were performed. Patterns in the form of microgrooves or microgrids were inscribed on films by using an interference pattern of light. Cells were mostly round when cultivated on flat or grid patterns, whereas they appeared to be highly elongated and aligned along the direction of linear microgrooved patterns (**Figures 8A,B,D**). Circularly polarized light was used to induce pattern erasure on pDR1 films, which caused a sharp decrease in the elongation of cells once cultivated on the erased pattern (**Figures 8C,D**). FA length did not display changes in the writing/erasing cycles, whereas FA orientation was very sensitive to the topography as parallel FAs were observed on the SRG only (**Figure 8E**). Authors from the same group further elaborated this concept by integrating pDR1 patterned with a single photon laser with the aim of developing dynamic substrates to study how spatio-temporal variations of topographic patterns affect the behavior of vital

cells in real-time (Rianna et al., 2016). A conventional confocal microscope was used to guide laser beam path to emboss complex patterns on pDR1m thin films with submicrometric resolution. Fibroblasts cells responded to the dynamic patterns by altering their morphology and migration because of cytoskeleton and focal adhesion reorganization. Furthermore, irradiating the substrate with an incoherent and unpolarized light of a mercury lamp enabled the erasure of topographic patterns on cell-populated pDR1m. A further peculiarity of the confocal technique consists in enabling multiple pattern inscriptions on the same platform simultaneously, so that isolated cells can be exposed to different patterns in real-time (**Figures 9A–C**). The possibility to imprint and erase several topographic patterns on azopolymer films with an easy and versatile method may pave the way to an investigation of complex processes involved in cell material-crosstalk (**Figures 9D–G**). Rossano et al. have recently exploited this technique to study NIH3T3 fibroblast response to dynamic circular topographic patterns (Rossano et al., 2018). The authors found that cells reacted to transition from flat to circular topographies quickly and that patterns of circular ridges 2µm wide and spaced of 10µm most effectively affected cell shape and local orientation. Since stress fibers and FAs cannot grow in a bent fashion, circular topographies proved very effective in destabilizing cytoskeletal structures and hence in decreasing the mechanical properties of cells.

Koçer et al. integrated microscale topographic cues with dynamic and reversible surface nanoroughness by exploiting the light induced azobenzene isomerization within liquid crystal networks (Koçer et al., 2017). More specifically, the polymer network containing the azobenzene moiety was photocrosslinked in a network whose microstructure was in the cholesteric phase. UV irradiation decreased the order of the phase causing an increase in surface nanoroughness, from 9.0 ± 1.2 to 11.1 ± 1.2 nm. NIH3T3 cells cultivated on the dynamic surfaces dynamically changed the migration behavior from motile to stationary according to the underlying topography.

### Temperature Responsive Topographies

Biocompatible shape memory polymers (SMPs) enabled to investigate cell recognition and response to time changing topographies by exploiting consolidated and inexpensive technologies such as casting and hot embossing (Ratna and Karger-Kocsis, 2008; Meier et al., 2015). Generally, SMPs can recover permanent shapes upon stimulation of temporary ones. The latter can be implemented by means of mechanical deformation, whereas the application of an external stimulus, including heat, light, or solvent exposure promotes the recovery of the permanent shape (Mather et al., 2009). SMPs possess the advantage of retaining microscale topographic features, with high fidelity and—if conveniently engineered—SMPs can

20µm. Adapted from Rianna et al. (2016). Copyright Wiley-VCH Verlag GmbH and Co. KGaA. Reproduced with permission.

switch under cytocompatible conditions (Le et al., 2011). One of these materials is polycaprolactone (PCL) that is able to change topography if heated above the transition temperature (Ttrans) of 40◦C (Le et al., 2011; Ebara et al., 2012). In one example, Le et al. crosslinked star shaped PCL liquid prepolymer under UV light in presence of diethoxyacetophenone photoinitiator into specifically designed molds, thus allowing the material to acquire the equilibrium primary topography. Mechanical forces were applied above Ttrans to impress the transient, secondary topography, and cooled down the Ttrans. Reduction of the molecular mobility prevented the material to relieve the mechanical stress, thus enabling the secondary topography to remain impressed. Heating above Ttrans provided sufficient mobility to the polymers to restore the original shape. The authors exploited this to investigate the response of hMSCs to dynamically changing microtopographies. Cells seeded on temporary 3:5µm grooved pattern elongated and aligned along the pattern direction. However, when the substrate was heated above Ttrans the primary-planar shape emerged, causing the cell to acquire a stellate morphology. While this method is simple (not requiring expensive device to actuate the transition) and effective, it is not truly reversible as secondary shapes cannot be reacquired. Furthermore, cells on secondary topographies need to be cultivated below the Ttrans (28◦C) which might be an issue for long-term cultures. Beside adhesion and orientation, patterned SMPs were shown to dynamically regulate the structural and contractile properties of cardiomyocytes (Mengsteab et al., 2016). Neonatal rat ventricular myocytes (NRVMs) were cultivated on a PCL-based SMP exhibiting a primary nanogrooved pattern (400 nm wide ridges; 500 nm

grooves; 150 nm depth) and a secondary transient nanogrooved pattern with same dimensions but rotated of 90◦ with respect to the primary one. This design was aimed at recapitulating the orientation of collagen fibrils in the myocardium. NRVMs on the transient nanopattern exhibited a homogeneous population of elongated FAs, nuclei alignment along the pattern direction and a unidirectional and a homogeneous contractile behavior. The surfacing of the primary pattern determined an increased dispersion of FA orientations and after 48 h cells showed a bimodal distribution of the directions of contraction. These data suggest that nanotopographic patterns do not dictate organization and contractile properties permanently making dynamic topographies a suitable signal to alter the organization of cells on a collective level.

Another class of polymers exhibiting shape memory are the crosslinked polyurethane-based adhesives, commercially available under the name NOA63. Davis et al. processed NOA63-based films displaying dynamic topographies. A flat surface was recovered from a microgrooved pattern by changing the temperature from 30 to 37◦C (Davis et al., 2011). Such a temperature preserved cell viability. Mouse embryonic fibroblasts accompanied the transition by remodeling their cytoskeleton and changing their shape accordingly. A drawback of this method relies in the long time-frame required for the topographic switch to occur (∼hours).

## DYNAMIC STIFFNESS

Materials whose mechanical properties can be changed dynamically represent unique platforms to investigate complex biological phenomena such as morphogenesis and wound healing. In fact, the mechanical properties of the culturing microenvironment—and more specifically its stiffness—are known to potently regulate cell differentiation (Engler et al., 2006; Ivanovska et al., 2015). However, how changes in ECM mechanics accompany or anticipate physiological or pathological states, are not thoroughly understood and artificial platforms recapitulating the dynamics of stiffness changes in vitro would be of great importance for studying the development of physiopathological states. ECM stiffening is a typical phenomenon accompanying several biologic processes such as in vivo organogenesis or the progression of pathologies (Berry et al., 2006; Georges et al., 2007; Krieg et al., 2008). Dynamic changes of substrate stiffness usually rely on degradation or crosslinking of hydrogels that can either proceed spontaneously or can be activated by external triggers such as light.

By exploiting photodegradation, Kloxin et al. aimed at addressing myofibroblastic differentiation in response to microenvironment softening (Kloxin et al., 2010). Fibroblast to myofibroblast differentiation is a crucial step in wound healing. Wounds are highly dynamic environments (in terms of biochemical microenvironment, architecture and mechanical properties) and today the role of biochemical/biophysical signals in governing the dynamics of healing is poorly understood. Dynamic platforms may provide valuable insights into the biology of regenerative/reparative processes. To this aim, valvular interstitial cells were cultivated on a PEG based hydrogel containing photodegradable crosslinkers. The photodegradable hydrogel allowed to change moduli in the physio-pathologic range within minutes after irradiation with UV light. Cells cultivated for 5 days on stiff (32 kPa) hydrogels displayed the characteristic phenotype of activated myofibroblasts, whereas cells cultivated on soft (7 kPa) gels were in quiescent form. However, the stiff to soft transition (3 days on stiff, followed by 2 days on soft) was sufficient to deactivate cells.

Not only the absolute magnitude of the stiffness affects fate decisions, but the way the mechanical stimulus is presented also plays an important role in regulating the differentiation process. Along these lines, Yang et al. used a photodegradable polyethylene glycol di-photodegradable acrylate hydrogel functionalized with RGD to investigate whether stem cells possess a "mechanical memory", i.e., stem cells make fate decisions based on the temporal integration of the perceived dynamic mechanical stimuli (Yang et al., 2014). To this aim, hMSCs were subjected to a mechanical dosing, i.e., they were cultivated on stiff substrates for different time intervals (0, 1, 5, or 10 days) and then the substrate was softened down to 2 kPa. hMSCs supplemented with a small dose of mechanical stimulus (1 day) exhibited transient levels of genes involved in osteogenesis. Conversely, an irreversible activation of the differentiation programme was observed when cells were cultivated for 10 days on stiff hydrogels prior to softening. The authors suggested that stem cells do possess a mechanical memory for which Yes-associated proteins/Transcriptional coactivators with PDZ-binding motif (YAP/TAZ) acts as mechanical rheostats since they persist in the nucleus, hence promoting osteogenesis, if cells are continually exposed to stiff environments.

In the context of addressing the role of dynamic changes of material stiffness on stem cell differentiation, Guvendiren and Burdick developed a methacrylated hyaluronic acid (HA) based gel whose stiffness can be increased through light-mediated radical polymerization (Guvendiren and Burdick, 2012). hMSCs were cultivated on soft gels for selected time intervals, after which gels were stiffened by light irradiation. Cells quickly reacted to material stiffening by increasing area and the expression of contractile forces. Furthermore, early stiffening (i.e., cells exposed to a stiff environment for a long timeframe) promoted osteogenesis, whereas late stiffening promoted adipogenesis. Besides differentiation, the timing of the materials stiffening proved to be a crucial parameter for activating pathological states in in vitro models (Caliari et al., 2016). Using hepatic stellate cells cultivated on a methacrylated HA gel sensitive to blue light irradiation (∼470 nm), Caliari et al. showed that a soft-to-stiff transition occurring after 6 days prompted an accelerated signaling kinetics of both YAP/TAZ and alphasmooth muscle actin whose time-course was similar to observed in vivo activation dynamics. Anticipating the transition at day 1 did not produce the same trend.

These examples suggest that the parameter stiffness has not to be intended as a "state variable" for which the cell response only depends on the initial and final values of the stiffness, being independent from the specific pathway that led to the change. The systems described above operate on step-wise changes of the stiffness occurring in short-timeframes (minuteshours). Cells are exposed to continuous changes of the in vivo biophysical microenvironment (Montell, 2008; Bonnans et al., 2014). Therefore, it is likely that the kinetics of stiffness variations may also play a role in affecting cell functions and fates. To address this, it is necessary to develop chemical modifications producing crosslinks or bond breakings whose reaction kinetic eventually match the time-variations observed in vivo. To verify whether time-dependent stiffening plays a role in establishing an adequate microenvironment for myocardial maturation Young and Engler synthesized a thiolated-hyaluronic acid hydrogel crosslinked with (ethylene glycol)diacrylate whose crosslinking kinetics and hence stiffening dynamics was tuned to match the temporal variations of stiffness observed in developing embryos (Young and Engler, 2011). Precardiac cells cultivated on collagen coated dynamic thiolated-HA gels showed significantly higher expression of mature cardiac markers and improved assembly of contractile fibers with respect to what observed when cells were plated on conventional and static polyacrylamide gels. Even though this experimental model does not allow to change mechanical properties at will, it clearly demonstrates how undifferentiated cells are sensitive to biophysical changes of the microenvironment.

More recently, hydrogels with reversible mechanical properties were fabricated through modular modifications of HA (Rosales et al., 2017). Three different moieties were used: o-nitrobenzyl acrylates (photodegradable), methacrylates (crosslinks with a photoinitiator) and RGD. Hydrogels could be softened and then hardened (or vice versa) through sequential irradiation with specific wavelengths. MSC cultivated on the HA modified gels showed a decrease in the cell area and nuclear localization of YAP/TAZ following in situ softening (from 14 to 3.5 kPa), whereas subsequent stiffening (from 3.5 to 28 kPa) increased the cell area and nuclear localization of YAP/TAZ. Authors of the same groups further elaborated the design of the photoresponsive system and successfully produced a supramolecular hydrogel constituted by hyaluronic acid functionalized with the photoresponsive guest-host pair azobenzene and β-cyclodextrin (Rosales et al., 2018). Azobenzene groups in the trans configuration form links with cyclodextrins that can be disrupted upon UV irradiation that promotes the cis conversion under cytocompatible conditions. Such a design possesses the advantage of decoupling changes in network mechanical properties from changes in chemical composition. This endows the system with well-defined physical/chemical characteristics suitable for studies in mechanobiology.

### CONCLUSIONS AND FUTURE PERSPECTIVES

Adhesion mediated signaling regulates various aspects of cell behavior including proliferation, migration, and differentiation. Cell adhesions have a direct impact on cytoskeleton arrangement, cell contractility, and nucleus shape, which altogether affect gene expression and eventually cell functions (Dalby et al., 2014; Murphy et al., 2014; Anderson et al., 2016; Ventre et al., 2018). Therefore, adhesion—cytoskeleton—contractility constitute a tripartite module in which changes in one element invariably induce alterations in the other thus affecting cell behavior. In this context material signals regulating cell adhesion acquire a central role, perhaps as important as chemical supplements or drugs. Culturing substrates, scaffolds or gels proved remarkable effectiveness in finely modulating even complex biological events. Yet, these culturing systems are mostly known for displaying signal in a static manner. Static signal platforms have provided tremendous insights in understanding the optimal configurations of signals to be displayed to elicit a specific response. However, in the way they are conceived static platforms cannot be employed to address other relevant questions. As soluble biochemical signals elicit very different cellular responses according to their dose and on the way they are delivered in time, do biophysical signals behave similarly? How does the variable spatio-temporal presentation of biophysical signals affect cell behavior? Does an optimal spatio-temporal presentation of bound signal to elicit a specific cell response in the most effective manner exist? Answering these questions may not only provide valuable insights for developing more effective and versatile culturing systems but might aid in shedding light on intricate biological processes involved whenever modifications in the microenvironment occur, for example in tissue and organ morphogenesis or in the development or progress of diseases. Dynamic platforms may certainly contribute to address these issues. However, many technical hurdles need to be overcome. First, a fine modulation of adhesion events and exerting a control on the dynamics of specific arrangements of FAs is necessary. Such a spatio-temporal resolution in the actuating elements of dynamic platforms has not been achieved yet. Second, dynamic surfaces have reached a level of performances that enables controlling adhesion processes at a subcellular level in timescales that are compatible to conventional cell cultures. However, different biological responses may occur on different time scales, i.e., adhesion takes minutes to hours, whereas differentiation occurs in days or weeks. Achieving a proper time control is still an issue. Some triggers are faster than others, but once they are turned "on" or "off " the whole system responds with its own kinetics. Therefore, tuning the dynamics of the signal change to that of a specific biological phenomenon may require a careful design of the molecular switch or the integration of various switches. Third, while 2D surfaces are valuable to understand the behavior of epithelia and endothelia as they more closely mimic the native microenvironment, certain biological events specifically occur in a 3D context. Patterning signals in space with a submicron resolution is technically challenging and requires expensive equipment. Actuating switchable signals in 3D is even more complex. Laser two photon or holographic techniques might be valuable instruments for achieving a spatial control of the chemical/physical properties of the 3D environment with an adequate resolution (Applegate et al., 2015; Tong et al., 2016). Notable examples of dynamic 3D environments able to regulate cell functions and fate decisions have been already reported (Khetan et al., 2013; Das et al., 2016; Brown et al., 2018). However, integrating diverse stimuli in 3D and

achieving a control on individual adhesion points still require further developments. Moreover, activating and deactivating signal exposure in a completely reversible manner is key central. Non-reversible or partially reversible systems might only partly recapitulate the dynamic behavior of the ECM or might not be entirely versatile. Creating complex spatio-temporal patterns of signals requires implementing signal display/conceal in a fast and effective manner. The uncontrolled presence of signal remnants may interfere with the intended signal presentation providing inconsistent results.

Literature data has provided compelling evidence that cell functions and fate are not dictated by a single signal type, but it is rather a complex interplay of multiple signals acting on different length- and time-scales that determines the final cell state (Dalby et al., 2014; Murphy et al., 2014). To actuate different signal acting at different timepoints, specific switches must be engineered. Photoswitches might be particularly suitable for this purpose. In fact, several molecules have been developed, for example the azoderivatives, whose response may fall within a broad range of wavelengths. In principle, by combining different light sensitive switches reacting to different wavelengths it would be possible to fabricate arrays of signals whose display could be controlled in an orthogonal manner, thus increasing the complexity of the system. This, however, requires a sapient positioning of light sensitive elements with non-overlapping absorption spectra.

Furthermore, dynamic platforms are usually engineered and fabricated to display/conceal binding sites. However, dynamic changes of the ECM are not limited to adhesion cues and involve multiple signal changes according to different chronoprogrammes. In fact, the literature on dynamic platform mostly concerns platforms regulating adhesion/detachment and migration. Yet, recent examples demonstrated enormous potentialities of dynamic signal display, particularly in what concerns fate decisions and in the acquisition of specific tissue functions (Young and Engler, 2011; Yang et al., 2014; Roberts et al., 2016). Therefore, beside exploiting soluble

### REFERENCES


signals, engineering complex patterns of bioadhesive signals, the introduction of the parameter "time" acquires a crucial importance in the design of novel and more efficient culturing systems. However, dynamic surfaces, although introducing space, and time variations of biologic properties fall short of capturing more complex features of the native ECM. An intriguing strategy to tackle this issue has been recently proposed by Hay et al. who introduced the concept of bacteria-based materials (Hay et al., 2018). These finely altered the adhesive microenvironment upon external commands, eventually dictating stem cell fate decision. Non-pathogenic bacteria can be engineered as micromachines able to assemble animal proteins or release factors upon external stimulation. Nowadays, biologic genetic circuits can be integrated in bacteria to control the yield and production rate (Bittihn et al., 2018). We expect future research to be focused on integrating different strategies and technologies for the fabrication of multi-functional and dynamic platforms able to recapitulate intricate biochemical, biophysical and cellular processes that are fundamental to promote and guide biologically relevant phenomena. Applications like organoid generation or the development of pathological models would largely benefit from these advancements. Therefore, we expect that the field of dynamic platform engineering can potentially contribute not only to unravel complex biological events such as migration and differentiation, but could impact the medical and pharmaceutical areas thanks to the development of devices for drug screening and discovery and systems for tissue regeneration.

### AUTHOR CONTRIBUTIONS

CC, LR, and MV gathered information and data from the literature and drafted the manuscript. MV and PN took the lead in organizing the final version of the manuscript. MV wrote the final version of the manuscript and edited figure panels. All authors provided critical feedback to shape the final version of the manuscript.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Cimmino, Rossano, Netti and Ventre. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Composite of Elastin-Based Matrix and Electrospun Poly(L-Lactic Acid) Fibers: A Potential Smart Drug Delivery System

#### Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Elisa Mele, Loughborough University, United Kingdom Marc Antoni Fernández Yagüe, National University of Ireland, Ireland

> \*Correspondence: Antonella Bandiera abandiera@units.it

#### †Present Address:

Luisa Stella Dolci, Department of Pharmacy and Biotechnology, Alma Mater Studiorum - Università di Bologna, Bologna, Italy

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 12 July 2018 Accepted: 24 August 2018 Published: 12 September 2018

#### Citation:

Bandiera A, Passamonti S, Dolci LS and Focarete ML (2018) Composite of Elastin-Based Matrix and Electrospun Poly(L-Lactic Acid) Fibers: A Potential Smart Drug Delivery System. Front. Bioeng. Biotechnol. 6:127. doi: 10.3389/fbioe.2018.00127

#### Antonella Bandiera<sup>1</sup> \*, Sabina Passamonti <sup>1</sup> , Luisa Stella Dolci 2† and Maria Letizia Focarete2,3

<sup>1</sup> Department of Life Sciences, University of Trieste, Trieste, Italy, <sup>2</sup> Department of Chemistry "G. Ciamician" and National Consortium of Materials Science and Technology (INSTM, Bologna RU), Alma Mater Studiorum - Università di Bologna, Bologna, Italy, <sup>3</sup> Health Sciences and Technologies–Interdepartmental Center for Industrial Research, Alma Mater Studiorum - Università di Bologna, Bologna, Italy

Stimuli-responsive hydrogel matrices are inspiring manifold applications in controlled delivery of bioactive compounds. Elastin-derived polypeptides form hydrogel matrices that may release bioactive moieties as a function of local increase of active elastases, as it would occur in several processes like inflammation. In view of the development of a patch for healing wounds, recombinant elastin-based polypeptides were combined with a proteolysis-resistant scaffold, made of electrospun poly-L-lactic acid (PLLA) fibers. The results of this study demonstrated the compatibility of these two components. An efficient procedure to obtain a composite material retaining the main features of each component was established. The release of the elastin moiety was monitored by means of a simple protocol. Our data showed that electrospun PLLA can form a composite with fusion proteins bound to elastin-derived polypeptides. Therefore, our approach allows designing a therapeutic agent delivery platform to realize devices capable of responding and interacting with biological systems at the molecular level.

Keywords: elastin, electrospun matrix, composite, smart release, drug delivery

## INTRODUCTION

The biomaterial field has enormously evolved in the last decades. Many different materials have been developed and are available to realize devices that can be used for a wide variety of applications. With the advent of bioengineering and regenerative medicine, the field of biomaterials entered the phase of developing devices capable of actively interacting with the biological system, rather than passively integrating within it (Hench and Polak, 2002). However, despite the huge work done in this field, there is a constant demand for innovative solutions in order to address many still unmet biomedical needs (Holzapfel et al., 2013).

Among the many composite biomaterials that have been developed and tested for medical applications, the combination of electrospun fibers and hydrogels recently attracted the attention of researchers. Due to their individual features, combining the advantages of both components results in a product with superior properties, that has a high potential for expanding the range of applications of the final construct (reviewed in Bosworth et al., 2013; Xu et al., 2016).

The advantages achieved by the combination of electrospun fibers and hydrogels have been demonstrated in the field of controlled drug delivery, where controlled release could be obtained by exploiting the release characteristics of the two components (Han et al., 2013; Bruggeman et al., 2017). In the field of tissue engineering, combining the biomimetic properties, hydrophilicity, and softness of hydrogels with the mechanical strength of electrospun sheets, allows to mimic the structure of tissue extracellular matrix (Gualandi et al., 2016). Other fields of application of electrospun fibers/hydrogels composites are those of biotechnology and biosensors (Xu et al., 2016).

Electrospun fibrous mats, made of natural or synthetic polymers, exhibit high porosity, high surface area to volume ratio, and good mechanical properties. Moreover, these properties can be easily tailored by changing the fiber diameter through proper control of the electrospinning process. For these reasons, electrospun mats represent a valuable platform for drug delivery and tissue engineering and regeneration (Chen et al., 2018).

Elastin-like polypeptides are an emerging class of biotechnologically derived biopolymers that are inspired to the tissue structural protein elastin (MacEwan and Chilkoti, 2010; Girotti et al., 2011). In our lab, starting from design, cloning and expression of synthetic genes, a family of recombinant proteins named Human Elastin-like Polypeptides (HELPs) was produced (Bandiera, 2010). This versatile platform can be readily customized by the fusion of bioactive domains of interest, thus embedding the new functionality in the final construct. A method for the preparation of hydrogel matrix based on these HELPs was set up (Bandiera, 2011) and the specific stimuli-induced release was demonstrated (Bandiera et al., 2014).

To the best of our knowledge, there are only few examples of elastin-like based composites that have been developed till now (reviewed in Kakinoki et al., 2014; Yeo et al., 2015). Here, we describe an approach to obtain a new composite material based on deposition of elastin-like based on electrospun poly-L-lactic acid (PLLA-HELP).

### MATERIALS AND METHODS

### HELP Biopolymers

HELP and mHELP, the latter being a construct of HELP obtained by fusing at the C-terminal region a functional domain (unpublished data) were produced exploiting their inverse phase transition properties as already detailed (Bandiera, 2010). The purified products were checked by SDS-PAGE and lyophilized.

### Electrospun PLLA Scaffold Fabrication

Poly(L-lactic acid) (PLLA) (Lacea H.100-E) (Mw = 8.4 10<sup>4</sup> g mol−<sup>1</sup> , PDI = 1.7) was supplied by Mitsui Fine Chemicals. Dichloromethane (DCM) and dimethylformamide (DMF), were purchased by Sigma–Aldrich and were used without any further purification. PLLA was dissolved in a mixed solvent, DCM:DMF = 65:35 v/v, at a concentration of 13% w/v. The polymeric solution was electrospun by means of an electrospinning apparatus (Spinbow srl, Italy) by applying the following processing conditions: applied voltage = 18 kV, feed rate = 0.015 ml min−<sup>1</sup> , needle-to-collector distance = 15 cm. The electrospun mat was produced at RT and at relative humidity of 40–50% and was kept under vacuum over P2O<sup>5</sup> at RT overnight in order to remove residual solvents.

### Water Contact Angle (WCA) Measurements

Static WCA measurements were performed at RT under ambient conditions by using an optical contact angle and surface tension meter KSV's CAM 100 (KSV, Espoo, Finland). Milli-Q water was used for measurements. The water drop profile images were collected in a time range of 0–60 s, every 1 s. Sixty seconds was selected as the upper time limit since it was verified that, after that period, the WCAs reached a constant value. Optical contact angle and pendant drop surface tension software was used for image processing. Results (WCA at 60 s) were averaged on at least five measurements obtained at different areas of the sample.

### HELP Deposition and Cross-Linking on PLLA

Five percent (w/v) of water solution of HELP and the other proteins tested were prepared and 5 µl of each sample were spotted on small sheets of PLLA mat. Water was evaporated at RT and the dried samples were stored (controls) or washed with excess water, respectively. Washed samples were dried as well. For cross-linking, 2 µl of microbial transglutaminase (60 mg/ml, N-Zyme Biotec GmbH, Darmstadt, Germany) were added to 30 µl of 5% (w/v) HELP or its fusion in 10 mM Tris/HCl pH 8 (Sigma-Aldrich, #T1503). Spots of 5 µl were deposited on PLLA mat and the reaction was carried on for 1 h at RT or at 5◦C overnight in a wet chamber to avoid drying. After cross-linking the samples were washed overnight as described above and stored dry.

### Evaluation of Protein Retention on PLLA

Samples of PLLA with adsorbed or cross-linked HELP were stained for 10 min in 0.5 mg/mL Amidoblack (Serva, #12310), 50% Ethanol and then rinsed twice with water for 10 min. Each spot was cut off and soaked in 200 µl of 50 mM Tris/HCl pH 7.5, 1 mM CaCl2, in the absence or in presence of 0.5 µg elastase (Sigma-Aldrich, #E7885). Samples were incubated overnight at 37◦C. Supernatants were read at 620 nm by a microplate reader (Synergy H1, Bio Tek).

### Scanning Electron Microscopy (SEM) Analysis

After cross-linking samples were rinsed with excess water. They were frozen at −20◦C and lyophilized. Slices were cut, mounted onto stubs using a double-sided adhesive and sputter coated with gold. Analysis was performed using a Leica Stereoscan 430i Scanning Electron Microscope.

### RESULTS

Our strategy to prepare a PLLA-HELP composite is schematized in **Figure 1A**.

The approach used in this study is the physical combination of two materials, through deposition of HELP-based protein onto the surface of PLLA mat. Blending of HELP protein and PLLA solution before electrospinning was not possible since the organic solvents required for electrospinning PLLA were not compatible with the protein. Moreover, only a small amount of the protein would have been available at the surface of the fibers. The first step of composite fabrication consisted in the evaluation of the compatibility of these two materials, since WCA measurement demonstrated that electrospun PLLA is a hydrophobic material (WCA = 120◦ ± 3 ◦ ), whereas the HELP protein is soluble in aqueous solution. Both HELP and one of its fusion products were dissolved in water and dropped on a PLLA sheet, avoiding the physical contact with any surface beneath it. A solution of Bovine Serum Albumin (BSA) at the same concentration was tested as reference. Interestingly, as shown in **Figure 1B**, PLLA samples in contact with HELP-based protein solutions immediately became wet and the drop spread over the sheet, whereas when a drop of BSA was deposited onto the PLLA sheet, no wetting was observed. After water evaporation, the sample deposited on PLLA was no longer observable. To assess the presence of HELP protein on PLLA, we stained the samples with Amidoblack. After de-staining, the protein became evident (**Figure 1C**, top left).

In parallel, a replica sample was prepared and submitted to an overnight wash before the staining procedure. As shown in **Figure 1C** (top right), almost no stain was detectable after the overnight wash. On the contrary, when the transglutaminase enzyme was added to the HELP protein to determine its crosslinking (Bandiera, 2011), no difference could be detected between the unwashed (**Figure 1C**, bottom left) and the overnight washed replica (**Figure 1C**, bottom right). This indicated that the enzymatic cross-linking stabilized the HELP protein on the PLLA sheet.

We set up a method to estimate the stability of HELP deposited on PLLA, by exploiting the susceptibility of HELP to protease-dependent elastolysis (Corich et al., 2017). **Figure 1C** shows that after the wash, cross-linked HELP-PLLA samples retained up to 96% of the HELP protein after the wash step, whereas <20% of it remained in the non-cross-linked samples. This suggested that cross-linking caused the formation of a stable HELP-PLLA composite.

FIGURE 1 | HELP-PLLA composite design. (A) Schematic description of the experimental approach; (B) Representative images of different protein aqueous solutions 0.5 min after deposition on PLLA matrix. Bar is 5 mm; (C) Stabilization effect of the HELP cross-linking process (CRLK) compared to simple deposition of the protein (free) on PLLA matrix: representative images of unwashed and washed samples, Amidoblack staining. Bar is 5 mm. Evaluation of sample stability after washing is reported in the histogram. Each value represented the mean ± SD, n = 3. Differences between free and cross-linked samples after extensive washing were assessed by one-way ANOVA, \*\*\*p < 0.001 compared to unwashed samples.

Since this work was undertaken with the aim of realizing smart devices endowed with environmentally-controlled functionality, it was of great interest to assess if fusion-modified HELP could be stabilized in the PLLA matrix as well. For this reason, mHELP, a fusion-modified HELP that we recently obtained by cloning a functional binding domain to the HELP backbone (unpublished results), was employed. The same procedure used for HELP-PLLA composite was applied to mHELP-PLLA samples and the analyses were performed giving results similar to those shown in **Figure 1C** (not shown).

To obtain further information on the morphology of the stabilized mHELP-PLLA samples, SEM analysis was performed. **Figure 2A** shows SEM analysis of PLLA electrospun fibers. Cross-linking the composite, even after extensive overnight washing, changed the morphology of PLLA, which appeared as coated by a continuous, amorphous layer, filling the pores among the fibers (**Figure 2C**). In the absence of cross-linking, tiny deposits of protein were observed on PLLA only after a quick wash (**Figure 2B**).

These results indicate that both HELP macromolecule and its fusion modifications could be employed for preparation of new composites endowed with functional activity. Analysis of activity retention is currently ongoing.

### DISCUSSION

HELP matrix has been shown to possess stimuli-responsive properties, undergoing selective degradation in the presence of elastolytic activity. Exploiting this feature represents an attractive option to realize smart systems that can be employed for therapeutic delivery, tissue engineering, and biosensing (Corich et al., 2017).

In this view, we explored the possibility of integrating the HELP hydrogel matrix with an electrospun PLLA support and our approach is schematized in **Figure 1A**. Besides being endowed with adequate mechanical tensile strength, electrospun PLLA has a high degree of biocompatibility and has been extensively employed in the fabrication of bioresorbable scaffolds for tissue engineering applications (Chen et al., 2018). Electrospun PLLA is known to be hydrophobic. To increase its hydrophilicity and wettability, surface modifications with functional groups are needed. These enable conjugation or chemical interactions with hydrogels (Dolci et al., 2014; Gualandi et al., 2016). In our approach, the first step of composite fabrication consisted in the evaluation of the compatibility and in the exploration of the conditions for a stable integration of the HELP-based hydrogel in the PLLA electrospun mat. Interestingly, we found that no treatment was needed to enhance PLLA wettability for HELP deposition. The presence of the protein in solution allowed the instantaneous PLLA fiber permeation by the solution (**Figure 1B**). The finding that physical adsorption of HELP-based proteins onto PLLA mat increased the wetting properties of electrospun PLLA surface is in agreement with previous studies on substrates modified with other elastin-based molecules (Jordan et al., 2007; Srokowski and Woodhouse, 2013) and it is likely related to the relatively high hydropathy index of the HELP protein (Bandiera et al., 2010). Notably, the cross-linking process was the key step to stabilize the HELP moiety on the PLLA mat (**Figure 1C**).

Keeping in mind that the HELP macromolecule can be tailored by addition of a bioactive domain, in principle, these conditions could be extended to any HELP derivative, obtained by C-terminal domain fusion. Thus, the resulting composite will be endowed with a new, specific functionality. Indeed, in this work we demonstrated that similar result could be obtained

FIGURE 2 | Representative SEM images of PLLA mat (A), mHELP deposited on the PLLA matrix in the absence of cross-linking (B) and mHELP undergone cross-linking after deposition on PLLA matrix (C). Bar is 10µm.

using both a HELP protein and mHELP, a fusion-modified HELP, in which the fusion domain represents about one-fifth of the whole macromolecule. SEM analysis was used to achieve the structural evidence of the integration of both materials in a new composite obtained after the cross-linking process (**Figure 2C**).

This achievement represents the second step of our strategy, i.e., the realization of a composite with a tailored HELP fusion, endowed with a specific functionality (**Figure 1A**).

A third step can be foreseen, consisting in the stimuliinduced release of the active domain. Future work will be dedicated to fully characterize the new material that opens the way for the realization of stimuli-responsive biomedical devices for the delivery of therapeutic and bioactive substances and for regenerative medicine, as well as for the development of biosensors.

### CONCLUSION

In this work we explored the opportunity to realize a composite material made of HELP-based hydrogels and PLLA electrospun fibers. The results clearly show that the fabrication of a composite is feasible and that the features of the two materials can be successfully integrated. This strategy is useful to combine the individual performances of the two constituents and future study will clarify the properties of the new hybrid material. The stimuli-responsive nature of the HELP moiety has been already

### REFERENCES


proven and described (Bandiera et al., 2014; Corich et al., 2017) and it represents an advantage that can be conferred to any new material derived from it. Moreover, HELP-based proteins represent a platform that is readily customizable by molecular fusion of exogenous domains, to confer specific functionality to the final product.

We believe that this strategy will contribute to bypass shortcoming and to improve the performance of the single components, opening the way to the realization of devices potentially able to release bioactive compounds upon specific stimuli.

### AUTHOR CONTRIBUTIONS

All the authors conceived and planned the experiments. AB and LD carried out the experimental work. All authors discussed the results and contributed to the final manuscript, reading, and approving the submitted version.

### ACKNOWLEDGMENTS

The authors gratefully thanks Dr. Francesca Vita for performing SEM analysis. Dr. Lucia Corich is also acknowledged. This work was supported by Beneficentia Stiftung (BEN 2014/125), Vaduz, Lichtenstein to AB, Crossborder Cooperation programme Italy-Slovenia 2007-2013 (Trans2Care project) to SP and by a Regione Emilia Romagna grant (POR-FESR 2014–2020) to MF.

affects fibroblast morphology. Plasma Process. Polym. 11, 203–213. doi: 10.1002/ppap.201300104


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Bandiera, Passamonti, Dolci and Focarete. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

**40**

# Cellular Response to Surface Morphology: Electrospinning and Computational Modeling

#### Anna Denchai <sup>1</sup> , Daniele Tartarini <sup>2</sup> and Elisa Mele<sup>1</sup> \*

<sup>1</sup> Department of Materials, Loughborough University, Loughborough, United Kingdom, <sup>2</sup> Department of Civil Engineering, University of Sheffield, Sheffield, United Kingdom

Surface properties of biomaterials, such as chemistry and morphology, have a major role in modulating cellular behavior and therefore impact on the development of high-performance devices for biomedical applications, such as scaffolds for tissue engineering and systems for drug delivery. Opportunely-designed micro- and nanostructures provides a unique way of controlling cell-biomaterial interaction. This minireview discusses the current research on the use of electrospinning (extrusion of polymer nanofibers upon the application of an electric field) as effective technique to fabricate patterns of micro- and nano-scale resolution, and the corresponding biological studies. The focus is on the effect of morphological cues, including fiber alignment, porosity and surface roughness of electrospun mats, to direct cell migration and to influence cell adhesion, differentiation and proliferation. Experimental studies are combined with computational models that predict and correlate the surface composition of a biomaterial with the response of cells in contact with it. The use of predictive models can facilitate the rational design of new bio-interfaces.

Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Elia Ranzato, Università degli Studi del Piemonte Orientale, Italy Simona Martinotti, Università degli Studi del Piemonte Orientale, Italy

> \*Correspondence: Elisa Mele e.mele2@lboro.ac.uk

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 29 August 2018 Accepted: 08 October 2018 Published: 24 October 2018

#### Citation:

Denchai A, Tartarini D and Mele E (2018) Cellular Response to Surface Morphology: Electrospinning and Computational Modeling. Front. Bioeng. Biotechnol. 6:155. doi: 10.3389/fbioe.2018.00155 Keywords: bio-interfaces, surface topography, electrospinning, micro-patterning, mathematical modeling

## INTRODUCTION

The natural regeneration process of human tissues is strongly regulated by the interaction of cells with the extracellular matrix (ECM) (Lutolf and Hubbell, 2005; Liu and Wang, 2014). ECM is a dynamic and complex fibrous network of proteins and polysaccharides, such as collagen, elastin, fibronectin, laminin, proteoglycans and glycosaminoglycans. Cells interact with ECM by transmembrane receptors, known as integrins, that ligate with specific motifs of ECM proteins, for example arginine, glycine and arginylglycylaspartic acid (RGD) peptides (Anderson et al., 2016; Dalby et al., 2018). Cells continuously remodel the ECM environment, which, in turn, influences cell behavior and fate (differentiation, proliferation and migration) by biochemical, physical and mechanical signals (Geiger et al., 2001), and provides structural support to cells. Recent studies have investigated the effects of ECM physical properties, particularly porosity, topography and hierarchical 3D architecture, on cellular functions, and extrapolated rules to design structures for effective tissue regeneration (Li et al., 2017; Marino et al., 2017; Lin et al., 2018).

One of the technologies that is widely used to produce ECM-mimicking structures and particularly to replicate the fibrillar architecture of ECM is electrospinning (Khorshidi et al., 2016). The electrospinning technique allows the production of networks of fibers with a diameter in the range of few nm to few µm via the application of electrical forces to polymer solutions or melts (Bhardwaj and Kundu, 2010; Mele, 2016; Zhang et al., 2016). Structural modifications of electrospun nanofibres, such as altering topographical characteristics and inducing porosity, can be achieved by controlling and varying the process parameters (polymer concentration, applied voltage, evaporation rate of the solvent used). Similarly, changes to the final makeup of the fibrous network, such as alignment and patterning of fibers, can be obtained by modifications of the electrospinning apparatus or post-processing.

This mini review analyses a selection of recent works on the use of solution electrospinning to create nanofibres with engineered surface topography (random, aligned and patterned fibers) for controlling adhesion, differentiation, and migration of different cells lines. The mini review is divided in two main sections: the first one will focus on experimental studies on electrospun fibers that provide physical cues for cell growth and differentiation; the second section will discuss computational models to predict cell behavior on micropatterns. Although mathematical models that simulate cell behavior on electrospun fibers are not currently available, the computational approaches here discussed can be adapted, in the future, to electrospun scaffolds and used to elucidate the underlying mechanisms responsible for cell-fiber interaction.

### EFFECTS OF FIBER TOPOGRAPHY AND MICRO-PATTERNING ON CELLULAR RESPONSE

Multiple studies have demonstrated that the morphology and roughness of fibers produced by electrospinning influence cell adhesion, proliferation, and orientation (Sill and von Recum, 2008; Xie et al., 2008; Bergmeister et al., 2013; Cirillo et al., 2014; Zhu et al., 2015; Sun et al., 2018). All factors that are imperative for successful tissue regeneration (Agarwal et al., 2008). Cells can sense topographical structures on a surface by filipodia that are actin-rich protrusions (0.1–0.3µm in diameter) of the cell membrane and are involved in cell contact guidance (Mattila and Lappalainen, 2008; Dalby et al., 2014). If nanoscale aligned features are present onto a surface, filopodia tend to orient along the direction of the features and determine cytoskeleton orientation. Focal adhesions at the cell membrane mediate the initial cell-biomaterial interaction, with integrin ligands in direct contact with the substrate and connected to the actin microfilaments of the cell cytoskeleton by a 40-nm stratum, which includes focal adhesion kinase (FAK), paxillin, talin, and vinculin (Kanchanawong et al., 2010).

This section of the review will discuss how electrospun mats with controlled porosity and surface morphology have been used to influence the behavior of mesenchymal stem cells (MSCs) (Jiang et al., 2015; Yin et al., 2015; Baudequin et al., 2017; Lin et al., 2017; Liu et al., 2017; Nedjari et al., 2017; Su et al., 2017; Zhang et al., 2017; Ghosh et al., 2018; Jin et al., 2018; Rahman et al., 2018; Sankar et al., 2018) and human umbilical vein endothelial cells (HUVECs) (Fioretta et al., 2014; Xu et al., 2015; Shin et al., 2017; Taskin et al., 2017; Yan et al., 2017; Ahmed et al., 2018). The literature on other cell lines, such as on myoblasts (Mele et al., 2015; Jun et al., 2016; Park et al., 2016; Tallawi et al., 2016; Abarzúa-Illanes et al., 2017; Yang et al., 2017) and neuron-like cells (Binan et al., 2014; Xie et al., 2014; Malkoc et al., 2015; Xue et al., 2017; Hajiali et al., 2018; Xia and Xia, 2018), will not be analyzed in detail here but a summary of it is reported in **Table 1**.

## Mesenchymal Stem Cells

MSCs are multipotent stem cells that are primarily isolated from bone marrow, but they can also be found in adipose tissue, dental pulp, placenta, umbilical cord and other vascularized tissues throughout the body (Lv et al., 2014; Tartarini and Mele, 2015). MSCs are of great interest in regenerative medicine, because of their therapeutic effects, such as: ability to differentiate into various cell types and therefore promote regeneration of a wide range of tissues (bone, cartilage, muscle, marrow, tendon, ligament, nervous tissue, and skin); secretion of bioactive molecules for tissue repair; migration to inflamed tissues and modulation of local inflammation; immunomodulatory functions (Sharma et al., 2014).

In a recent research, Zhang et al. have studied how the topography and fibrillar organization of electrospun poly (εcaprolactone) (PCL) fibers influences the recruitment of MSCs in vivo and ex vivo (Zhang et al., 2017). PCL mats (randomly distributed fibers) were implanted into the subcutaneous tissue of rats and the results were compared with solid PCL films (not electrospun). It was observed that, during the initial postimplantation period (1 day), a great number of macrophages with M1 phenotype (pro-inflammatory) were recruited to the PCL fibers, differently from solid PCL. This was attributed to the high surface area of the fibers and the porosity of the electrospun mats that promoted protein adsorption from the surrounding tissue, such as complement C3a (a chemo-attractant responsible to activate and recruit immune cells), fibronectin and vitronectin. After 4 and 7 days of implantation, the PCL fibers attracted host MSCs and modulated macrophages polarization with an increased number of cells exhibiting M2 (pro-healing) phenotype. While the number of M2 cells continuously increased over the entire period of implantation for PCL fibers, this was not the case for solid PCL where a large population of M1 cells was retained. Migration of MSCs was also observed in ex vivo experiments conducted with the implanted PCL samples. It was found that the macrophages at the implanted PCL mats secreted high levels of SDF-1, a chemokine that mediates MSCs recruitment by interacting with CXC chemokine receptors on the MSCs membrane. The study concluded that the physical organization of the PCL electrospun network induced the phenotype M1-to-M2 transition of macrophages that attracted MSCs at the implantation site by releasing SDF-1. This cascade of events was beneficial to stimulate tissue repair. PCL electrospun fibers have been used also to stimulate the production of pro-angiogenic and anti-inflammatory paracrine factors in rat adipose-derived MSCs (Ad-MSCs) (Su et al., 2017) and in skin excisional wound-healing model in rats (**Table 2**). Ad-MSCs were seeded on three types of electrospun PCL fibers, random (REF), aligned (AEF) and with a mesh pattern (MEF). It was observed that scaffolds with oriented fibers (AEF and MEF) promoted the expression of PGE2 TABLE 1 | Summary of the recent literature on the use of electrospun fibers to control morphology, alignment and differentiation of diverse cell lines.


(Continued)

#### TABLE 1 | Continued


(Prostaglandin E2, a potent inflammatory mediator), iNOS (inducible Nitric Oxide Synthase), VEGF (vascular endothelial growth factor) and HGF (hepatocyte growth factor), compared to REF scaffolds. In order to elucidate the molecular signaling mechanism responsible for the paracrine secretion of Ad-MSCs, the cells were treated with an inhibitor of NF-kB (a transcription factor that induces the expression of pro-inflammatory genes) and this significantly reversed the paracrine response of MSCs to the electrospun scaffolds. The authors therefore speculated that, in the presence of the scaffolds, MSCs behaved as if they were exposed to an external inflammatory stimulus. Similar results have been recently reported for MSCs cultured on electrospun fibers of PCL/polytetrahydrofuran (PTHF) urethane (P fibers) and PCL-PTHF urethane/collagen I (PC fibers) (Jiang et al., 2018). In this case, down-regulation of genes that contribute to inflammation and suppression of the NF-kB pathway signaling pathway were achieved by changing the mechanical properties of the fibers. PC fibers with a Young's modulus of 4.3 MPa were able to suppress inflammation, differently from P fibers (Young's modulus of 6.8 MPa).

Another study has investigated how electrospun PCL scaffolds with a novel random-aligned-random structure can be used to mimic bone-ligament connections and native ligaments (Lin et al., 2017). The authors designed a fiber collecting device for the fabrication of electrospun scaffolds with a controlled spatial distribution of random and aligned fibers.

TABLE 2 | Summary of main results reported in selected recent papers on electrospun scaffolds used in vivo experiments.


(Continued)

#### TABLE 2 | Continued


The regions of the scaffold with random fibers were then mineralized with Ca-P. In vitro tests on human bone marrow MSCs (hBMSCs) revealed that fiber anisotropy modified cells' morphology: polygonal, round-shaped cells without alignment were detected in the random, mineralized regions of the scaffold; while elongated spindle-shaped cells aligned along the fiber direction were visible in the aligned region. Moreover, the aligned fibers significantly up-regulated tendon-specific and tendon-related markers (Tnmd, Mkx) and therefore guided tenogenic phenotypes of hBMSCs; while, the regions with random, mineralized fibers determined the expression of bone-specific markers (Runx-2, Ocn, Opn) and consequently hBMSCs osteogenic phenotypes. Although the authors have not elucidated the underlying cell signaling mechanisms, this work demonstrates that electrospun scaffolds with engineered fiber anisotropy are advantageous to achieve region-specific distribution of tendon- and bone-related genes and find potential application in ligament repair and regeneration of bone-ligament connections.

The possibility to mediate the expression of signaling biomolecules by electrospun fibers and hence guide MSCs differentiation has been demonstrated also by Rahman and co-workers (Rahman et al., 2018). They investigated the odontoblastic differentiation of human dental pulp MSCs (DP-MSCs) on polystyrene (PS) random fibers. The cells cultured on PS mats strongly increased the expression of bone morphogenetic proteins (BMPs) and Wnt ligands that are essential in tooth development: Wnt3a transcript expression was more than 50 folds higher after 4 days of culturing on PS fibers than on standard petri dishes. The levels of odontoblast/osteoblast markers, such as dentin sialophosphoprotein (DSPP), osteocalcin, and bone sialoprotein, were also higher for DP-MSCs cultured on electrospun fibers. The results of this study indicate that nanofibres mimicking the in vivo microenvironment are crucial to stimulate the differentiation of DP-MSCs into odontoblasts (specialized cells responsible for dentin formation) by mediating the production of signaling molecules including Wnt3a, and to promote dentinogenesis. Osteogenesis of MSCs has been reported also on random Poly-L-lactic acid (PLLA) fibers, due to cytoskeletal rearrangements and tensions, which in turn influence intracellular mechanotransductive pathways (Yin et al., 2015). In fact, when the cells were treated with Rho kinase (ROCK) inhibitor Y-27632 (inhibitor of myosin-generated cytoskeletal tension), loss of lineage commitment was detected, and cells' morphology was not affected by the fibers topography.

The works here summarized and the others conducted on the interaction of MSCs with electrospun substrates (**Tables 1**, **2**) demonstrate that networks of polymer fibers (random, aligned and hierarchical) are effective in providing topographical and physical cues to guide differentiation of stem cells. These observations have led to the development of bioinspired scaffolds with potential future implications in diverse clinical areas, including the regeneration and repair of bone, tendon, ligament, dentin, and skin.

### Human Umbilical Vein Endothelial Cells

Vascular endothelial cells are of fundamental importance for the entire circulatory system, because they are involved in fluid filtration, homeostasis and prevention of thrombosis (Rajendran et al., 2013). Endothelial cells and particularly HUVECs, which are isolated from human umbilical cord veins, are widely used to study cardiovascular diseases and develop biomedical devices for vascular tissue engineering (Lei et al., 2016). One important aspect to consider when designing scaffolds for endothelial cells is the role played by surface topography, at micro- and nanoscale, on cell adhesion, proliferation and migration, to create a physiological environment that stimulates the formation of a functional endothelium.

A recent work of Ahmed and co-workers has reported on the influence of the diameter of electrospun fibers on HUVECs migration (Ahmed et al., 2018). Aligned poly(lactic-co-glycolic acid) (PLGA) fibers with five different diameters (0.5, 1, 2, 4, and 10µm) were analyzed. The greater cell displacement in a scratch wound assay was measured for HUVECs seeded on fibers with intermediate diameter (1 and 2µm) with peak migration velocity of 24 µm/h after 12 h of cell culture. HUVECs were able also to move on scaffolds with 0.5µm size fibers but at lower migration rates. On these scaffolds cell alignment and polarization, and higher levels of FAK expression were detected. FAK is a non-receptor tyrosine kinase that regulates cell shape, adhesion and motility. The fiber diameter influenced the focal adhesion of HUVECs but not their metabolism or the formation of cell-matrix anchorage points. At 12 h, a significant increase in phosphorylated FAK (pFAK, associated with actin regulation and adhesion dynamics) was detected, which is linked to the peak migration velocity. On the contrary, limited cell motility was observed for scaffolds with 4 and 10µm fibers. Investigations of the spatial distribution of pFAK revealed that pFAK was localized in the HUVECs cytosol for 0.5, 1.0, and 2.0µm fibers, and at the cell periphery for 4 and 10µm fibers (non-uniform distribution). This promoted uniaxial cell morphology and stimulated the migratory process to occur preferentially along the fiber longitudinal direction for scaffolds with intermediate fiber diameter. A similar conclusion has been drawn by other researchers working on HUVECs cultured on micropatterned scaffolds with spatially heterogeneous alignment

of poly(D,L-lactide) (PDLLA)/PCL electrospun fibers of 0.5– 1µm size (Xu et al., 2015). Fibrous scaffolds with patterns of random and well-aligned PDLLA/PCL fibers were prepared using a wire spring as template collector. It was observed that the micropatterned scaffolds induced the proliferation of HUVECs and modifications to their cytoskeleton morphology (**Figure 1**). The lowest values of mean cell body shape index (a parameter indicating the degree of cell polarization) were measured for cells cultured on patterned scaffolds having the longest distance (1,500µm) between regions with random and aligned fibers, indicating the highest degree of cell polarization and alignment. Furthermore, those scaffolds stimulated the cells to express high levels of angiogenesis-related genes and therefore they have potential applications in vascular tissue engineering. The combination of electrospinning and micro-pattering techniques has proven to be effective for creating hierarchical biointerfaces that direct the arrangement of endothelial cells and their biological functions (Shin et al., 2017; Yan et al., 2017).

### COMPUTATIONAL MODELS

The literature that has been discussed so far in this review provides experimental evidences that the surface topography of biomaterials influences cellular behavior, including cells' alignment, elongation, migration, phenotype transitions and differentiation. In vitro and in vivo studies are incredibly beneficial to collect data and results on how artificially created micro- and nano-features perform in realistic applications (**Tables 1**, **2**). The underlying mechanisms of cell-material interactions are only partially understood and further investigations are required to define the best scaffold design for promoting the regeneration of a target tissue (Kennedy et al., 2017; Paim et al., 2018). However, time and cost requirements for in vitro and in vivo tests pose limitations on the use of and reliance on experimental studies alone, together with ethical issues when animal models are concerned. Computational modeling has the significant advantage of facilitating research by conducting thousands of simulated trials with a wide range of variations and for a plethora of complex biological systems (Geris et al., 2018).

Albert and Schwarz have developed mathematical models to predict the dynamics of cell shape and forces on micropatterned substrates (Albert and Schwarz, 2014, 2016a,b). Their models are based on the cellular Potts model (CPM) that allows to simulate the behavior of single or interacting cells by describing them as internally structureless but spatially extended objects on a regular lattice (Voss-Böhme, 2012; Tartarini and Mele, 2015). The number of lattice sites belonging to a single cell defines the area occupied by the cell. By changing the lattice resolution and the indices of the lattice sites, cells with arbitrary shape and shape evolutions can be represented. Initially, the authors compared simulations with experimental data on single cell attached on crossbow, Y and H patterns (Albert and Schwarz, 2014). The model well described how the cell contour adapted to the pattern's geometry and reconstructed the traction forces in agreement with experiments. The forces were higher at the extremities of the patterns (adhesive edges of the contour) and increased with the curvature of the contour depending on the availability of receptors for focal adhesion. The CPM-based model was then used to predict the collective behavior of cells on micropatterns, including cell division, cell-cell contacts and migration (Albert and Schwarz, 2014). The model predicted, for example, that for a cell dividing on a L shaped pattern, the two daughter cells were most likely to be located on the two arms of the L, as confirmed by experimental results. In order to identify the optimal adhesive patterns to control cell functions, CPM was combined with genetic algorithms (GAs) (Albert and Schwarz, 2016c) (**Figure 2**), which are computational techniques inspired by natural evolution for the heuristic search of problem solutions (McCall, 2005). The migration of cells on rachet micropatterns in a linear arrangement was analyzed and the algorithm predicted that a triangular shape was ideal to guide cell migration in the direction of the tip of the triangle, as also demonstrated experimentally. Differently from what expected though, the most effective pattern to achieve unidirectional migration of cells consisted of asymmetric triangles that were rotated and

connected to one another to form a pattern with an almost straight horizontal edge. The computational model developed is a useful tool to predict cell interactions with structured scaffolds and it can be adapted to simulate diverse cellular processes.

2016 of the Royal Society of Chemistry.

With a distinct lack of literature on computational/numerical modeling that predicts how cells interact with electrospun nanofibrous structures (role of roughness and topography), there is a clear gap in the field which has great potential if correctly pursued. This will open even more possibilities to design and create novel fibrous scaffolds with engineered surface structures (Ziebert and Aranson, 2016). For example, computer aided characterization of complex biointerfacial interactions of specific polymer fibers could be created. Computational algorithms and numerical solutions could be formulated to generate a method of predicting the most suitable surface topography of electrospun mats for specific cells and to prompt tissue regeneration processes. In the development of computational models that describe how cell behavior is affected by the surface properties of electrospun scaffolds, geometrical parameters to be considered include fibers diameter, fibers organization and degree of alignment, porosity of the mat, presence of nanostructures or nanopores on single fiber surface, overall roughness of the electrospun mat. All these aspects have been evaluated experimentally, as discussed in the previous section of this mini review.

### CONCLUSIONS

Electrospun nanofibers have become vital structures for a plethora of different applications, with the field of biomedical

#### TABLE 3 | Clinical trials of electrospun scaffolds.


The data are obtained from ClinicalTrials.gov, a resource provided by the U.S. National Library of Medicine (Accessed on September 2018).

engineering being, arguably, one of the most important. Thanks to the versatility of electrospinning, nanofibrous scaffolds can be tailored and modified to improve their biocompatibility for applications such as tissue engineering, drug delivery and wound dressings. For example, electrospun mats have been used in clinical studies for the treatment of arterial occlusive disease, skin cancer and diabetic foot (**Table 3**). As discussed in this review, fiber alignment, micropatterning, and controlled porosity of nanofibrous mats have all been found to have significant effect on cellular behavior, inducing cell attachment, migration and differentiation. Extensive research has been conducted on exploring morphological cues provided by 2D electrospun mats, and only recently fibrous 3D scaffolds have been proposed to closely mimic the ECM structure (Cai et al., 2013; Lee et al., 2014; Cho et al., 2016; Hwang et al., 2018; Unnithan et al., 2018). The studies conducted so far have demonstrated that a fine tuning of the 3D porosity of the electrospun scaffolds is crucial to promote cell infiltration. Future research in the field should combine experimental studies with numerical and

### REFERENCES

Abarzúa-Illanes, P. N., Padilla, C., Ramos, A., Isaacs, M., Ramos-Grez, J., Olguín, H. C., et al. (2017). Improving myoblast differentiation on electrospun poly(ε-caprolactone) scaffolds. J. Biomed. Mater. Res. A. 105, 2241–2251. doi: 10.1002/jbm.a.36091

computational modeling for the design and fabrication of novel micro- and nanostructured 3D scaffolds. Computer aided simulations could not only be used to predict cell interaction with specific topography but be formulated in a manner which then advises on the most suitable functional group (or biological molecule) that ought to be immobilized on the surface or embedded within the scaffold. This would require taking in consideration a complex combination of parameters that include the chemical composition of the scaffold (exposed chemical groups, wetting properties, and biodegradation), micro- and nano-porosity, organization of the fibrous network (random or aligned fibers) and mechanical properties of the scaffold.

### AUTHOR CONTRIBUTIONS

AD and EM contributed to the conception and design of the study. AD and EM wrote the first draft of the manuscript. DT wrote sections of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

Agarwal, S., Wendorff, J., and Greiner, A. (2008). Use of electrospinning technique for biomedical applications. Polymer 49, 5603–5621. doi: 10.1016/j.polymer.2008.09.014

Ahmed, M., Ramos, T., Wieringa, P., Blitterswijk, C. V., Boer, J., and Moroni, L. (2018). Geometric constraints of endothelial cell migration on electrospun fibres. Sci. Rep. 8:6386. doi: 10.1038/s41598-018-24667-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Denchai, Tartarini and Mele. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Borrowing From Nature: Biopolymers and Biocomposites as Smart Wound Care Materials

#### Giulia Suarato1,2 \*, Rosalia Bertorelli <sup>2</sup> and Athanassia Athanassiou<sup>1</sup> \*

*<sup>1</sup> Smart Materials, Istituto Italiano di Tecnologia, Genoa, Italy, <sup>2</sup> In vivo Pharmacology Facility, Istituto Italiano di Tecnologia, Genoa, Italy*

Wound repair is a complex and tightly regulated physiological process, involving the activation of various cell types throughout each subsequent step (homeostasis, inflammation, proliferation, and tissue remodeling). Any impairment within the correct sequence of the healing events could lead to chronic wounds, with potential effects on the patience quality of life, and consequent fallouts on the wound care management. Nature itself can be of inspiration for the development of fully biodegradable materials, presenting enhanced bioactive potentialities, and sustainability. Naturally-derived biopolymers are nowadays considered *smart materials*. They provide a versatile and tunable platform to design the appropriate extracellular matrix able to support tissue regeneration, while contrasting the onset of adverse events. In the past decades, fabrication of bioactive materials based on natural polymers, either of protein derivation or polysaccharide-based, has been extensively exploited to tackle wound-healing related problematics. However, in today's World the exclusive use of such materials is becoming an urgent challenge, to meet the demand of environmentally sustainable technologies to support our future needs, including applications in the fields of healthcare and wound management. In the following, we will briefly introduce the main physico-chemical and biological properties of some protein-based biopolymers and some naturally-derived polysaccharides. Moreover, we will present some of the recent technological processing and green fabrication approaches of novel composite materials based on these biopolymers, with particular attention on their applications in the skin tissue repair field. Lastly, we will highlight promising future perspectives for the development of a new generation of environmentally-friendly, naturally-derived, smart wound dressings.

Keywords: wound healing, biomimetic, alginate, chitosan, hyaluronic acid, silk fibroin, keratin, antibacterial

### INTRODUCTION

Skin is our major external defense system, in charge of protecting our inner body structures from microorganisms' attacks, and the adverse effects of the external environment. Adult skin is composed of three layers: epidermis or stratum corneum, mainly consisting of keratinocytes; dermis, the connective tissue rich in collagen; and hypodermis or subcutaneous layer, composed of fat tissue, which provides thermal isolation and mechanical protection to the body (Gurtner et al., 2008).

#### Edited by:

*Gianni Ciofani, Politecnico di Torino, Italy*

#### Reviewed by:

*Gozde Ozaydin Ince, Sabanci University, Turkey Satoshi Arai, Waseda University, Japan Filippo Rossi, Politecnico di Milano, Italy*

#### \*Correspondence:

*Giulia Suarato giulia.suarato@iit.it Athanassia Athanassiou athanassia.athanassiou@iit.it*

#### Specialty section:

*This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology*

Received: *27 July 2018* Accepted: *13 September 2018* Published: *02 October 2018*

#### Citation:

*Suarato G, Bertorelli R and Athanassiou A (2018) Borrowing From Nature: Biopolymers and Biocomposites as Smart Wound Care Materials. Front. Bioeng. Biotechnol. 6:137. doi: 10.3389/fbioe.2018.00137* Wounds are breaks or defects within the skin, which may form due to physicochemical or thermal damage. Acute wounds define injured tissues that need a healing period over 8–12 weeks, (e.g., burns, chemical injuries, cuts). In contrast, chronic wounds are a fallout of diseases, such as venous or arterial vascular insufficiency, pressure necrosis, cancer, and diabetes (Sen et al., 2009; Moura et al., 2013). They require longer healing time (weeks-months to years) and often fail to reach a normal healthy state, persisting in a pathological condition of inflammation (Guo and Dipietro, 2010). Therefore, delayed or impaired wound healing poses a significant socio-economic burden on patients and health care systems worldwide, in terms of treatment costs and waste production (Sen et al., 2009).

Insight into the intricate biochemical events activated during skin repair is crucial to design appropriate wound dressings (Weller and Sussman, 2006; Gurtner et al., 2008; Pereira et al., 2013). The healing process can be divided into the following, overlapping stages: homeostasis, inflammation, proliferation, and remodeling (Martin, 1997; Gurtner et al., 2008; Bielefeld et al., 2013; Das and Baker, 2016). Homeostasis is the immediate response of the body to an injury, in order to stop blood loss at the wound site, by means of fibrin cloths as temporary barriers (Sinno and Prakash, 2013). Inflammation (form 24 h to 4–6 days) is mediated by neutrophils and macrophages (Broughton et al., 2006), that sweep the wound bed from foreign particles and tissue debris. Cytokines and enzymes are released to stimulate fibroblasts and myofibroblasts (Das and Baker, 2016), while the wound exudate provides the essential moisture for the recovery. In the proliferation phase epithelialization occurs and newly formed granulation tissue begins to fill the wound area, producing new extracellular matrix (ECM). Finally, during the remodeling phase, collagen-based cross-linking is responsible for a tight 3D network formation, increasing the tensile strength of the new tissue (Sinno and Prakash, 2013).

Given the multiple mechanisms involved in the skin wound healing and the interplay of several external factors, the choice of suitable dressing materials is compelling. Specifically, for biodegradable natural materials, their degradation needs to follow the dynamics of the wound repair, guaranteeing the physiological healing evolution, and releasing active principles when needed. At last but not least, proper consideration should be put onto the environmental sustainability of these biomaterials, in terms of green chemistry fabrication approaches, and complete biodegradation without harmful by-products. While numerous reviews on traditional wound dressing biomaterials have been extensively published (Sell et al., 2010; Mogo¸sanu and Grumezescu, 2014; Norouzi et al., 2015; Mele, 2016), in this mini-review we will focus our attention on the most recent naturally-derived, active systems, pursuing the quest for an environmentally sustainable wound management.

### MIMICKING NATURE AS A THERAPEUTIC STRATEGY

Successful wound management relies on understanding the healing process combined with a know-how on the properties of the various dressing materials available. Principal purpose of any wound treatment is to maximize the treatment efficiency (Weller and Sussman, 2006). Currently, standard care procedures consist of swabbing the infection, cleaning the wound bed from tissue debris, and applying the dressing (Dreifke et al., 2015). In case of extended skin lesions, the use of split-thickness skin autografts or allografts might be required, carrying safety issues related to disease transmission and immune rejection.

An ideal dressing should remove excessive exudate to avoid tissue maceration and promote autolytic debridement, while keeping moisture, adequate oxygen and water vapor permeability within the wound. It should be adhesive and flexible, to favor mechanical compliance to the patient body and ease the application/removal. Deliverable bioactive compounds, such as antibiotics, essential oils, and natural antioxidants, stimulate the dressing interaction with the wound microenvironment and further enhance the therapeutic action via antimicrobial, antifungal, and antiseptic activities (Pereira et al., 2013). A number of fabrication techniques, such as film-casting, electrospinning, self-assembly, freeze-drying, emulsions, microsphere injection, have been employed to produce wound dressings, either based on synthetic macromolecules or on materials of natural origin (Sell et al., 2007, 2010; Wei and Ma, 2008; Huang and Fu, 2010; Zhong et al., 2010; Rieger et al., 2013; Mogo¸sanu and Grumezescu, 2014; Norouzi et al., 2015; Liakos et al., 2016; Mele, 2016).

Recently, natural biopolymers have largely attracted the scientific community interest. On top of their notable biocompatibility and biodegradability, natural occurring proteins and polysaccharides allow to achieve the highest level of biomimicry, recapitulating the native ECM biological and physico-chemical features. Further architectural resemblance can be obtained with an appropriate processing (e.g., nanofibers, sponge-like hydrogels; Huang and Fu, 2010; Mogo¸sanu and Grumezescu, 2014; Liakos et al., 2015; Mele, 2016). Despite their batch-to-batch variations in terms of mechanical properties and their rather limited shelf-life, naturally-derived biopolymers confer ECM support (collagen, gelatin, hyaluronic acid), present cell-recognition domains and biomolecule binding sites (RGD and LDV sequences in silk fibroin and keratin), and may possess inherent antibacterial and anti-inflammatory properties (chitosan, alginate). Moreover, in the past decades, clinical understanding advancements have directed significant exploitation of natural materials in clinical trials (Vyas and Vasconez, 2014; Dhivya et al., 2015). By looking into our natural surroundings and by re-using some of the discarded natural resources, several functional biomaterials can be easily identified and implemented for promising wound healing applications, with a reduced impact on the environment (**Figure 1**).

## PROTEIN-BASED BIOPOLYMERS

## Collagen and Gelatin

Collagen is the most abundant animal protein, which provides mechanical strength to tissues and stimulates cell-adhesion and proliferation (Neel et al., 2013; An et al., 2016). Twentynine different types of collagen have been identified, displaying a triple-helical tertiary structure of polypeptide sequences (**Figure 2a**), but only a few are used in the production of

collagen-based biomaterials. As animal-derived proteins may be responsible for allergic reactions and pathogen transmissions (Koide, 2007), an alternative is constituted by collagen from heterologous expression in mammalian, insect and yeast cells (Olsen et al., 2003), or produced by Escherichia coli (Pinkas et al., 2011). High biocompatibility and biodegradability by endogenous collagenases make collagen ideal for biomedical applications (Parenteau-Bareil et al., 2010; Chattopadhyay and Raines, 2014). During wound healing, fibroblasts produce collagen molecules that aggregate to form fibrils with diameter in the range of 10–500 nm. This fibrous network facilitates cell migration to the wounded site, actively supporting tissue repair (Baum and Arpey, 2005).

Thanks to a facile chemical functionalization of the protein structure, various dressing architectures have been exploited. Collagen-based wound dressings, either in forms of hydrogels, electrospun fibers, or nanocrystal-containing scaffolds, have been applied to cover burn wounds, treat ulcers (Ghica et al., 2017; Guo et al., 2017; Bhowmick et al., 2018; Yoon et al., 2018), reduce tissue contraction and scarring, and increase epithelialization rate (Powell et al., 2008). Collagen sponges and fibrous membranes were found particularly promising, due to their wet-strength that allows suturing to soft tissues and provides a template for new tissue growth. Composites with other natural materials, such as dextran, chitosan, hyaluronic acid, and alginate (Karri et al., 2016; Ghica et al., 2017; Wei et al., 2018) or constructs based on collagen and synthetic biopolymers, such as poly-α-hydroxyl esters (Hall Barrientos et al., 2017; Albright et al., 2018; Bhowmick et al., 2018) have been extensively exploited. Moreover, acetylated, succinylated, methylated, or biotinylathed collagen have been used to immobilize therapeutic enzymes or growth factors and to control drug delivery (Lima et al., 2015; Mele, 2016; Qu et al., 2018; Zhu et al., 2018). Albright and coworkers (Albright et al., 2018) proposed a multi-structured nanofibrous dressing, composed of poly-ε-caprolactone/collagen electrospun matrix, loaded with transforming growth factor TGF-β1 and modified with polypeptide-based nanocarriers incorporating tannic acid and gentamicin. The multifunctional platform showed anti-bacterial and anti-inflammatory properties, while retaining a favorable topography for cell proliferation, thus accelerating healing and wound closure. A similar construct was proposed by Karri et al. (2016), where a composite scaffold of collagen and alginate was impregnated with curcuminloaded chitosan nanoparticles to obtain an all-natural wound dressing.

A collagen-derivative with promising biomedical values is gelatin. Gelatin is obtained by an incomplete denaturalization of collagen extracted from connective tissues, skin, and boiling bones (Jaipan et al., 2017). It has been employed to fabricate strong hydrogel-like membranes (Thanusha et al., 2018), microspheres (Thyagarajan et al., 2017), sponges, and electrospun mats (Chen et al., 2016), for dermal tissue applications and to treat severe burn wounds. Various combinations of gelatin and modified chitosan have been proposed (Han et al., 2014; Agarwal et al., 2016), as well as blends with poly-vinyl alcohol based via enzymatic crosslinking, to support fibroblast culture and proliferation (Hago and Li, 2013).

Despite their rather extensive usage as biomaterials for scaffold design, collagen and gelatin remain sustainable materials with highly engineering potential yet unexplored (Hall Barrientos et al., 2017; Golser et al., 2018).

(e,f) wool keratin sponges, reprinted from Patrucco et al. (2016) Copyright©2016 with permission from Elsevier; (g) calcium cross-linked alginate beads and (h) film incorporating antiseptic PVPI complex, reprinted from Liakos et al. (2013) Copyright©2013 with permission from Elsevier; (i,l) mycelia material from *P. ostreatus* after 20 days of growth on potato-dextrose broth and cellulose, presenting a 3D network of hyphae. Panels (i,l) are adapted from Haneef et al. (2017). This material is licensed under the Creative Commons Attribution 4.0 International Public License (https://creativecommons.org/licenses/by/4.0/legalcode).

### Silk Fibroin

Silks are proteins produced in the epithelial cells of specialized glands of various arthropods, such as spiders and silkworms. The secreted silk fibers present a highly repetitive sequence, consisting mainly of glycine (43%), alanine (30%), and serine (12%)—[GAGSGA]<sup>n</sup> motifs, arranged in β-sheets regions embedded in an amorphous matrix (Chutipakdeevong et al., 2013; Reimers et al., 2015), which confer high toughness and elasticity (**Figure 2a**). Regarding more specifically silk from the cocoons of Bombyx mori silkworms, two kinds of proteins are its major components: the fibroin and the sericin. The fibroins are composed of three types of protein fibers: 350 kDaheavy chain, 30 kDa-glycoprotein, and 25 kDa-light chain, the latter conferring hydrophilicity, water uptake ability and cell adhesion properties. Light and heavy chains are connected by disulfide bonds, while the glue protein sericin coats the silk fibers. Thanks to high mechanical resistance, enzymatic-driven biodegradability and favorable cell attachment, silk fibroins have been successfully exploited for skin tissue engineering and wound healing applications.

Bombyx mori silkworm-derived fibroins are obtained from cocoon and separated from sericin by degumming in alkaline boiling water and following solubilization in hot LiBr solution (Reimers et al., 2015). Regenerated silk water soluble form, silk I, can be converted into insoluble silk II, by modifying the α-helical chain arrangements into β-sheets via alcohol treatment or water vapor annealing (WVA; Min et al., 2006; Wharram et al., 2010; Hu et al., 2011). In fact, by controlling the protein secondary structure, fibroin scaffolds' biodegradation can be properly tuned, in order to modulate the release of bioactive molecules (Hofmann et al., 2006), such as antibiotics (Pritchard et al., 2013; Chouhan et al., 2017), growth factors (Schneider et al., 2009; Chouhan et al., 2017; Pignatelli et al., 2018), and anti-oxidant compounds (Fan et al., 2012; Sheng et al., 2013; Lin et al., 2016). In their study, Pignatelli and coworkers (Pignatelli et al., 2018) encapsulated human platelet lysate into electrospun silk-PEO patches, to prolong the growth factor shelf life and ease its handling during wound management. By changing the crystallinity degree of the fibrous matrices via WVA from 21 to 35 or 44%, the 24-h drug release drastically decreased from 100 to 80 and 46%, respectively.

Proteins' versatile nature allows for a plethora of processing techniques, with the consequent fabrication of multiple scaffold morphologies, such as films (Srivastava et al., 2015), foams and sponges (Roh et al., 2006), gels, and fibrous matrices. In the last decade, electrospun fibroin has been extensively proposed for the design of anti-bacterial, anti-inflammatory and anti-oxidant patches (Lin et al., 2016; Selvaraj and Fathima, 2017; Yang et al., 2017). To ease a water-based electrospinning process, silk has been processed either in combination with natural polymers, such as cellulose (Guzman-Puyol et al., 2016; **Figures 2c,d**), gelatin (Shan et al., 2015), sericin (Hang et al., 2012), chitosan (Cai et al., 2010), alginate (Roh et al., 2006), elastin (Zhu et al., 2016), and hyaluronic acid (Yan et al., 2013), or mixed with synthetic materials, such as polyethylene oxide (Schneider et al., 2009; Wharram et al., 2010; Chutipakdeevong et al., 2013), polyvinyl alcohol (PVA; Chouhan et al., 2017), and polyhydroxy esters (Lian et al., 2014; Shahverdi et al., 2014; Suganya et al., 2014; Shanmugam and Sundaramoorthy, 2015). Silk/PVA mats loaded with Ciprofloxacin and epidermal growth factors (Chouhan et al., 2017) enhanced human dermal fibroblasts and keratinocytes proliferation in vitro, and favored re-epithelization, mature collagen deposition and complete wound closure at 14 days in a in vivo wound healing rabbit model. In a different work, Ju and coworkers (Ju et al., 2016) investigated the intrinsic anti-inflammatory effects of a porous fibroin/PEO electrospun nanomatrix in a mice burn-model, observing downregulation of pro-inflammatory cytokines IL-1α and IL-6.

### Keratin

Keratins (Ker) are the most abundant group of insoluble and filament-forming proteins produced in epithelial cells of mammals, birds, reptiles, and humans. As structural components of wool, nails, horn, feathers, and hair, they exploit mechanical support and protective functions against the environment (Reichl, 2009; Wang et al., 2016). Keratins present a complex intermediate filament (IF)-matrix hierarchical structure and are categorized according to the polypeptide chain secondary assembly. α-Ker (40–68 kDa) comprise α-helices arranged in coiled-coil heterodimers to form 7-nm IF, while β-Ker (10–22 kDa) consist of packed β-sheets disposed in 3-nm IF. The highsulfur containing matrix (γ-Ker, below 10 kDa), rich in cysteine, tyrosine, glycine and phenylalanine residues, present a globular assembly (Dowling et al., 1986; Fraser et al., 1986; Steinert and Marekov, 1993; Rouse and Van Dyke, 2010; Wang et al., 2016). The secondary structure of keratinous materials largely affects their mechanical resistance, solubility, and hydration sensitivity (Wang et al., 2016). Tons of Ker-containing biomasses are produced every year, from meat and poultry market, wool industry and hair salons, leading to continuous accumulation of wastes in the ecosystem. The challenges associated with this waste disposal have been considered by the European Parliament and Council regulation EC 1774/2002 (Sharma and Gupta, 2016). Due to the presence of strong disulfide and H-bonds, keratin extraction from biomasses involves rather complicated methods, such as microbial and enzymatic hydrolysis, mechanical treatments, or chemical protocols with alkali, reducing agents or ionic liquids (Yamauchi et al., 2003; Ozaki et al., 2014; Sharma and Gupta, 2016; Shavandi et al., 2017).

However, thanks to its biocompatibility, biodegradability, and hemostatic properties, keratin constitutes a potential green secondary raw material for wound healing, tissue repair, drug delivery, and cosmetics applications (Sharma and Gupta, 2016; Arslan et al., 2017; Shavandi et al., 2017). Since its earliest documented use in medicinal applications (China, sixteenth century; Rouse and Van Dyke, 2010), in the past decades several Ker-based biomaterials have been proposed, given the ability of this biopolymer to self-assembly into 3D networks favorable for cell infiltration, and its intrinsic bioactivity for the presence of cell binding motifs (such as EDS, LDV, and RGD; Rouse and Van Dyke, 2010). Keratin extracted from chicken feathers, wool and human hair have been processed in films (Yamauchi et al., 2003; Fujii and Ide, 2004; Tonin et al., 2007; Reichl, 2009; Cui et al., 2013), sponge-like and hydrogel-like scaffolds for tissue engineering and wound healing (**Figures 2e,f**; Tachibana et al., 2002; Verma et al., 2008; Hill et al., 2010; Saul et al., 2011; Richter et al., 2012; Wang et al., 2012; Xu et al., 2013; Patrucco et al., 2016; Singaravelu et al., 2016). Electrospun fibers have also been obtained in combination with PEO (Aluigi et al., 2007, 2008; Fan et al., 2016; Ma et al., 2017), PVA (Choi et al., 2015; He et al., 2017; Wang et al., 2017), fibroin (Zoccola et al., 2008; Yen et al., 2016), poly-caprolactone (Boakye et al., 2015; Edwards et al., 2015; Li et al., 2016; Zhu et al., 2017), poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (Yuan et al., 2015), chitosan (Singaravelu et al., 2016), and gelatin (Yao et al., 2017).

### NATURALLY-DERIVED POLYSACCHARIDES

### Hyaluronic Acid

Hyaluronic acid (HA) is a non-immunogenic polysaccharide consisting of glucuronic acid and N-acetyl-D-glucosamine units (**Figure 2b**). This glycosaminoglycan is one of the main components of the connective tissue in mammals (Mele, 2016). Due to its hygroscopic nature, HA has been used to prepare hydrogel-like constructs, to support keratinocyte migration and angiogenesis, and promote a scar-free wound healing (Mogo¸sanu and Grumezescu, 2014; Dreifke et al., 2015). The molecular weight (MW) plays a key role in the process (Tolg et al., 2014): low MWHA degradation products were found to be proinflammatory (Campo et al., 2010; Dreifke et al., 2015), while high MWHA appeared to inhibit nutrient supply. Interestingly, medium MWHA (100–300 kDa) showed enhanced wound closure capability through up-regulation of adhesion molecules (Ghazi et al., 2012). Moreover, the hydrophilicity of the HA chains allows the 3D network swelling and the consequent gradual release of encapsulated active compounds, making this biomaterial suitable as drug delivery platform (Maeda et al., 2014). HA-based electrospun fibers, either pure or in combination with other biomacromolecules (Xu et al., 2009; Hsu et al., 2010; Uppal et al., 2011; Dogan et al., 2016), have been proposed for tunable degradation and sustained release in vitro and in vivo.

### Chitosan

Chitosan (CS), a deacetylated chitin-derivative found in the exoskeletons and shells of crustaceans, is a linear polysaccharide consisting of β(1-4)-D-glucosamine and N-acetyl-D-glucosamine groups randomly distributed (**Figure 2b**). Owing to its intrinsic antifungal, antibacterial, hemostatic, and muco-adhesive properties, chitosan has been widely exploited in the biomedical field for wound and burn treatments (Dash et al., 2011; Croisier and Jérôme, 2013; Norouzi et al., 2015; Zhao et al., 2015). Several dressing architectures have been proposed: CS-Aloe vera membranes (Wani et al., 2010), thyme oil-CS films (Altiok et al., 2010), CS-gelatin sponges (He et al., 2007), CS-silk hydrogels (Silva et al., 2012), CS-cellulose films (Niyas Ahamed and Sastry, 2011; Romano et al., 2015a), cinnamon oil-CS/polyethylene oxide nanofibers (Rieger and Schiffman, 2014), and CS/poly(3-hydroxybutyrateco-3-hydroxyvalerate) scaffolds (Veleirinho et al., 2012). In addition, water-soluble derivatives, such as carboxymethyl-CS and methacrylate glycol CS have been synthetized and investigated for wound healing applications (Romano et al., 2015b).

### Alginate

Alginate (Alg) is a linear co-polymer of β-D-Mannuronic acid and α-L-Glucuronic acid (**Figure 2b**). This polysaccharide is mostly abundant in Brown Algae or produced by some bacteria (Khan and Ahmad, 2013). It is highly hydrophilic, biocompatible, and able to absorb wound exudate, maintaining a moist microenvironment (Chiu et al., 2008). The combination of alginate with antimicrobial and enzymatic components can promote elimination of necrotic tissues and microbial bodies, while the polysaccharide base can stimulate reparative wound processes (Patel et al., 2007). Alginate dressings are also useful as delivery platforms, in order to provide a controlled release of therapeutic substances to exuding wounds (e.g., pain-relieving, antibacterial, and anti-inflammatory agents; Maver et al., 2015; Szekalska et al., 2016; Setti et al., 2018). Biodegradable Na-Alg/PVPI (povidone iodine complex) films and Ca-Alg/PVPI beads have displayed antimicrobial and antifungal activities (Liakos et al., 2013; **Figures 2g,h**). Moreover, Na-Alg/PVPI films have shown to reduce the inflammatory response and accelerate the wound healing providing a controlled release of PVPI (Summa et al., 2018). To treat UV-induced skin burns, instead, electrospun nanofibers loaded with lavender essential oil have been used: the composite mats exhibited antibacterial and antiinflammatory properties, being able to reduce the production of pro-inflammatory cytokines both in vitro and in vivo (Hajiali et al., 2016).

## SELF-GROWING MYCELIUM-BASED BIOMATERIALS

Mycelium, the fungi vegetative part, comprises a network of filamentous hyphae, which penetrate the substrate. Hyphae are tubular structures of micrometric diameter, composed of aligned and elongated cells, separated by walls, called septa (**Figures 2i,l**). A continuous cell wall protects the hyphae and confers mechanical strength and shape to the mycelium (Haneef et al., 2017; Jones et al., 2017). Being constituted of chitin, chitosan, glucans, mannoproteins, and glycoproteins (Synytsya and Novák, 2013), the cell wall is a biopolymer composite that prevents the hyphae from collapsing during their sprouting and movement (Cairney, 2005).

Peculiarity of these living, self-growing composites is the possibility to tune the physico-chemical properties during their growth phase, reducing sophisticated processing and byproduct formation, while allowing for ready-to-use systems. Throughout its dynamic growth, the mycelium "senses" the substrate and responds to the surrounding, depending on edaphic conditions, substrate pH and composition, or the presence of other living organisms (Krull et al., 2013). A polarized extension of the cell wall occurs at the apical region of the hyphae, as the mycelium secretes a variety of enzymes, hydrolyzes the substrate and absorbs the solubilized nutrients. By properly exploiting different feeding substrates for hyphae digestion, the resulting properties of the interwoven fibrous mycelium material can be efficiently tailored. Type and amount of absorbed nutrients may affect mycelia growth rate, extension and biological activity (Frimpong-Manso et al., 2011; Da Silva et al., 2012; Larsen, 2015; Anderson and Cairney, 2018). Similarly, culture conditions and feeding substrate highly influence the final chemical composition (Krull et al., 2013; Haneef et al., 2017), either stimulating plasticizer biosynthesis (lipids and small glycoproteins) or promoting rigid macromolecule production (chitin, β-D-glucans; Synytsya and Novák, 2013). In this regard, mycelia can be considered as 3D smart, micro-reactors, able to bio-convert various agro-residues into enzymes, polysaccharides, and bioactive metabolites (Vassilev et al., 1995; Krull et al., 2013; Vamanu, 2014; Yan et al., 2014; Salati et al., 2017), with potential bio-pharmaceutical and neutraceutical relevance.

In the past, mycelia-derived scaffolds have been exploited (Su et al., 1999, 2005; Hung et al., 2001). Su et al. (1997) developed a filament-structured membrane from the residue of Ganoderma tsugae, called Sacchachitin, composed of β-1,3-glucan (60%), and N-acetylglucosamine (40%), to be used as skin substitute. The new biomaterial demonstrated wound healing potential in vivo, by promoting fibroblast proliferation and migration. In a following study (Su et al., 2005), the Sacchachitin membranes appeared to boost keratinocytes proliferation and prevent metalloproteinase-related ECM degradation, contributing to accelerate the healing process of a chronic wound in vivo model. Furthermore, a micronized Sacchachitin nanogel has been investigated to treat superficial chemical corneal burns in vivo (Chen et al., 2012), while the anti-oxidant and immunomodulating effects of extracts from mycelia of some medicinal fungi have been investigated for skin aging (Kim et al., 2014), dermatitis (Hwang et al., 2012), and UV-protection (Nanbu et al., 2011; Bae et al., 2012), suggesting a promising mycelia biological value yet unexplored.

### CONCLUSIONS - A QUEST FOR AN ENVIRONMENTALLY SUSTAINABLE WOUND MANAGEMENT

Environmental sustainability has nowadays become an imperative issue to front, in an effort to balance both the industrial productivity and the planet ability to generate resources, with the neutralization of wastes and the mitigation of polluting processes. Material and energy consumption related to the healthcare industry, ranging from complex material manufacturing processes, to drug packaging, to high-volume

### REFERENCES


medical wastes, might heavily contribute to increase the overall pollution, with an unsought negative impact on the human health (Jameton and Pierce, 2001). Nature itself can be of inspiration to develop cost-competitive, low-energy consumption and fully biodegradable materials, presenting greater environmental sustainability. The increasing interest of the scientific community in the use of either protein-based or polysaccharide-derived dressings is striking, and it reflects the growing perspective of giving back what we borrowed from Nature. In this regard, self-growing mycelia, which are biocomposites constituted of both proteins and polysaccharides, can represent a smart strategy to fabricate healthcare products of the future, as economically and environmentally valid alternatives to synthetic materials. In conclusion, naturally derived biopolymers provide a versatile, multifunctional, and tunable platform to design appropriate extracellular environments, able to actively contrast the onset of infections and inflammations, while promoting tissue regeneration, and scar remodeling.

### AUTHOR CONTRIBUTIONS

GS and RB conceived and wrote the main manuscript text. AA edited and reviewed the whole manuscript.


on corneal epithelium. Int. J. Nanomed. 7, 4697–4706. doi: 10.2147/IJN.S 34530


of biocompatible surfaces. ACS Biomater. Sci. Eng. 4, 2106–2114. doi: 10.1021/acsbiomaterials.7b00583


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Suarato, Bertorelli and Athanassiou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# One-Step Synthesis of Hexagonal Boron Nitrides, Their Crystallinity and Biodegradation

Özlem ¸Sen† , Melis Emanet † and Mustafa Çulha\*

Department of Genetics and Bioengineering, Faculty of Engineering, Yeditepe University, Istanbul, Turkey

Hexagonal boron nitrides (hBNs) have recently been investigated for several novel applications due to their unique properties such as biocompatibility, superhydrophobicity, electrical insulation, and thermal and chemical stability. In addition, their biodegradation products have recently reported to have therapeutic effect on certain cancer types. hBNs are easily synthesized from boron and nitrogen precursors at moderately low temperatures. However, crystallinity and yield vary depending on the type of precursor, reaction temperature, and duration. In this study, a simple one-step hBNs synthesis method is reported without a catalyst, which might be an undesired contaminant for biomedical applications. The influence of boron precursors (boric acid, colemanite, or boron trioxide) on hBNs crystallinity, stability, and biodegradation in suspensions containing oxidative and hydrolytic degradation agents is investigated with the aim of their possible application in biomedicine. We found that the choice of boron precursor is a critically important parameter controlling the hBNs crystallinity and dependently influencing the biodegradation rate.

#### Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Attilio Marino, Fondazione Istituto Italiano di Technologia, Italy Kaan Kececi, Istanbul Medeniyet University, Turkey Yuriy Bogdanovich Stetsyshyn, Lviv Polytechnic, Ukraine

\*Correspondence:

Mustafa Çulha mculha@yeditepe.edu.tr

†These authors have contributed equally to this work.

### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

> Received: 29 April 2018 Accepted: 01 June 2018 Published: 21 June 2018

#### Citation:

¸Sen Ö, Emanet M and Çulha M (2018) One-Step Synthesis of Hexagonal Boron Nitrides, Their Crystallinity and Biodegradation. Front. Bioeng. Biotechnol. 6:83. doi: 10.3389/fbioe.2018.00083 Keywords: hexagonal boron nitride, synthesis, biodegradation, boric acid, boron trioxide, colemanite

### INTRODUCTION

In recent years, boron nitride (BN) based nanomaterials have grasped attention of researchers for their possible use in medical applications due to their biocompatibility (Chen et al., 2009; Salvetti et al., 2015; Li et al., 2017), chemical and mechanical stability (Lahiri et al., 2011; Liu et al., 2016). Since they are composed of boron and nitrogen, they can be excellent needed as boron and nitrogen source in biosystems. Recently, it was demonstrated that BNs can be used as a boron source to promote wound healing and treat prostate cancer, and their degree of crystallinity is an important factor due to its influence on biodegradation (Dousset et al., 2002; Li et al., 2017).

Although hexagonal boron nitrides (hBNs) were discovered in 1842, their first stable form was obtained about a century later. The BNs do not naturally occur, thus they must be synthetically synthesized from boron and nitrogen precursors to prepare structures analog to their counterpart, graphene (Arenal and Lopez-Bezanilla, 2015). In hBNs, boron and nitrogen atoms are covalently bound in a hexagonal structure and their layers are stacked on top of each other. Similar to graphene, the two-dimensional (2D) BN layers interact with each other through van der Waals forces (Du Frane et al., 2016), and constitute hBN films (Shi et al., 2010) or spherical nanostructures (Bhimanapati et al., 2014), Given their excellent physical and chemical properties, hBNs are considered as promising nanomaterials in many technological applications (Han, 2010).

The use of hBNs in a range of applications was reported in the literature including their exploitation as 2D dielectric thin films and deep UV emitters at around 215 nm (Kubota et al., 2007; Song et al., 2010). It was also reported that composite of hBNs with polyether ether ketone showed improved mechanical and thermo mechanical properties, since hBNs possess high elastic modulus, excellent lubrication properties, and good thermal conductivity (Liu et al., 2016). In last few years, the hBNs are also gaining attention as a promising nanomaterial in medical and biomedical applications due to their high biocompatibility (Sukhorukova et al., 2015). Their use as a tablet lubricant in pharmacy and additive in cosmetics are also encouraging points for their consideration in the biomedical field (Turkoglu et al., 2005; Shi et al., 2010; Liu et al., 2016). In our recent study, their cytotoxicity was assessed, and conjugation with doxorubicin and folate has been proposed to enable their potential in cancer therapy (Emanet et al., 2017).

There are several hBNs synthesis methods reported in the literature by using a variety of boron and nitrogen precursors under different experimental conditions (Singhal et al., 2008). The synthesis methods influence reaction yield, number of layers, shape, and size of hBNs (Li et al., 2011). In one study, boric acid and urea were mixed and heated under ammonia atmosphere to produce hBNs (Chakrabartty and Kumar, 1995). In another, it was reported that hBN films were synthesized using diborane and ammonia as gas precursors, and the number of hBN layers can be controlled at various temperatures, gas pressure and flow rates (Ismach et al., 2012). Moreover, spherical BNs were synthesized using trimethoxyborane [B(OMe)3] under ammonia gas flow via chemical vapor deposition (CVD) method (Tang et al., 2008). It was found that the spherical morphology of the BNs strongly depends on the elimination yield of Me2O groups from the BN structures during the intermediate phase of the synthesis (Tang et al., 2008). In addition, hollow BN spheres were synthesized using B(OMe)<sup>3</sup> and Ar instead of ammonia in the second stage of the synthesis that is the key point to form hollow structures. Furthermore, an increase in post-treatment temperature decreases the wall thickness of the hBNs while increasing the crystallinity of the structures (Li et al., 2017).

The studies reporting the inhibitory effect of boron containing compounds on cancer cell proliferation including prostate and breast cancer by reducing the release of stored Ca2<sup>+</sup> ions into the cytosol of cells suggests that boron compounds can be promising agents for cancer treatment (Henderson et al., 2009). However, the frequent administration of boron compounds due to their short half-life for systemic administration limits their use because it requires continues administrations. Since boron-based nanomaterials such as hBNs can behave as a controlled boron release source (Li et al., 2017). Their use can open up new venues in cancer treatment. Therefore, it is important to synthesize hBNs as pure as possible with desired crystallinity, which might also be an important factor for further chemical modification with targeting ligands. These mentioned points could also be important factors for their dissolution/degradation in biological medium for a successful therapeutic approach (Barranco and Eckhert, 2004, 2006; Scorei et al., 2008; Emanet et al., 2015; Li et al., 2017).

In this study, from three different boron compounds, boric acid, colemanite, and boron trioxide were used to synthesize hBNs under ammonia gas atmosphere with CVD technique. The synthesized hBNs were characterized with imaging and spectroscopic techniques. The colloidal stability of the hBNs was investigated by monitoring zeta potential and time dependent size distribution with dynamic light scattering (DLS) technique. Thermogravimetric analysis (TGA) was carried out to observe the resistance against heat decomposition. Finally, the biodegradation behavior of the hBNs in oxidative and hydrolytic degradation conditions was assessed by TGA, ICP-MS, and Raman spectroscopy.

### MATERIALS AND METHODS

### hBNs Synthesis

Boric acid, colemanite, and boron trioxide, as boron and ammonia as nitrogen source were used. The synthesis was performed based on a method previously reported by our group (Emanet et al., 2017). First, 2 g of boric acid or colemanite or boron trioxide were suspended in 3 mL of 13.38 M ammonia solution. This mixture was transferred onto a silicon carbide boat and dried on a hot plate adjusted to 100◦C for approximately 20 min. Then, this silicon carbide boat was placed in a Protherm Furnace PTF 14/50/450 and heated until 1,300◦C with a heating rate of 10◦C/min under ammonia gas flow for 2 h. Following the heating, the silicon carbide boat was removed from furnace at around 550◦C and hBNs were scratched from the surface of the silicon carbide boat with the help of spatula and stored under room conditions.

### SEM and TEM Imaging

The morphology and size of the synthesized hBNs were characterized using SEM and TEM. SEM (Helios Nano-Lab 600i FIB/SEM, FEI) imaging was carried out on samples previously gold-sputtered for 25 s at 60 nA, obtaining a 3-nm thick conductive layer over the hBNs. TEM images were acquired with a JEOL-2100 HRTEM microscopy system at 200 kV (equipped with LaB6 filament and an Oxford Instruments 6498 EDS system).



### UV-Vis, XRD, FT-IR, and Raman Spectroscopy

The hBNs were dispersed in double distilled water (ddH2O) by sonication for 2 min at before the analysis (Bandelin Sonopuls HD 3100). A Perkin Elmer Lambda 25 UV-Vis spectrometer was used to obtain absorption spectra. IR spectra were acquired with a Thermo NICOLET IS50 Spectrometer. XRD analysis was performed using a Shimadzu XRD-6000 with a ICDD PDF 4 software. The scanning area was in continuous mode with a scanning range of 2.000–69.980◦ and a scanning speed of 2.0000◦ /min. The sampling pitch was set to 0.0200◦ , and the preset time was set to 0.60 s. Raman spectra of the hBNs were recorded using a Renishaw In Via Reflex Raman Microscopy system (Renishaw Plc., New Mills, Wotton-under-Edge, UK) equipped with a 514 nm Argon ion laser. A minimum of 16 spectra was acquired from a 16-µm<sup>2</sup> hBNs sample area. All measurements were performed at least three times.

### Thermogravimetric Analysis (TGA)

TGA analyses were performed using a Mettler Toledo TGA/SDTA 851 instrument. The samples were analyzed under 20-mL/min N<sup>2</sup> gas flow. The temperature was increased up to 700◦C and was set at 10◦C/min.

### Dynamic Light Scattering (DLS) and Zeta Potential Measurements

The colloidal stability analysis of the hBNs was carried out by monitoring the size distribution and the Z-potential using a Malvern Zetasizer Nano ZS. The hBNs were sonicated for optimal dispersion in ddH2O at for 2 min. The concentration of the hBNs was fixed as 1 mg/mL. Then, the sonicated samples were analyzed using DLS at different times (0 and 60 min) to assess their time-dependent colloidal stability. Furthermore, zeta potential measurements were performed and all experiments were repeated at least three times.

### Biodegradation Studies

Biodegradation of hBNs was performed under two different conditions: in lysosome mimicking solution (LMS) for oxidative degradation, and in phosphate-buffered saline (PBS) for hydrolytic degradation. To observe the oxidative degradation of hBNs in lysosome mimicking condition, a LMS was prepared according to **Table 1** (Russier et al., 2011). The pH of the solution was brought to 4.5. Then, 25 mg of hBNs were dispersed in 25 mL of LMS, which contains physiological concentration of hydrogen peroxide (H2O2, 1 mM). Then, they were sonicated for 1 h. To maintain the physiological oxidizing environment of lysosomes, 1 mM of H2O<sup>2</sup> was weekly added. For hydrolytic degradation of hBNs in PBS, 25 mg hBNs were dispersed in 25 mL PBS and



sonicated for 1 h. After the sonication, all samples for oxidative and hydrolytic degradation were placed in dark and incubated at 37◦C while shaking at 180 rpm up to 30 days. In both cases, 1 mL sample was taken from the LMS or PBS suspensions at 0, 1, 3, 7, 14, and 30 days of the incubation, and stored at 4◦C in the dark until characterization.

### Characterization of Degradation Products

After the degradation process in the LMS and PBS suspensions, samples taken from the suspensions were used for TGA analysis. An unprocessed hBNs suspension was used as a control for comparison. The suspensions were centrifuged at 10,000 rpm for 10 min, the supernatants were removed, and the pellets including hBNs were dried under vacuum. The TGA analysis was carried out using the pellets.

The degradation products were evaluated with ICP-MS to determine boron content. First, 1 mL of the samples was centrifuged at 10,000 rpm for 10 min at 4◦C to remove nondegraded hBNs from solution as a pellet. Then, 0.5 mL of the supernatant containing degradation products was mixed with 4.5 mL of 1% nitric acid solution. Thereafter, the samples were filtered with a 0.22µm Millipore filter to further remove any non-degraded hBNs from the samples, which were evaluated with an X Series 2 ICP-MS (Thermo Scientific) instrument equipped with a CETAC asx-520 autosampler. In the ICP-MS system, the plasma power, the plasma gas, the nebulizer gas flow rate, the clutch duration, and the wash duration were set to 1,350 W, Ar, 0.95, 35, and 35 s, respectively. VHG Labs Z frequency 1007-100 multi-element standard stock solution (1 mg/mL) containing Al, B, Cu, Ag, As, Cd, Fe, Ni, Sr, Zn, Ca was used for calibration. The standards were prepared at the concentrations of 0.1, 1.0, 10.0, and 100.0µg/mL in 5% nitric acid using this stock solution. The calibration curves were created for each metal and the experiments were performed at least three times.

The characterization of hBNs after the degradation process was further analyzed by Raman spectroscopy. Five microliters of LMS and PBS suspensions containing hBNs were placed on a CaF<sup>2</sup> slide and allowed to dry. A minimum of 16 spectra was recorded from a dried droplet area by choosing arbitrary points. All experiments were performed at least three times and their average values were calculated.

## RESULTS AND DISCUSSION

### Morphological Characterization of hBNs

The size and morphology of hBNs can vary depending on synthesis conditions and the type of boron precursor. **Figure 1** shows the SEM and TEM images of the hBNs synthesized from three different boron precursors. As seen, the hBNs synthesized from each precursor have platelet-like morphology with varying sizes. The hBNs synthesized from boric acid (hBNs\_boric acid) show a more uniform structure with lateral size dimension between 50 and 70 nm as seen in **Figure 1A** while the lateral size dimensions of the hBNs synthesized from colemanite (hBNs\_colemanite) are between 50 and 80 nm. The hBNs\_colemanite also include boron nitride nanotubes (BNNTs) in the product mixture (**Figure 1B**). This can be explained with the presence of metal oxides such as CaO, B2O3, Fe2O3, Al2O3, MgO, and Na2O in colemanite acting as catalysts (Gür, 2007). Transition metals and their oxides play an important role in synthesis of BNNTs as catalysts to roll the hBN layers into tube shape (Pakdel et al., 2012). The hBNs synthesized from boron trioxide (hBNs\_boron trioxide) show non-uniform lateral sizes in the range of 200–300 nm indicating a poor regularity as seen in **Figure 1C**. **Figures 1D–F** show TEM images of the synthesized hBNs. As seen in **Figure 1D**, the hBNs\_boric acid present uniform platelet-like structures while both hBNs and incomplete BNNTs are visible in the hBNs\_colemanite sample reported in **Figure 1E** (inset image) similar to the case observed in SEM images. **Figure 1F** shows that the hBNs\_boron trioxide have varying sizes confirming their structures as observed from their SEM images. **Figures 1G–I** show the parallel straight-line like crystalline features of the synthesized hBNs indicating the high quality of the samples. The honeycomb lattice structure of atomic boron and nitrogen was observed using HRTEM as shown in **Figure 1H** as inset image.

### Spectroscopic Characterization of hBNs

The hBNs were characterized with UV-Vis, XRD, FT-IR, and Raman spectroscopy. **Figure 2A** shows comparison of UV-Vis spectra of hBNs synthesized from all three precursors. A characteristic maximum absorption band at around 210 nm is observed on the spectra. The highest absorption and background intensity are observed with the hBNs\_colemanite probably due to the presence of BNNTs and other impurities in the final product (Behura et al., 2015).

**Figure 2B** shows the XRD pattern of the synthesized hBNs. The peaks originating from hBNs were observed at 2θ angles of 26.8◦ and 41.8◦ indicating successful synthesis for all hBNs derived from three different precursors. The synthesized hBNs show high crystallinity exhibiting narrow peak width around 26.8◦ and 41.8◦ . However, hBNs\_colemanite show the highest crystallinity with respect to the hBNs\_boric acid and hBNs\_boron trioxide characterized with the increased peak intensities. Furthermore, hBNs\_colemanite have varying peaks around 30–40◦ attributed to characteristic peaks of colemanite (Bayca et al., 2014).

The FT-IR spectra show that the hBNs have broad peaks at around 1,364 and 820 cm−<sup>1</sup> that is attributed to the B–N vibrations as shown in **Figure 2C**. Moreover, a weak band at around 3,400 cm−<sup>1</sup> attributed to O–H stretching in hBNs\_boric acid and hBNs\_boron trioxide, which is an important indicator for the degradation tendency of samples (Tang et al., 2008).

**Figure 2D** shows the Raman spectra of the hBNs with a sharp peak at around 1,364 cm−<sup>1</sup> originating from B–N vibrations. The spectroscopic evaluation of the hBNs obtained from the boric acid, colemanite, and boron trioxide shows the unique characteristic spectral features for hBN structure.

### Colloidal Stability of hBNs

Zeta potential of nanomaterials can be used as an indicator for their colloidal stability in aqueous environment. It is a fact that neutral and low density of charged molecular structures or nanomaterials cannot resist to attractive forces and form aggregates while highly charged particles repel each other and show better stability (Zhang et al., 2008). The zeta potentials of the synthesized hBNs in this study were measured and shown in **Table 2**. As seen, the zeta potential values are −20.72 ± 0.11 for the hBNs\_boric acid, −18.02 ± 0.07 for the hBNs\_colemanite, and −18.96 ± 0.08 for the hBNs\_boron trioxide. These values indicate that the stability of colloidal suspensions of hBNs in aqueous environment is quite good.

In addition, their hydrodynamic stabilities were investigated by measuring their particle size distributions at different timepoints from 0 to 60 min with DLS (**Figure 3**). The samples were sonicated for 2 min for an optimal dispersion. The size

distribution of the hBNs\_boric acid was found between 50 and 91 nm as shown in **Figure 3A**, and the maximum hydrodynamic size increased to 78 from 58 nm after 60 min. The size distribution of the hBNs\_colemanite was found to be between 78 and 190 nm as shown in **Figure 3B**, and their maximum size was also approximately increased to 141 from 122 nm. The size distribution of hBNs\_boron trioxide was between 78 and 164 nm as shown in **Figure 3C**, and their maximum size approximately increased to 122 from 105 nm after 60 min.

The results indicate that the hBNs\_boric acid are more efficiently dispersed in aqueous environment considering their smaller average size as compared to the hBNs\_colemanite and hBNs\_boron trioxide. Furthermore, it has been observed that the size distribution of hBNs\_boric acid is narrower as compared to the hBNs\_colemanite and hBNs\_boron trioxide indicating higher uniformity of hBNs\_boric acid. Besides, the low differences in size distribution of each sample at 0 and 60 min show their high colloidal stability confirming the Z-potential results.

### Thermal Stability of hBNs

TGA analysis was performed to investigate thermal stability of the hBNs within the temperatures range of 30–700◦C. **Figure 4** shows TGA analysis of the synthesized hBNs. As seen, hBNs synthesized from all precursors have high thermal stability since no significant weight loss was observed with heating up to 700◦C. A slight different result is observed for the hBNs\_colemanite. This may be due to presence of metal oxides originating from colemanite. Since the hBNs\_colemanite may include incomplete BNNTs in the sample, the high thermal stability of BNNTs also contributes to the thermal stability of the mixture (Ferreira et al., 2015; Kalay et al., 2015, 2016).

### Biodegradation of hBNs

Biodegradability of nanomaterials is an important issue to be investigated before their use in medicine (Naahidi et al., 2013). In addition to the therapeutic effect, biodegradable nanomaterials have an additional advantage since they can be eliminated from the body through their degradation (Kalashnikova et al., 2015).

Furthermore, boron compounds such as boric acid can be used in some cancer types as therapeutic agents as mentioned earlier (Barranco and Eckhert, 2004; Scorei et al., 2008). However, frequent administration is needed because of short half-life circulation and low bioavailability of boron compounds. Thus, slow release of boron derivatives is important for therapeutic applications particularly in a cancer treatment (Li et al., 2017). Boron exists as boric acid at the physiological pH and it is considered having the therapeutic effect (Barranco and Eckhert, 2006).

The biodegradation of the hBNs under oxidative and hydrolytic conditions was examined for 30 days. After 30 days, hBNs in LMS or PBS suspensions were centrifuged to remove the supernatants. Then, TGA analysis was performed on the remaining pellets. **Figure 5** shows the result of the experiments. The control hBNs\_boric acid was found thermally stable during

the temperature gradient. The approximate weight loss of hBNs\_boric acid incubated in LMS was 12% while it was only 3% for hBNs\_boric acid incubated in PBS. The mass loss of the hBNs\_colemanite incubated in LMS was around 1% while it was 2% incubated in PBS indicating that hBNs\_colemanite were chemically stable in LMS and PBS suspensions after a 30 day incubation as seen in **Figure 5B**. Furthermore, **Figure 5C** shows that approximately 12% weight of hBNs\_boron trioxide incubated in LMS was lost while the mass loss of the hBNs incubated in PBS was found to be around 6%. The data show that hBNs\_colemanite are more resistant to both oxidative and hydrolytic degradation processes compared to the hBNs synthesized from other two precursors. This might be again due to the presence of incomplete BNNTs along with the hBNs. In addition, the hBNs\_boric acid and hBNs\_boron trioxide were more resistant in hydrolytic degradation compared to the oxidative degradation. It is hypothesized that adding hydrogen peroxide into the test suspensions could damage the B-N bonds and these defects leads to the degradation of the hBNs. Furthermore, as seen in **Figure 2C** in FT-IR results, a weak band at around 3,400 cm−<sup>1</sup> , which is attributed to O–H stretching,

could make the hBNs\_boric acid and hBNs\_boron trioxide more prone to oxidative degradation (Tang et al., 2008). The mass loss of the hBNs demonstrates that the degradation of the hBNs depends on their precursors and dependently crystallinity in both oxidative and hydrolytic degradation environments. Please note that after the centrifugation, it is possible that all hBNs are not precipitated as a pellet. However, the same experimental conditions were applied for all types of hBNs, we think that the trend in the data should be the same.

The amount of the released boron from hBNs in the LMS and PBS suspensions along with their degradation is investigated using ICP-MS. It was found that as the incubation time increased, the boron content in their suspensions increased. After 30 days, the amount of released boron from hBNs\_boric acid increased to 142.0 ± 6.4 ppm from 88.1 ± 3.0 ppm in LMS and PBS suspensions of hBNs\_boric acid (**Figure 6A**). The boron concentration was found to be lower in both LMS and PBS suspensions of hBNs\_colemanite as seen in **Figure 6B**, because of higher crystallinity of hBNs\_colemanite. As mentioned earlier, crystallinity of the hBNs affects the boron release from the hBNs (Li et al., 2017). The amount of released boron from hBNs\_colemanite increased to approximately 20.7 ± 0.4 ppm from 6.2 ± 0.9 ppm in PBS solution while the amount of released boron increased to approximately 31.4 ± 0.9 ppm from 22.0 ± 0.1 ppm in LMS. Thus, a decreased amount of boron release (approximately 7-fold) was found for hBNs\_colemanite in PBS due to its crystallinity. After 30 days, the boron content in both suspensions of the hBNs\_boron trioxide was found similar to the case of hBNs\_boric acid as shown in **Figure 6C**. The amount of released boron from hBNs\_boron trioxide increased to approximately 142.1 ± 5.4 ppm from 112.5 ± 1.8 ppm in PBS solution while the amount of released boron increased to approximately 144.5 ± 7.3 ppm from 117.2 ± 10.8 ppm in LMS.

A further demonstration of hBNs degradation process was provided with Raman spectroscopy analysis. **Figure 7A** shows the Raman spectra of hBNs\_boric acid before and after degradation process. The decrease of peak intensity at around 1,360 cm−<sup>1</sup> suggests the degradation of the hBNs after 30-day incubation in LMS. For hBNs\_colemanite, a sharp peak at around 1,360 cm−<sup>1</sup> decreases as seen in **Figure 7B** resulting from the degradation of the hBN structures. The wide and decreased peak at around 1,360 cm−<sup>1</sup> for hBNs\_boron trioxide indicates the degradation of the hBNs after 30 days (**Figure 7C**). A decrease in the peak intensity at around 1,360 cm−<sup>1</sup> demonstrates the biodegradation of the hBNs. These results are also in good agreement with TGA and ICP-MS results.

### REFERENCES


### CONCLUSION

In this study, hBNs were synthesized from three different precursors; boric acid, colemanite, and boron trioxide, and their crystallinity, size, shape, and biodegradation behavior were investigated. The morphological and spectroscopic characterization results indicate that the hBNs\_boric acid show high crystallinity, uniform, and unique platelet-like structures. In the sample prepared from hBNs\_colemanite, incomplete BNNTs were observed along with the highest crystalline hBNs. The formation of BNNTs is attributed to the presence of metal oxides behaving as catalysts. The hBNs\_boron trioxide have also high crystallinity despite irregular lateral size dimensions. It is found that the size of the hBNs varies and their colloidal suspension is quite stable. The biodegradation studies show that hBNs\_colemanite are more resistant to both oxidative and hydrolytic degradation (approximately 7-fold) due to their high crystallinity while hBNs\_boric acid and hBNs\_boron trioxide are more prone to both oxidative and hydrolytic degradation. The degradation study indicate a slow boron release possible in the form of borate, which can be an important finding in the treatment of certain cancer types (Barranco and Eckhert, 2006). The hBNs are suggested as possible therapeutic agents for use in some cancer types such as prostate due to boron content. In addition, using hBNs as a boron source might be a significant improvement for wound healing applications due their slow degradation. Over all, the type of precursors affects the synthesized product crystallinity, structural uniformity, and their biodegradation behavior. Thus, an appropriate boron precursor should be chosen based on the target application.

### AUTHOR CONTRIBUTIONS

ÖS and ME carried out the experimental work and the analysis of the data. All authors contributed to manuscript revision, read and approved the submitted version.

### FUNDING

The authors acknowledge the financial support from Yeditepe University.

### ACKNOWLEDGMENTS

The authors also acknowledge the help of Seda Kele¸stemur with ICP-MS experiment.


heterostructures with MoS2 and WS2. J. Am. Chem. Soc. 137, 13060–13065. doi: 10.1021/jacs.5b07739


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 ¸Sen, Emanet and Çulha. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The ABC Guide to Fluorescent Toolsets for the Development of Future Biomaterials

#### Ferdinandus <sup>1</sup> and Satoshi Arai 2,3 \*

*<sup>1</sup> Waseda Bioscience Research Institute in Singapore, Singapore, Singapore, <sup>2</sup> Research Institute for Science and Engineering, Waseda University, Tokyo, Japan, <sup>3</sup> PRIME-AMED, Tokyo, Japan*

In recent decades, diversified approaches using nanoparticles or nano-structured scaffolds have been applied to drug delivery and tissue engineering. Thanks to recent interdisciplinary studies, the materials developed have been intensively evaluated at animal level. Despite these efforts, less attention has been paid to what is really going on at the subcellular level during the interaction between a nanomaterial and a cell. As the proposed concept becomes more complex, the need for investigation of the dynamics of these materials at the cellular level becomes more prominent. For a deeper understanding of cellular events, fluorescent imaging techniques have been a powerful means whereby spatiotemporal information related to cellular events can be visualized as detectable fluorescent signals. To date, several excellent review papers have summarized the use of fluorescent imaging toolsets in cellular biology. However, applying these toolsets becomes a laborious process for those who are not familiar with imaging studies to engage with owing to the skills gap between them and cell biologists. This review aims to highlight the valuable essentials of fluorescent imaging as a tool for the development of effective biomaterials by introducing some cases including photothermal and photodynamic therapies. This distilled information will be a convenient short-cut for those who are keen to fabricate next generation biomaterials.

Keywords: fluorescent probes, bioimaging, chemical indicators, fluorescent proteins, single-cell studies

## HOW DO NANOMATERIALS AFFECT CELLULAR ACTIVITIES?

Recent advances in the development of biomaterials have yielded various types of nanomaterials for medical imaging, drug delivery, and regenerative medicine. The words nanomaterials and nanomedicine increasingly appear in these studies (Wagner et al., 2006). A nanometer-sized object is compatible with the scale of a single cell. For instance, drug-containing nanoparticles or nanorods are able to be taken up into a micron-sized cell efficiently, reach subcellular compartments, and lead to the alteration of cellular activities (Peer et al., 2007). To further benefit therapeutic efficiency, the surface of the nanomaterials can be modified to target the cells of interests (Chauhan and Jain, 2013; Yhee et al., 2014). Apart from nano-sized objects, two or three dimensional bulk-sized materials have also emerged, where their thickness is in the order of nanometers and/or their surface possesses a sophisticated nanostructure (Fujie, 2016). These nanomaterials effectively interact with biomacromolecules located at the cellular surface at a nanoscale interface (Luo et al., 2013). As potential applications, the modification of medical devices decorated with functional surfaces can prevent undesirable side effects, such as immune responses, when they are

### Edited by:

*Giada Graziana Genchi, Fondazione Istituto Italiano di Technologia, Italy*

#### Reviewed by:

*Stefan G. Stanciu, Politehnica University of Bucharest, Romania Stefano Luin, Scuola Normale Superiore di Pisa, Italy Robert E. Campbell, University of Alberta, Canada*

#### \*Correspondence:

*Satoshi Arai satoshiarai@aoni.waseda.jp*

### Specialty section:

*This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology*

Received: *15 October 2018* Accepted: *07 January 2019* Published: *23 January 2019*

#### Citation:

*Ferdinandus and Arai S (2019) The ABC Guide to Fluorescent Toolsets for the Development of Future Biomaterials. Front. Bioeng. Biotechnol. 7:5. doi: 10.3389/fbioe.2019.00005* introduced into the body (Franz et al., 2011). Also, customized culture dishes with nanostructured surfaces enable the induction of cellular differentiation and the morphological control of multicellular components (Discher et al., 2005; Marino et al., 2015b).

When developing ideas for biomedical applications, one starts with the design of a nanomaterial that can alter cellular functions at the subcellular level. After a proof of concept is demonstrated in simple cellular studies, one may go further and carry out studies in animal models. In recent years, material scientists carried out animal studies themselves due to the growth in interdisciplinary studies between medicine and material science. However, despite these efforts, the evaluation of whether the designed material is really able to function as expected has been overlooked. In particular, little is known about the spatiotemporal information in real time regarding the behavior of the nanomaterials in cells.

Thanks to advances in molecular and chemical biology, various toolsets are easily accessible to light up cellular events using fluorescence imaging. As the first step, it is necessary to choose an adequate microscope (epi-microscope, confocal microscope, etc.) and then make a decision on the appropriate fluorescent indicators (probes or sensors). Although the development of indicators has progressed tremendously in chemistry and biology, the strategy adopted in both fields is likely to be common, which means that these indicators are capable of reporting changes in intracellular events as fluorescent signals. In general, the indicators are classified into small chemical or genetically encoded indicators. This review will skip over detailed descriptions as these have already been reviewed in depth (Giepmans et al., 2010; Kang et al., 2011; Newman et al., 2011; Specht et al., 2017). In brief, genetically encoded indicators can specifically target organelles or events and exhibit long-term (for example, days) stability under microscopic observation (Greenwald et al., 2018). On the other hand, chemical indicators are easy to handle and thus accessible for anyone without the need for expertise in cell biology (**Figure 1**). Owing to recent developments, there is a risk to oversimplify the difference between the two types of indicators. For instance, the modified Baculovirus allows a safe and easy-handling protocol for genetically encoded indicators for researchers without any molecular biology skills. Such technology could contribute to solve the gap between the two indicators in future.

This review paper stresses the significance of the interaction between nanomaterials and biospecimens at the submicron scale. It therefore focuses on the tools for observing the dynamics of cellular events in real-time rather than simple immune staining. We will also skip any references to commercially available kits or indicators.

### TOOLSETS TO VISUALIZE INTRACELLULAR EVENTS IN REAL-TIME

### (i) How to See Changes in Cell State

Nano-sized carriers invade cells and then release drugs; so what happens to the cells and how do we visualize it? In the case of anticancer drug-loaded carriers, a cell viability assay (**live/dead assay kit**) could be the primary option. Of further interest would be the pathway through which cell death is executed, either apoptosis or necrosis. Necrosis involves the collapse of the outer cellular membrane, which can be evaluated with **propidium iodide (PI)**. During apoptosis, phosphatidylserine appears on the outer leaflet of the lipid bilayer, which can be identified with **fluorescent labeled Annexin V**. Looking at apoptosis more closely, indicators for mitochondrial membrane potentials (**JC1**) and Caspase activities (**Caspase-3/7 Green**) would be helpful.

To date, most drug delivery systems are designed to cause cell death in diseased cells and in such cases a simple live/dead test is frequently done. If the drug does not lead to complete cell death, it is possible to overlook the chance that nanomaterials could be affecting other cellular functions such as the cell cycle and proliferation. In fact, it was reported that even old-fashioned hyperthermia therapy perturbed the cell cycle (Dewey, 2009). In such cases, a genetically encoded indicator, Fluorescent Ubiquitination-based Cell Cycle Indicator (**Fucci**), could be used to distinguish S, G2, and M phases (Sakaue-Sawano et al., 2008).

In addition to the cell cycle, the change in energy metabolism resulting from nanomaterials is worth considering, though rarely discussed. A proliferating cell, such as a cancer cell, adopts a unique way to produce ATP as an energy source via strong glycolytic pathways, which is known as the Warburg effect (Heiden et al., 2009). Cancer cells therefore depend primarily on external glucose to survive while differentiated, non-cancerous cells rely on oxidative phospholyration (Oxphos) in mitochondria. ATP measurements are commonly performed using the well-known luciferin-luciferase assay (Manfredi et al., 2002). However, this method loses the spatiotemporal information of intracellular ATP dynamics. In order to visualize ATP fluctuations in real-time inside the cell, several fluorescent indicators have been developed, starting with genetically encoded indicators (**ATeam** and **Perceval**/**PercevalHR**), followed by the advent of small molecule indicators (Kurishita et al., 2012; Tantama et al., 2013; Tsuyama et al., 2013; Yaginuma et al., 2014). More recently, a team including an author of this review published a paper on the success of red, green, blue (RGB) color fluorescent indicators (**MaLions**) (Arai et al., 2018). This handy colorful toolset allows the observation of ATP dynamics at different organelles in the same cell simultaneously.

Other tools for visualizing energy metabolism include genetically encoded indicators sensing glucose (Ye and Schultz, 2003) and lactate (San Martín et al., 2013). These provide direct evidence for changes in metabolic pathways.

### (ii) How to image Signaling molecules

In recent years, pioneering groups have attempted to generate stimulation devices that allow remote control of physiological functions in brains and muscles (Stanley et al., 2012; Marino et al., 2015a). The targeted functions are frequently associated with intracellular Ca2<sup>+</sup> flux by external stimulation, which can be visualized using commercially available chemical indicators such as Fluo 4 and Fluo 8. Surprisingly, even a couple of decades after the advent of its prototype, Fluo 4 is still commonly used as a gold

standard (Russell, 2011). Although there is no doubt regarding the viability of such indicators, their major drawback is their short cellular retention time. Due to the multiple carboxyl groups of the indicators in the anionic charged state, these chemical indicators are likely to be pumped out from cells via the anion transporter (Takei et al., 2013). On the other hand, genetically encoded Ca <sup>2</sup><sup>+</sup> indicators such as **GCaMP** enable long-term observations (Nakai et al., 2001). Moreover, the expanded color set of genetically encoded indicators prevails over chemical ones as color palettes, such as the R-, G-, B-**Geco** series, have already been established (Zhao et al., 2011). For more quantitative analysis, one should consider using the fluorescence resonance energy transfer (FRET)-based **Cameleon** series (Miyawaki et al., 1997). Considering its potential applications, the genetically encoded glutamate indicator, **iGluSnFr**, could be useful to visualize synaptic communications (Marvin et al., 2013). Also, as the other second messenger in addition to Ca2+, indicators to detect cAMP, such as the **Flamindo** series, are effective (Odaka et al., 2014; Harada et al., 2017).

Reactive oxygen species (ROS) play a critical role in several cascades. A commercially available indicator, 2′ ,7′ dichlorodihydrofluorescein diacetate (**H2DCFDA**), is capable of capturing any kind of ROS non-specifically. For advanced studies, more specific indicators are available for distinguishing between several ROS species (hydroxyl radical; **OxiOrange**TM, HClO; **HySOx**, H2O2; **HYDROP**, ONOO radical; **NiSPY-3**, Molecule Oxygen; **LOX-1**). Although limited to H2O2, genetically encoded indicators (**Hyper** series) are also applicable for monitoring ROS dynamics at target organelles (Bilan et al., 2013).

### (iii) How to see invisible physicochemical elements

Many biomedical applications, such as therapeutic devices and nanomaterials, are associated with external physical stresses, such as force, magnetic fields, and temperature. For example, how does a single cell experience the temperature change produced by heat therapeutic devices? **ER thermo yellow** (or **ERthermAC**) and **Mito thermo yellow** are available to monitor temperature changes at the ER (endoplasmic reticulum) and mitochondria, respectively; these were developed by teams including an author of this review (Arai et al., 2014, 2015). As genetically encoded indicators, **gTEMP** and **tsGFPs**, which can target organelles, were developed (Kiyonaka et al., 2013; Nakano et al., 2017) and allow thermometry at the target organelles.

In addition to temperature as an obvious physical parameter for biomedical applications, there are several other elements to be examined in the biophysics field. For instance, several indicators were reported to visualize invisible factors, such as viscosity (Battisti et al., 2013; Liu et al., 2014), molecular crowding (Boersma et al., 2015), polarity (Sunahara et al., 2007; Abbandonato et al., 2018), and tension (Grashoff et al., 2010). Although the importance of these factors for biomaterials development still remains vague, these options should be kept in mind.

One may also have the interests in measuring the change of intracellular pH as a common physicochemical parameter. To monitor the pH, small chemical indicators (**SNARF** series) and genetic ones (**pHluorin** series) are available, some of which can be applied for two photon, ratiometric, and fluorescence life time imaging (Miesenböck et al., 1998; Bizzarri et al., 2009; Shen et al., 2014).

### (iv) How to avoid the misinterpretation due to Artifacts

It is fascinating that a wide range of indicators are available for detecting various elements, however, one should take great care regarding pH fluctuations. Most indicators, regardless of whether they are genetically encoded or are chemical, follow the mechanism that the fluorescence intensity is altered as the change in the concentration of targeted analyte. Yet, the fluorescence of many indicators is also sensitive to pH changes owing to their pH-sensitive components, such as phenol and carboxylic groups. Though the cytoplasmic pH is maintained around 7.4 (different organelles being slightly different), the metabolic stress may cause an acidic change (Matsuyama et al., 2000). When the experimental conditions result in exposure to severe pH changes, this can be accounted for using pH indicators. This is vitally important to clarify whether the fluorescence change originates from the pH fluctuation or from the change in the concentration of targeted analyte (Berg et al., 2009). For example, one may correct the pH fluctuation using an additional pH indicator as reported (Berg et al., 2009). In another unique approach, Sato et al. published a success of simultaneous imaging of chloride ion and pH, where the pH correction was done using the indicator capable of detecting pH and chloride ion simultaneously (Sulis Sato et al., 2017). In addition to pH fluctuation, the fluorescence signal can also be affected by focus drift, different concentration of indicators, photobleaching, photoactivation, and several other variables. To avoid misinterpretation of the results, the detection limits of the indicators should be accounted for and the predicted artifacts considered.

### "MUST HAVE", NOT "NICE TO HAVE"!

Although it is quite obvious that imaging technology is a powerful tool for the development of biomaterials, most researchers consider it as supplementary, that is, "nice to have," but not "must have." We would like to stress the importance of imaging studies as the concept of the materials gets more advanced. In this section, we present a few examples of the effective use of indicators.

### (i) What happens inside a cell during "photothermal therapy"?

Thermal therapy to kill tumors by means of elevated temperature is a popular therapeutic approach as medical treatment for cancers, in which one of key things is the mechanism to heat up cells and tissues. Among several thermal therapeutic ways, photothermal therapy (PTT) has been the most appealing means which enables local treatment with minimal invasiveness. This method requires photothermal materials which absorb near infrared (NIR) light and convert the energy into heat. To date, innumerable materials such as inorganic materials, organic dyes, and semicondutive polymers, have been fabricated in order to achieve more efficient photothermal therapeutic effect (Zhang et al., 2013, 2015). However, these materials also possess the risks of generating ROS, aside from producing heat. In some cases, ROS is harmful for normal tissues and cells, thus making the development of a pure photothermal material challenging. Recently, Jung et al. successfully generated novel photothermal materials using the organic dyes, cryptocyanines, which are ineffective ROS generators and also specifically target the mitochondria (Jung et al., 2017). It is easy to imagine that a strategy to target a critical organelle would benefit therapeutic efficiency. The team clarified the mechanism using a fluorescent ROS indicator targeting mitochondria (**MitoSOX**); the heat generated from the dye by photoirradiation perturbed the electron transport chain, resulting in a change in endogenous ROS dynamics at the mitochondria, leading to cell death. In other words, the change in ROS dynamics is not triggered by ROS derived from the materials.

As the other point in PTT, we need to pay attentions to how a single cell senses a rise in intracellular temperature caused by external heating. For example, when a temperature change was brought about in the multicellular spheroidal HeLa cells by external heating, a mitochondrial targeting temperature indicator, **Mito thermo yellow**, showed that each cell in the spheroidal aggregates experienced a different temperature change depending on its position (Arai et al., 2015) (**Figure 2A**). The maximum temperature difference between cells turned out to be around a couple of degrees even within the same bunch of cells. This result implies that maximal therapeutic efficiency cannot be achieved if heating is not homogeneous at the single cell level. This thermometry was also applied to evaluate the stimulation of skeletal muscles by heat (Marino et al., 2017). Attilio et al. demonstrated that once a gold nanoparticle, as a heat generator, was taken into the skeletal muscles, its contraction could be induced by heat with the irradiation of a NIR laser. A small fluorescent temperature indicator, ER thermo yellow, assisted in monitoring the intracellular temperature change while the muscle was contracting. For a more advanced example, Zhu et al. reported an intelligent nanocomposite that comprises a photothermal material and a fluorescent thermometer together (Zhu et al., 2016) which allows more accurate thermometry at the heat spot owing to the zero distance between the heater and the thermometer (**Figure 2B**).

### (ii) What happens inside a cell during "photodynamic therapy"?

Much like photothermal therapy (PTT), photodynamic therapy (PDT) has a long history. However, there are scarce reports where cellular events were observed during the photodynamic therapy. For example, Kuimova et al. reported a unique porphyrin-dimer that possesses dual functionalities as an intracellular viscosity indicator as well as a photosensitizer to generate singlet oxygen species acting as cytotoxic agents (Kuimova et al., 2009) (**Figure 2C**). After photoirradiation, the generated singlet oxygen led to cell death while also causing a change in viscosity, which was visualized using their ratiometric viscosity indicator. Interestingly, it turned out that the increase in intracellular viscosity was not homogeneous and decreased the diffusivity of intracellular species. They suggested that the increase in the viscosity resulting from the photosensitizer altered the formation of singlet oxygen species and also their decay. The viscosity at subcellular compartmentalized spaces could be an important factor as it may also alter the transport efficiency of drugs.

Very recently, Wang et al. proposed a new type of therapeutic mechanism where the free radical generator is coupled with the NIR light (**Figure 2D**) (Wang et al., 2018). Unlike common previous approaches, their designed material does not generate ROS from intracellular oxygens but instead produces free radical species from the thermal initiator (V057: 2,2′ -Azobis [N-(2-carboxyethyl)-2-methylpro-pionamidine] hydrate). More specifically, the gold nanocage (AuNc) which acts as an efficient photothermal material was decorated with V057 and a fluorescent dye (Cy5.5). Once the temperature is elevated on the surface of AuNc by NIR irradiation, the free radical is produced due to the cleavage of the thermal initiator, V057. Simultaneously, the free radical generation could be monitored with the fluorescence intensity change of Cy5.5 since the fluorescence is recovered from quenched state involved in the free radical production. Importantly, the material is also designed

to target to mitochondria in order to maximize the therapeutic efficiency. Thus, the total system allows the induction of the cell death with concurrent monitoring of the free radical generation in the targeted mitochondria. It was also noted that this system works in both normoxic and hypoxic conditions because of the oxygen-independent principle.

#### (iii) Tackles for "unknown mechanism" and future perspective

Though not quite often, we sometime face complicated situations where a material works well in animals but its mechanism remains vague. In such cases, imaging technology can provide vital clues by visualizing the dynamics at a cellular level. Mitsunaga et al. pioneers in photoimmunotherapy, developed a near-infrared dye-labeled antibody that binds specifically to cancer cells and then effectively induces cell death via NIR illumination (Mitsunaga et al., 2011). The system is also able to work efficiently in animal level, although the mechanism still remains unclear at cellular level. Crucially, it is unlikely to result in ROS or heat production. Recently, Nakajima et al. investigated the mechanism regarding this photoimmunotherapy and then verified that the plasma membrane was damaged for a short time by observing the sodium ion influx using a fluorescent indicator (Nakajima et al., 2018). This minute damage on the membrane resulted in the increase (**Figure 2E**) of its permeability, cell swelling, and cell death. In this case, though it is not still perfectly understood, the imaging technique provided a hint to elucidate the unknown mechanism.

The importance of the physical stimulus will evolve following the development of biomaterials. This is because attempts to manipulate cellular functions using physical stresses is one of the biggest topics in biophysics, known as mechanobiology, providing fresh insights for material science. Pioneers in this field are dedicated to revealing how cells sense mechanical stress and what the molecular players are (Uhler and Shivashankar, 2017). Concurrently, leading biologists have also produced various tools to help unveil the sensing mechanism at a molecular level; these tools will be of benefit to material scientists as well in the development of novel biomaterials. Ideally, the development of indicators, that are applicable for two-photon and NIR imaging, should be also significant to achieve the subcellular resolution for in vivo imaging using animals (Hong et al., 2017). We positively believe the promising future of imaging studies which will claim their place as a "must have" approach in biomaterials development.

### REFERENCES


### AUTHOR CONTRIBUTIONS

F and SA conceived and revised the manuscript.

### FUNDING

This work was partially supported by Japan Agency for Medical Research and Development (AMED) PRIME (JP18gm5810001), a Grant-in-Aid for Scientific Research on Innovative Areas Chemistry for Multimolecular Crowding Biosystems (18H04566) and a Grand-in-Aid for Young Scientists (A) (16H06044).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ferdinandus and Arai. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Bright Dots and Smart Optical Microscopy to Probe Intracellular Events in Single Cells

Hideaki Fujita<sup>1</sup> , Chongxia Zhong<sup>2</sup> , Satoshi Arai 3,4 and Madoka Suzuki 2,5 \*

*<sup>1</sup> WASEDA Bioscience Research Institute in Singapore, Singapore, Singapore, <sup>2</sup> Institute for Protein Research, Osaka University, Osaka, Japan, <sup>3</sup> Research Institute for Science and Engineering, Waseda University, Tokyo, Japan, <sup>4</sup> PRIME-AMED, Tokyo, Japan, <sup>5</sup> PRESTO, Japan Science and Technology Agency, Saitama, Japan*

Probing intracellular events is a key step in developing new biomedical methodologies. Optical microscopy has been one of the best options to observe biological samples at single cell and sub-cellular resolutions. Morphological changes are readily detectable in brightfield images. When stained with fluorescent molecules, distributions of intracellular organelles, and biological molecules are made visible using fluorescence microscopes. In addition to these morphological views of cells, optical microscopy can reveal the chemical and physical status of defined intracellular spaces. This review begins with a brief overview of genetically encoded fluorescent probes and small fluorescent chemical dyes. Although these are the most common approaches, probing is also made possible by using tiny materials that are incorporated into cells. When these tiny materials emit enough photons, it is possible to draw conclusions about the environment in which the tiny material resides. Recent advances in these tiny but sufficiently bright fluorescent materials are nextly reviewed to show their applications in tracking target molecules and in temperature imaging of intracellular spots. The last section of this review addresses purely optical methods for reading intracellular status without staining with probes. These non-labeling methods are especially essential when biospecimens are thereafter required for *in vivo* uses, such as in regenerative medicine.

### Edited by:

*Gianni Ciofani, Politecnico di Torino, Italy*

### Reviewed by:

*Tzu-Ming Liu, University of Macau, China Satya Ranjan Sarker, Jahangirnagar University, Bangladesh Takayuki Uchihashi, Nagoya University, Japan*

\*Correspondence: *Madoka Suzuki suzu\_mado@protein.osaka-u.ac.jp*

#### Specialty section:

*This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology*

Received: *11 October 2018* Accepted: *12 December 2018* Published: *04 January 2019*

#### Citation:

*Fujita H, Zhong C, Arai S and Suzuki M (2019) Bright Dots and Smart Optical Microscopy to Probe Intracellular Events in Single Cells. Front. Bioeng. Biotechnol. 6:204. doi: 10.3389/fbioe.2018.00204* Keywords: fluorescent protein, nanoparticle, infrared, nanodiamond, quantum dot, Raman, temperature

## INTRODUCTION

Since Robert Hooke observed a cork and named a compartment he found within it a "cell" using a simple microscope composed of two lenses in the 17th century, microscopic observation has become one of the most essential techniques in cell biology. Most animal cells are almost transparent and various techniques were developed to image these transparent cells. The simplest method is to label the cells with colored probes. Haematoxylin-Eosin staining was introduced back in the 19th century and it is still commonly used in histological labs for diagnostics. Now, there are many probes with various functions and characteristics (**Figure 1**), which are briefly summarized in section Genetically encoded probes vs. chemical dyes to highlight the variety. These fluorescent probes were not designed to work on single molecules. The signal is captured as an average value of many molecules. When many molecules are concentrated in a small space, a bright fluorescent nanoprobe can be obtained. When optical microscopy is optimized for nanoprobes, location tracking as well as environment probing are made possible using a single particle.

Section Bright nano-dots as single probes for tracking of this review focuses on nanoprobes that are designed to be imaged one by one with a purpose to monitor intracellular parameters such as the location of biomolecules and spatial and temporal changes in temperature. Other methods do not involve staining. Differential interference contrast microscopy and phase contrast microscopy both utilize the phase shift between the illuminating light and the light which passes through the specimen and in so doing can visualize transparent cells without labeling. These are not discussed in this review as there is much less information on the cellular states that can be determined using these techniques when compared to the other methods we will discuss. Imaging utilizing the vibrational information of a molecule is discussed in section Imaging technologies without labeling of this review (**Figure 2**).

### GENETICALLY ENCODED PROBES VS. CHEMICAL DYES

Fluorescent sensors are defined as those in which the signal, such as fluorescence intensity, wavelength, or lifetime, is altered in response to a change in the environment. They are called fluorescent probes, indicators, and sensors (hereafter referred to as "fluorescent probes") (Zhang et al., 2002). The probes can be broadly categorized into two groups: genetically encoded probes and small chemical dyes. Which probe to choose depends on what intracellular events need monitoring and the duration of the observation.

### Genetically Encoded Fluorescent Probes

From the standpoint of long-term observation in fluorescence microscopy, genetically encoded probes are preferred. If we are keen to make observations for a few hours or days using small molecule indicators, we need to employ the strategy where the dyes are anchored to the target places or proteins covalently, which is still challenging (Wakayama et al., 2017). Also, to specifically observe an evenet of interest, genetically encoded probes once again prevail over small chemical dyes due to the working mechanism that they employ. The ability of natural proteins to bind to a specific target is adaptable in genetically encoded probes, while small chemical indicators use synthetic modules for sensing which are usually less specific to the target molecule than natural counterparts. Thus, genetically encoded probes exhibit better specificity to a target, especially against small molecules such as Ca2+, cAMP, ATP, and glucose (DiPilato et al., 2004; Imamura et al., 2009; Zhao et al., 2011; Hofig et al., 2018).

Regarding the delivery of probes into cells, genetically encoded probes require transfection using reagents, or physical delivery. To evaluate the probes for more than several days, preparation of a stable cell line into which the genetic code of the probe is integrated into the genome is necessary. Also, if transfection is difficult, such as in the case of primary cultured cells, viral infection is unavoidable. Another feature worth highlighting is that the location of the probes can be controlled at an organelle level. For example, a target sequence, such as for the nucleus (nuclear localization signal), mitochondria (subunit VIII of cytochrome c oxidase), or cellular membrane (growthassociated protein 43, neuromodulin), is conjugated to the N or C terminus of the probe, which then selectively localizes the probe to the target organelle (De Giorgi et al., 1999; Arai et al., 2018). Lastly, once the plasmid for a probe is purchased or given as a gift from the developer, it is always possible to amplify the plasmid at little cost.

## Organic Dyes

One may feel that genetically encoded probes are the perfect choice. However, there is one critical issue–ease of handling. Genetically encoded probes require around 1–3 days for expression in cells after delivery of the gene. On the other hand, small chemical indicators are easy to apply for staining, usually taking around 30 min to take effect before imaging can begin (Zhu et al., 2016). For those who rarely perform molecular biology protocols, transfection might be difficult to apply to their biospecimens or it may be that the use of viruses for infection is not allowed in their laboratory. In some cases, biological samples cannot withstand the 1 day incubation time needed for gene expression. In these instances small chemical dyes are the only remaining option. Across the field of research, imaging is often required for a variety of cell and tissue types, as well as different animal species, beyond that of commonly used cell lines (Vendrell et al., 2012). Gene delivery frequently proves to be a challenge under these conditions. Small chemical dyes can therefore prove to be helpful and even be considered the first choice for imaging. One can find in the literature positive examples of labeling using small near-infrared dyes in biomedical research at tissue and animal levels (Hong et al., 2017). The purpose of the experiment and the type of specimen should always be the primary factors in determining which probes to use, regardless of whether they are small chemical dyes, or genetically encoded probes.

These fluorescent probes were designed to take an average of many molecules to measure a given parameter. Thus, spatial resolution is limited by optical microscopy. Are there means to

improve accuracy in location while probing parameters using optical microscopy?

## BRIGHT NANO-DOTS AS SINGLE PROBES FOR TRACKING

The accumulation of fluorescent molecules creates a fluorescent nanoparticle. Quantum dots (Qdots) and nanodiamonds are also nanoparticles that emit photons upon excitation by light. These nanoparticles can be so bright that each dot is tracked in single camera frames repeatedly at 30 Hz or faster. When the nanoparticle is designed as a sensor, it is possible to probe the position of the nanoparticle as well as garner information about its local environment. If said probing is repeated frame by frame, time courses of the measured parameters can be obtained instantly. Among the large variety of applications of nanoparticles for probing, temperature sensing is one of the most advanced for biological usage. As the temperature is a fundamental parameter that can influence chemical and physical processes in any biological systems, the mechanism how living organism senses the temperature and how it releases heat in its body has been one of the major research topics in biology, and in animal, and plant physiologies. Recent advances in optical probes has brought our interest from the macroscopic (such as the temperature of the animal body or of the cell culture medium) to microscopic scale as small as a single cell. However, local event of heat release is still unclarified in live cells. Excellent reviews on fluorescent thermometers in general can be found in the literature (Brites et al., 2012; Wang et al., 2013). Lists detailing the advantages and disadvantages of materials and methods, operating principles, accuracy, and resolution of sensing have been described in reviews (Brites et al., 2012; Wang et al., 2013; Suzuki et al., 2016; Arai and Suzuki, 2018), as well as debates on single-cell thermometry (Baffou et al., 2014, 2015; Kiyonaka et al., 2015; Suzuki et al., 2015). Instead of repeating these details, this section introduces fluorescent nanoparticles that have been applied for probing purposes, including temperature sensing, with single particle analyses in physiological situations. Their functional ranges cover the temperature from ∼25◦C to ∼42◦C. Although the accuracy of determining the temperature depends on the optical setup that reads out signals from the probe, 0.5◦C or better values are frequently reported. The nanoparticles can be further "functionalized" to target biological molecules and organelles.

### Fluorescent Polymeric Nanoparticles for Probing Temperature Changes

We have developed fluorescent polymeric nanoparticles that are able to probe temperature changes in single cells (Oyama et al., 2012; Takei et al., 2014). Each nanoparticle contains temperaturesensitive fluorophores such as Eu-thenolytrifluoroacetonate (EuTTA) or Eu-tris(dinaphthoylmethane)-bis-trioctylphosphine oxide) (EuDT), of which emission intensities inversely correlate with temperature changes. In some cases, less temperaturesensitive fluorophores were also embedded in the same polymeric matrix to form a nanoparticle, whose surface is further covered by a hydrophilic polymer outer layer (Takei et al., 2014; Arai et al., 2015; Ferdinandus et al., 2016). Measurement of the temperature can be achieved by determining the reduction of emission intensity of EuTTA or EuDT, or by the ratio of the intensity of EuTTA or EuDT to that of the less temperature-sensitive fluorophores which act as internal references. The nanoparticle thermosensor assures strong emission intensity since it can pack multiple fluorophores into one particle. Furthermore, it is capable of measuring temperature without being affected by various intracellular factors, such as pH, viscosity, ionic strength, etc.

### Quantum Dots

Qdots are fluorescent semiconductor nanocrystals which possess distinct optical and electrical properties. These properties render Qdots as a promising alternative to organic dyes in biological applications from cell biology to in vitro diagnostics. Qdots have broad excitation spectra and narrow emission spectra, which are composition- and size-dependent. Furthermore, the strong photoluminescence and high photostability of Qdots enable them to play an important role in single particle tracking (Michalet et al., 2005; Ichimura et al., 2014b; Pisanic et al., 2014). Applications in intracellular temperature measurement have also been reported using the shift of emission spectra caused by temperature changes (Yang et al., 2011; Tanimoto et al., 2016).

Functionalization is essential for Qdots to be applied to biological investigations. Being mostly synthesized in organic solvents, Qdots are too hydrophobic to be dissolved in aqueous buffers. Ligand exchange with hydrophilic compounds or encapsulation within amphiphilic coatings are therefore necessary to achieve solubility in water. Surface modification of Qdots provides an interface for conjugating to target molecules (cf. section Targeting).

Blinking and cytotoxicity are two major weak points of Qdots, which limit their biological and therapeutic applications. Blinking, also called intermittent fluorescence, is the phenomenon of a single Qdot particle exhibiting light and dark periods under continuous laser illumination (Ko et al., 2011). Blinking interferes with the tracing process in single-particle tracking studies. Both cadminium-containing and cadminium-free Qdots were found to induce the elevation of intracellular reactive oxygen species levels and increase cell apoptosis, which contributes to a decrease in cell viability (Derfus et al., 2004). Qdots in high concentrations also affect embryo development (Dubertret et al., 2002).

### Fluorescent Nanodiamonds

One of the newly emerging fluorescent probes is nanodiamonds containing nitrogen vacancy centers (NVCs) (Chang et al., 2018). An NVC is composed of a substitutional nitrogen adjacent to a defect in the lattice. The NVC excited by green light emits light in the range of red. Thus, the nanodiamonds containing NVCs are called fluorescent nanodiamonds. When NVCs are embedded within the nanodiamond they show an outstanding robustness to changes in environmental parameters (Sekiguchi et al., 2018). Similar to Qdots, fluorescent nanodiamonds are extremely photostable. In contrast to Qdots, no blinking is observed. Modification of the surface is also relatively easy (Sotoma et al., 2018); e.g., Tsai et al. successfully used this advantage to attach gold nanorods as tiny heat sources onto the surface of nanodiamonds to probe the local temperature at the heated dot (Tsai et al., 2017). Biocompatibility has been shown to be good. Bearing these advantages in mind, fluorescent nanodiamonds can be considered close to the ideal fluorescent probe.

There remain at least two issues regarding the use of fluorescent nanodiamonds as probes. Firstly, the NVC can be exposed at the surface of the nanodiamond, thus eradicating its robustness to the surrounding environment. Greater care should therefore be taken as the diameter of the nanodiamond gets smaller. Secondly, measurement can require long periods of time when extreme accuracy is required or when the amount of NVCs in the nanodiamond is reduced. For example, it is possible to measure the temperature using the negatively charged NVC, NV−. All optical methods detect changes of the emission peak by changing temperature (Plakhotnik et al., 2014). The temperature-dependent status of NV<sup>−</sup> can also be detected by the downward peaks in emission intensity when the microwave is swept, known as optically detected magnetic resonance, or ODMR (Kucsko et al., 2013). Collecting enough photons or sweeping the microwave is required to draw spectra to determine the peaks. Efforts to shorten the sweeping process continue (Tzeng et al., 2015).

### Targeting

It is a challenging task to target nanoparticles to specific cells, organelles, or molecules within a living cell. Yet several attempts have been successful in biological uses. In in vivo conditions, a probe was prepared based on Qdots by immobilizing a ligand with a high affinity for hexahistidine (Arai et al., 2012). By using this Qdot probe, single molecule-tracking of his-tagged myosin V was achieved. Fujita et al. discovered the major sequestration mechanism for glucose transporter 4 in fully differentiated single 3T3L1 adipocytes (Fujita et al., 2010). Gonda et al. conjugated anti-protease-activated receptor1, a tumor cell membrane protein, with Qdots to uncover changes in membrane fluidity and morphology during tumor metastasis in living mice (Gonda et al., 2010). The extreme photostability of Qdots was one of the major reasons that these studies were made possible. More recently, single nanodiamonds surface functionalized with polydopamine followed by thiol terminated poly (ethylene glycol) were successfully targeted to single DNAs using the avidin-biotin system to track their locations (Jung et al., 2018).

There are other types of nanoparticles that are constructed step-by-step directly within living cells at the desired target regions. The first step is to prepare cells expressing single chain avidin (ScAVD) at the target region. The second and the third are to incubate the cells with dibenzocyclooctyne-biotin followed by azide-functionalized organic dyes to perform a copperfree click reaction to construct a ScAVD-biotin-functional dye conjugated nanoprobe (Hou et al., 2016a,b). These nanoprobes have increased brightness and photostability as well as improved localization specificity when compared to single fluorescent dyes.

### IMAGING TECHNOLOGIES WITHOUT LABELING

Although labeling is a powerful technique to visualize the structure of the cell, it is not always preferred. Even when using a bio-safe dye, the detection of cancerous tissue using a labeling method during a surgical operation can significantly prolong the surgery, which increases the chances of complications. Labeling is also not preferred when producing cells for regenerative medicine. In this section, we introduce imaging techniques that don't require labeling. Raman microscopy and fluorescent lifetime imaging microscopy (FLIM) using auto-florescence are discussed (**Figure 2**).

### Raman Imaging Spectroscopy

When light enters a molecule, three types of scatters are generated, Reyleigh scattering, Stokes Raman scattering, and anti-Stokes Raman scattering. Reyleigh scattering has the same energy as incident light, whereas Stokes and anti-Stokes Raman scattering have less and more energy than incident light, respectively. The shift in wavelength (Raman shift) relates to the chemical bonds within the molecule, thus the molecule can be identified using the Raman spectrum. Raman spectroscopy is a non-labeling and non-destructive analysis method, which can obtain information relating to the chemical bonds of a molecule.

Raman imaging spectroscopy usually uses visible light for illumination so the optical resolution is generally higher than Infrared (IR) imaging spectroscopy. Other than standard Spontaneous Raman spectroscopy, there are several other modes of Raman imaging techniques, including Coherent anti-Stokes Raman spectroscopy (CARS), Surface-enhanced Raman spectroscopy (SERS), and Stimulated Raman spectroscopy (SRS). The various types of Raman spectroscopy are reviewed in Krafft and Popp (Krafft and Popp, 2015).

Raman spectra can be used to identify the state of the cell, thus, clinical diagnosis is possible at a tissue level. Raman spectroscopy has been used diagnostically for cancer (Austin et al., 2016; Santos et al., 2017), atherosclerosis (MacRitchie et al., 2018) and liver steatosis (Pacia et al., 2018). Raman imaging of living cells was difficult because the Raman signal is far weaker than fluorescence; strong illumination with a visible laser and a long exposure is therefore needed, which significantly damages the cell. However, recent advances in detectors, spectroscopes, and optical filters have enabled Raman imaging of living cells (Hamada et al., 2008). Now, cell state can be monitored using Raman spectroscopy at a single cell level (Ichimura et al., 2014a, 2015, 2016).

### Fluorescent Lifetime Imaging (FLIM)

FLIM is one of the most important technologies in optical microscopy, especially in live cell imaging, as fluorescence lifetime is independent of fluorophore concentration and focus drift. FLIM has been frequently used in Förester resonance energy transfer (FRET) based on fluorescent sensors that comprise two fluorescent proteins as a donor and an acceptor. The lifetime of the fluorescent protein is altered during energy transfer, thus a FRET sensor can be applied for FLIM. As such, small chemical sensors are also possible for FLIM, which are sensitive to the surrounding environment in the fluorescence lifetime domain. For successful examples, intracellular conditions such as levels of Ca2<sup>+</sup> (Zheng et al., 2018), K<sup>+</sup> (Paredes et al., 2013), Na<sup>+</sup> (Roder and Hille, 2014), and Cl<sup>−</sup> (Chao et al., 1989), as well as pH (Schmitt et al., 2014), and temperature (Itoh et al., 2016) can be monitored using FLIM. FLIM can also be performed in nonlabel imaging as there are naturally occurring endogenous fluorescent biomolecules, such as reduced nicotinamide adenine dinucleotide (phosphate) [NAD(P)H] and flavines present in the cell. Using FLIM, NADH, and NADPH can be distinguished and the relative ratio can be calculated, which is not possible using fluorescence intensity as their spectra are identical (Blacker et al., 2014).

FLIM using auto fluorescence, which is non-labeling, is especially useful in diagnostic purposes. Detection of macrophages in an atherosclerotic model (Marcu et al., 2005), glioma in brain tissue (Yong et al., 2006), retinal pigment epithelium (Miura, 2018), and basal cell carcinoma in skin (Galletly et al., 2008) are made possible using FLIM. Twophoton FLIM is preferred for tissue diagnostics purposes as it reduces photo-damage to the cells and reaches deeper within the tissue. Two-photon excitation is now used to distinguish between papillary and reticular dermis (Shirshin et al., 2017), in the diagnostics of inflamed human skin (Huck et al., 2016), and in melanoma detection (Seidenari et al., 2013). Second Harmonic Generation (SHG) and two-photon FLIM can be combined to increase the information obtained and used in diagnosis as the same laser source can be used (Provenzano et al., 2009; Ranjit et al., 2016).

### Other Non-label Imaging Technologies

There are other modes of non-label imaging technologies such as SHG and infrared (IR) spectroscopy. The information obtained by IR spectroscopy complements that obtained by Raman spectroscopy as when the IR signal from a particular chemical bond is strong, the signal obtained from Raman spectroscopy is weak, and vice versa. A number of products capable of obtaining Raman, FLIM, or IR are released every year, which will further advance this field.

## SUMMARY

Nanotechnology provides tools to control, create, or cure biospecimens. The effect should be evaluated in a quantitative manner to gauge the efficacy of the methods or the quality of the results. Optical microscopy can provide essential information on how and why samples have reacted at sub-cellular resolution. Observation of morphological changes of cells (e.g., axon-like elongation for neuronal cells, multi-nucleus, and contractile structures in muscular cells, or lipid droplets in adipocytes) is one of the first steps to be performed. Probing intracellular reactions provides additional and much more detailed insights. In this article, we reviewed optical methods to probe intracellular events at sub-cellular resolutions in living cells using optical microscopes. Some methods use fluorescent nanoprobes, several of which are designed to change their properties in fluorescence in response to changes in the chemical and physical conditions surrounding the probe. Other methods do not require these tiny probes, but are performed in purely optical manner. Considering the ease of availability, optical microscopy should continue to be a standard method. With the wide variety in probes as well as in equipment, methods are available to meet a range of specific needs. Optimal choices will always be dependent on the sample being studied.

### AUTHOR CONTRIBUTIONS

HF, SA, and MS conceived the story. All the authors co-wrote and critically revised the manuscript.

### REFERENCES


### ACKNOWLEDGMENTS

This work was supported by JSPS KAKENHI Grant Numbers JP15K05251 (to MS), by the Japan Science and Technology Agency Grant Number JPMJPR15F5 (to MS), by the Human Frontier Science Program RGP0047/2018 (to MS), and by Kurita Water and Environment Foundation 17D002 (to MS).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fujita, Zhong, Arai and Suzuki. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Quartz Crystal Microbalance With Dissipation Monitoring: A Powerful Method to Predict the in vivo Behavior of Bioengineered Surfaces

Chiara Tonda-Turo1,2 \*, Irene Carmagnola1,2 and Gianluca Ciardelli 1,2,3

<sup>1</sup> Department of Mechanical and Aerospace Engineering, Politecnico di Torino, Turin, Italy, <sup>2</sup> POLITO BIOMedLAB, Politecnico di Torino, Turin, Italy, <sup>3</sup> Department for Materials and Devices of the National Research Council, Institute for the Chemical and Physical Processes (CNR-IPCF UOS), Pisa, Italy

#### Edited by:

Giada Graziana Genchi, Fondazione Istituto Italiano di Technologia, Italy

#### Reviewed by:

Giuseppe Maruccio, University of Salento, Italy Cristina Satriano, Università degli Studi di Catania, Italy Hamit Yurtseven, Middle East Technical University, Turkey

\*Correspondence:

Chiara Tonda-Turo chiara.tondaturo@polito.it

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 29 August 2018 Accepted: 10 October 2018 Published: 30 October 2018

#### Citation:

Tonda-Turo C, Carmagnola I and Ciardelli G (2018) Quartz Crystal Microbalance With Dissipation Monitoring: A Powerful Method to Predict the in vivo Behavior of Bioengineered Surfaces. Front. Bioeng. Biotechnol. 6:158. doi: 10.3389/fbioe.2018.00158 The Quartz Crystal Microbalance with dissipation monitoring (QCM-D) is a tool to measure mass and viscosity in processes occurring at or near surfaces, or within thin films. QCM-D is able to detect extremely small chemical, mechanical, and electrical changes taking place on the sensor surface and to convert them into electrical signals which can be investigated to study dynamic process. Surface nanotopography and chemical composition are of pivotal importance in biomedical applications since interactions of medical devices with the physiological environment are mediated by surface features. This review is intended to provide readers with an up-to-date summary of QCM-D applications in the study of cell behavior and to discuss the future trends for the use of QCM-D as a high-throughput method to study cell/surface interactions overcoming the current challenges in the design of biomedical devices.

Keywords: quartz crystal microbalance with dissipation monitoring (QCM-D), cell adhesion, cell death, cell cytoskeleton, bioengineered surface characterization

### INTRODUCTION

Quartz crystal microbalance with dissipation monitoring (QCM-D) is an easy, highly versatile as well as highly sensitive method to study processes occurring at the surfaces or within thin films. QCM was introduced in 1960s to monitor layer formation in vacuum and air thanks to the high mass sensitivity of the 5 MHz quartz system (0.057 Hz cm<sup>2</sup> ng−<sup>1</sup> ) which was superior compared with other available technologies (Janshoff et al., 2000). Then, it gained interest in biomedical application since 1980s when novel QCM systems able to work in liquid media was introduced (Marx, 2003). However, the QCM has achieved resounding success since 1995 when a new system able to measure changes in damping properties of adsorbed layers (QCM-D) were developed, patented and commercialized by Q-Sense (Rodahl et al., 1995; Rodahl, 1996).

To date, QCM-D can be applied to monitor processes in real-time by measuring changes in frequency and energy dissipation of the system composed by the surface or thin film of interest adsorbed on a piezoelectric quartz sensor. Quartz sensors are applied in the QCM-D devices as a mechanical oscillation of characteristic frequency can be produced on the crystal by applying an alternated electric field. When a change on the mass occurred at the surface interface or within the thin film, a frequency shift from the fundamental resonant frequency of the crystal (1f) is

measured making the QCM-D a quantitatively and ultrasensitive technology to measure mass changes. In addition to 1f, QCM-D simultaneously monitors the viscoelastic properties of the overlayer or thin film adsorbed to the quartz sensors by recording changes in the energy dissipation factor (D) (**Figure 1A**). When the generator is switched off the crystal oscillation amplitude decays exponentially and the energy dissipation factor can be measured by recording the amplitude of the oscillation as a function of time (Rodahl et al., 1995). Higher are the dissipative properties of the overlayer faster is the decay time resulting in substantial energy dissipation factor shift (1D) from rigid state.

In the case of rigid layer (1D = 0), when the change in mass occurs in air or in vacuum, the frequency shift (1f) is proportional to mass change and it can be quantified applying the Sauerbrey equation (Sauerbrey, 1959). However, biological processes take place in liquid environment where the Sauerbery equation is no more valid and the equation described by Kanazawa and Gordon has to be used to relate 1f to mass changes (Kanazawa and Gordon, 1985). For not rigid layers where the damping of the layer is significant (1D > 0), the Voight model is applied to analyze collected data and equations developed by Voinova et al. (1998) provide information on the mass and viscoelastic properties of the viscous layer. Latest QCM-D devices allow for simultaneously recording of 1f and 1D for multiple harmonics (up to the 13th overtone) which permit to better extract the adsorbed overlayer characteristics (such as density, storage, and loss moduli; Hovgaard et al., 2007). An historical perspective describing the QCM-D working principles and its applications to study complex systems was recently published by Chen et al. (2016).

In the past 15 years, QCM-D has been increasingly used as a tool to characterize the biological interactions of engineered surfaces. To date, QCM-D crystals have been mainly applied to monitor the binding efficacy of probing surface of biosensors (Chen et al., 2010; Xu et al., 2011; Ertekin et al., 2016) as well as to verify the biofunctionalization of nanoengineered surfaces through self-assembled monolayers (SAMs; Minsky et al., 2016) or layer-by-layer (LbL) technologies (Chen et al., 2010). In addition, proteins adsorption competitive studies on biomedical device surfaces have been performed thanks to the possibility to measure at the same time the mass and the viscoelastic properties of absorbed biomolecules, which can be related to protein conformation. Protein adsorption kinetics on polyHEMA (Teichroeb et al., 2008), stainless steel (Chandrasekaran et al., 2013), and titanium oxide (Pegueroles et al., 2012) were analyzed to study interaction between biomedical devices and biological fluids after implantation. Furthermore, QCM-D has been used to assess the antifouling properties of surfaces such those composed of branched polymers (Luan et al., 2017). Besides composition and nanoscale chemistry, different surface morphologies were investigated to relate nanostructures to protein adsorption kinetics (Hovgaard et al., 2008; Chandrasekaran et al., 2013).

Phenomena occurring at the interface within surfaces and biological fluids are fundamental to address specific cell response on biomedical surfaces. QCM-D allows to study these phenomena in real-time without requiring labeling processes which could interfere with the system, thus preventing any loss of information (Nowacki et al., 2014). Furthermore, QCM-D experimental setup is equipped with a temperature-controlled system able to mimic the physiological conditions within the equipment chamber. In the last 5 years a growing interest on QCM-D related to cell behavior studies has been observed as this technique appears to be particularly well-suited for realtime acquisition of cell responses and to monitor cytoskeleton changes after internal or external stimuli. Recently, novel crystal chamber set-ups have been commercialized to permit a multiple characterization of the processes occurring on the crystal surface. Among others, the open window chamber, which combined traditional QCM-D measurements with light microscopy, and the ellipsometric modulus are applied to monitor cell-surface interactions as they allow to follow in real-time changes in cell morphology (Marcus et al., 2012; Tymchenko et al., 2012) and in overlayer thickness (van der Meulen et al., 2014), respectively.

The purpose of this article is to provide a summary of the present state of the art of QCM-D as a tool to monitor cell behavior on surfaces. Then, the future trends in the use of QCM-D as a tool to study phenomena occurring at surface/cell interface, which are pivotal for innovation in the biomedical field, are envisioned.

### QCM-D AS A TOOL TO MONITOR CELL ADHESION

QCM-D sensors was applied in the study of cell-surface interactions as the piezoelectric mass sensing combined with the monitoring of dissipation changes was related to different colonization stages of cells on surfaces (as attachment, adhesion, and spreading; Saitakis and Gizeli, 2012; Tymchenko et al., 2012). When a cell is interacting with the sensor surface a shift in the 1f can be detected, however the majority of the information on cell state is provided by the energy dissipation factor (1D), index of the fact that the signal is mainly attributable to the viscoelastic behavior of the cell part adhering on the crystal. This is mainly ascribed to the size of cells, which overcome the sensitive limit of the instrument. Indeed, the QCM-D penetration signal is ∼0.25µm in water, allowing the detection of only a fraction of the adhered cell size and consequently effects on frequency shift are mainly associated to adhesion contact points located on the cell membranes (Saitakis and Gizeli, 2012). The contact of the cells with the substrate induces a small decrease of 1f which is not proportional to the real mass adsorbed while in contrast 1D strongly increases having higher shift when adhesion involved a small fraction of the cell membranes (low spreading and limited numbers of focal contact points). During the cell spreading no significant changes are detected in frequency shift compared to adhesion, whereas dissipation slightly decrease due to the rearrangement of cytoskeleton and a subsequent cell stiffening (Tymchenko et al., 2012) responsible for the decrease in damping events (**Figure 2**). The information collected from the temporal

FIGURE 1 | (A) Schematic representation of QCM-D plots reporting frequency factor (1f) and dissipation energy factor (1D) vs. time. QCM-D system is mainly applied in the study of binding efficacy of probing surface of biosensors, assembly of multilayers, competitive protein adsorption studies, and recently for cell/surface interactions. (B) Representative 1D-1f plot signatures for cell adhesion on sensor surfaces (I-initial adhesion, II-formation of attachment points, III-cell spreading, IV-steady state of spread cells, V-production of ECM), cytotoxicity studies (left side: I-water release from cytoskeleton, II-cell lysis; right side: I-cell detachment) and cytoskeleton modification such as platelets activation (I-initial interactions among surface sensor and platelets, II-platelets spreading and pseudopodia formation).

plot of 1f and 1D can be combined in a 1D-1f plot. In this contest, 1D-1f plot (frequency shift plotted vs. energy dissipation shift) provides greater insight into changes in the structural conformation of cells and it is used to monitor the different stages of the cell adhesion: from initial cell/surface interactions (initial adhesion, formation of attachment points) to the following changes in the cytoskeleton conformation characteristics of cell spreading. Westas et al. (2015) monitored the interaction of human gingival fibroblasts on osteoinductive surfaces functionalized with hydroxyapatite nanoparticles when a complete coverage of the sensor surface was achieved. Five consecutive stages were identified by analyzing the recorded QCM-D signals: (i) initial adhesion when cells were close to the surface and started making few attachment points [1f increased as cells behaving like a coupled oscillator (Granéli et al., 2004) and 1D slightly increased], (ii) the number of adhered cells increased as well as the attachment points (1f decreased as the number of attachment point increased and a more homogenous layer is formed at the sensor interface causing 1D increase), (iii) cell spreading occurred (1D rapidly decreased due to the formation of mature focal adhesion complex and 1f decrease as the number of mature focal adhesion complex was lower than the initial adhesion points formed at the second stage), (iv) steady state of spread cells (1f and 1D decreased slowly), and (v) production of extracellular matrix (ECM) from adhered cells (1f decreased due to new mass on the crystal; **Figure 1B**).

The real-time monitoring of cells behavior on surfaces was recently exploited in the optimization of implantable surfaces in order to design ad-hoc substrates having chemical-physical features able to enhance cell adhesion. As the electric charge of the surface material is crucial for the integration in the

overtones) over time of NIH3T3 fibroblasts cultured on a collagen I functionalized sensor. Five events were identified: (A) cell seeding, (B) cell adhesion, (C) cell spreading, (D) cytoskeleton rearrangement after addition of the cytoskeleton-perturbing agent cytochalasin D (2µg/ml) and (D′ ) cell recovery after cytochalasin D removal. Reproduced with permission from Tymchenko et al. (2012). The different phases of cell/surface interaction were confirmed by real-time monitoring of cell morphology and distribution through polarized light microscopy.

biological environment, Kao et al. (2017) employed the potential of QCM-D equipment to follow the NIH3T3 mouse embryonic fibroblast adhesion on surface with different ζ-potentials. Higher ζ-potential surfaces allowed the direct interaction between cells and surface resulting in a high adhesion and cell spreading while the electrostatic repulsions among surface and cells occurring in presence of lower ζ-potentials required the formation of an ECM-like layer prior to cell adhesion and spreading. Kushiro et al. (2016) used the same approach to monitor the cell adhesion process on various SAM surfaces. Interestingly, QCM-D signals allowed to distinguish from different material-cell interactions that were not detected by morphological analysis using confocal microscopy. These findings confirmed the strength of QCM-D setup in the detection of cell changes at the micro and nanoscale, which cannot be measured through conventional methods. Furthermore, Chronaki et al. (2016) employed QCM-D to analyze the difference between the adhesion pattern of normal thyroid and anaplastic thyroid cancer cells demonstrating that starting from dissipation over frequency curves it is possible to accurately discriminate cancer cell from healthy ones, envisioning the use of QCM-D as a promising diagnostic tool.

In addition to cell, bacteria-surface interactions were monitored using QCM-D device (Reipa et al., 2006; Olsson et al., 2010, 2011). Bacteria adhesion is a highly studied topic as adhesion of bacteria on biomedical device surfaces results in biofilm formation, which is one of the more frequent cause of implant failure. QCM-D was applied in the attempt to monitor the antibacterial efficacy of two well-known aminoglycosides, kanamycin A, and neomycin B (Joshi et al., 2015). The modeling of frequency and dissipation energy shifts allowed to get detailed knowledge on mechanism of action of different aminoglycosides in bacteria membrane disruption.

### QCM-D AS A TOOL TO MONITOR CYTOTOXICITY AND CELL VIABILITY

Cells are highly sensitive to perturbations in the surrounding environment and changes interfering with the cell activities can lead to cell death. The monitoring of cell behavior when cells are in contact with bioengineered surfaces, biomaterials, or drugs is fundamental to identify cytotoxic or non-cytotoxic effects. Traditional methodologies to study cytotoxicity and cell viability are based on microscopy and flow cytometry which are unable to monitor cell health in situ and in real-time. Hence, the ability of QCM-D signals to detect cytoskeletal changes can be exploited to assess the cell health in response to external perturbations. Fatisson et al. (2011a) investigated the cellular response during exposure to two different cytomorphic agents (Triton-X 100 and lipopolysaccharide—LPS) on bovine aortic endothelial cells (BAECs) adhering on QCM-D sensor. QCM-D sensors were pre-incubated with BAECs, then coated sensors were mounted into the device chamber and the chemicals were injected in the chamber flowing on the cells attached to the sensor. Both changes in frequency and in energy dissipation were recorded to model cytoskeleton changes in real time. 1D-1f plots revealed two different cell death mechanisms induced by the two cytomorphic agents. High concentrations of Triton-X 100 led to cell lysis, as confirmed by the 1D-1f plot. In this plot, a first phase of 1f and 1D drop was detected suggesting mass adsorption and more rigid layer formation (probably ascribed to Triton-X 100 micelles adsorption and water release from cytoskeleton), followed by the effective cell lysis which induced a 1f increase and a slow 1D drop (**Figure 1B**). On the other hand, LPS exposure resulted in a single slope of the 1D-1f plot with a steady increase of 1f and decrease of 1D. LPS induced a progressive detachment of the BEACs from the QCM-D sensor (from spread cells to round cells) which resulted in mass loss and reduction of damping. Kandel et al. (2014) assessed the viability of human umbilical vein endothelial cells (HUVECs) on different fibronectin-coated QCM-D sensors monitoring the cytoskeleton changes in presence or absence of an actin depolymerizing agent.

Further evidences of the ability of QCM-D to sense cell death were described by Nowacki et al. (2014) which investigated QCM-D signal changes when human breast cancer cells (MCF-7) adhering on quartz crystal surfaces were treated with staurosporine (STS), a protein kinase inhibitor known to induce cell apoptosis. 1D-1f plot displayed a simultaneous increase of 1f and a decrease of 1D, which is typical of apoptosis-mediated cell death.

These findings pave the way for the use of QCM-D not only in the study of cell cytotoxicity but also as a tool to investigate the efficacy of drugs in fighting cancer or other diseases where selective cell death is targeted. The high sensitivity of QCM-D and its ability to continuously monitor the events occurring on the quartz sensors allow to distinguish between the various cell death subtypes and pathways which are fundamental aspects for drug efficacy evaluation (Kepp et al., 2011). Pioneering work of Zhang et al. tracked changes in frequency and energy dissipation of MCF-7 cells treated with resveratrol reporting the use of QCM-D as a cancer drug screening technology for the first time (Zhang et al., 2015a).

### QCM-D AS A TOOL TO MONITOR IMPORTANT PHENOMENA IN CELLS

The opportunity to detect in real time the rearrangement of cell cytoskeleton through QCM-D technology is of pivotal importance in the monitoring of cell biological phenomena such as cell migration and cell differentiation. Data collected by the microgravimetric analyses using the QCM-D system could give information on biological mechanisms which are still difficult to predict. In this context platelets conformational modifications and cytoskeleton changes are involved in the platelet activation process, which is the major player in the formation of blood clotting on biomaterial surfaces. QCM-D was employed to evaluate how composition of bioengineered surfaces affected the platelet activation, which is crucial in the design of blood-contacting devices, as no activation is foreseen to avoid thrombogenic effects. Platelets activation was observed on fibronectin-coated sensors as fibronectin is a highly thrombogenic protein. As discussed above, 1f alone does not give sufficient information and the analysis 1D-1f plot is the more appropriate method to observe changes in cell shape. 1D-1f plot revealed two phases in platelet activation process, a first phase corresponding to initial interactions among surface sensor and cells with a decrease of 1f and a slight increase of 1D followed by a second phase characterized by an increase of 1D ascribed to the platelet spreading and pseudopodia formation (**Figure 1B**). The higher is the slope of the second phases, the lower is the thrombogenicity of the surface as a low anchorage of the platelets on the surfaces resulted in a less rigid layer (Fatisson et al., 2008). Assays on thrombogenicity of high-density polyethylene (HDPE) and polycarbonate (PC) were performed evaluating the effect of surface chemical composition, roughness, and wettability, on platelet activation highlighting that the hemocompatibility effect of these polymers was mainly due to their hydrophobic nature (Fatisson et al., 2011b).

QCM-D system was also applied by Zhang et al. (2015b) to study the difference between healthy erythrocytes (RBCs) and RBCs derived from type 2 diabetic patients, as vascular thrombosis is the most frequent complication in diabetic patients. 1D-1f plot revealed that RBCs from diabetic individuals are stiffer and less prone to cytoskeleton rearrangements than healthy cells. Furthermore, RBCs were co-cultured for 30 min on a QCM-D sensor previously coated with HUVECs. QCM-D signals revealed that pathological RBCs adhered to HUVECs while no adhesion was observed for healthy RBCs.

Among other potential applications, QCM-D was exploited to give novel insights in the study of cancer cell motility which has been recognized as an essential value to quantify the metastatic potential of cancer cells (Friedl and Gilmour, 2009). Tarantola et al. (2010) proposed a QCM-based sensor to determine the motility of four breast cancer lines with different invasiveness by correlating frequency fluctuations recorded by QCM system to collective motility. High fluctuations were associated to collective mobility, which is typical of tumor progression phase. The higher was the fluctuation the more invasive was the cell population as confirmed by the traditionally used Boyden chamber method. This new approach allows measuring the invasiveness of tumor in colonizing adjacent connective tissues helping to identify the degree of tumor malignancy and to define the more appropriate treatment.

### CONCLUSION AND FUTURE PERSPECTIVES

The collected literature highlighted the role of QCM-D technology as a new and non-conventional facility to monitor cells in real time using a label-free methodology. Several phenomena can be detected from initial cell binding, focal adhesion formation, cell spreading, and complex changes in cytoskeleton. The attachment and growth of cells is the key step to achieve device biointegration and many efforts are addressed to the design of bioengineered surface able to enhance cell growth. In vitro and in vivo tests currently applied lack the ability to follow the initial phase of cell/surface interactions which can be monitored modeling QCM-D signals. Complex phenomena occurring during cell/external environment interactions can be distinguished and analyzed by combining temporal frequency and energy dissipation shifts in 1D-1f plots.

To date, QCM-D is rarely applied in materials and surfaces characterization, however, the data collected from this measuring system are unique and could drive to a better understanding of biological phenomena which is fundamental to engineer devices with enhanced biological performances. In addition, taking advantage of the high sensitivity of QCM-D instruments in detecting tiny changes in cell conformation, it is possible to analyse the cytoskeleton changes induced by different drugs on pathological and healthy cells. Consequently, the more

appropriate pharmacological treatment can be defined helping in developing methodologies for advanced and/or personalized therapies as well as screening tools. Furthermore, recent upgrades of the QCM-D device allow to match the recorded 1f and 1D signals with morphology and optical properties of studied platforms by connecting the sensor chamber with additional apparatus such as microscope and ellipsometer.

In conclusion, the recent literature clearly indicates that QCM-D is already a promising methodology to study cell behavior since it provides data which are pivotal to achieve

### REFERENCES


a complete understanding of many biological phenomena consequently helping in tackling the current challenges in the biomedical field.

### AUTHOR CONTRIBUTIONS

CT-T revised the current literature on QCM-D and wrote the main text. IC summarized the state of the art focused on application of QCM-D in cytoskeleton changes analysis. GC revised the manuscript discussing the future trends of QCM-D applications.


fibroblasts on implant surfaces. J. Biomed. Mater. Res. A 103, 3139–3147 doi: 10.1002/jbm.a.35458


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Tonda-Turo, Carmagnola and Ciardelli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Photocatalytic Activity of Polymer Nanoparticles Modulates Intracellular Calcium Dynamics and Reactive Oxygen Species in HEK-293 Cells

Caterina Bossio<sup>1</sup> , Ilaria Abdel Aziz 1,2, Gabriele Tullii 1,2, Elena Zucchetti 1,2 , Doriana Debellis <sup>3</sup> , Mattia Zangoli <sup>4</sup> , Francesca Di Maria<sup>4</sup> , Guglielmo Lanzani 1,2 and Maria Rosa Antognazza<sup>1</sup> \*

#### Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Varpu Seija Marjomäki, University of Jyväskylä, Finland Valeria Piazza, Centro de Investigaciones en Optica, Mexico Grazia Tamma, Università degli studi di Bari Aldo Moro, Italy

> \*Correspondence: Maria Rosa Antognazza Mariarosa.antognazza@iit.it

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

> Received: 18 May 2018 Accepted: 20 July 2018 Published: 23 August 2018

#### Citation:

Bossio C, Abdel Aziz I, Tullii G, Zucchetti E, Debellis D, Zangoli M, Di Maria F, Lanzani G and Antognazza MR (2018) Photocatalytic Activity of Polymer Nanoparticles Modulates Intracellular Calcium Dynamics and Reactive Oxygen Species in HEK-293 Cells. Front. Bioeng. Biotechnol. 6:114. doi: 10.3389/fbioe.2018.00114 <sup>1</sup> Center for Nano Science and Technology@PoliMi, Istituto Italiano di Tecnologia, Milan, Italy, <sup>2</sup> Dipartimento di Fisica, Politecnico di Milano, Milan, Italy, <sup>3</sup> Electron Microscopy Facility, Istituto Italiano di Tecnologia, Genova, Italy, <sup>4</sup> Institute for Organic Synthesis and Photoreactivity, CNR-ISOF, Bologna, Italy

Optical modulation of living cells activity by light-absorbing exogenous materials is gaining increasing interest, due to the possibility both to achieve high spatial and temporal resolution with a minimally invasive and reversible technique and to avoid the need of viral transfection with light-sensitive proteins. In this context, conjugated polymers represent ideal candidates for photo-transduction, due to their excellent optoelectronic and biocompatibility properties. In this work, we demonstrate that organic polymer nanoparticles, based on poly(3-hexylthiophene) conjugated polymer, establish a functional interaction with an in vitro cell model (Human Embryonic Kidney cells, HEK-293). They display photocatalytic activity in aqueous environment and, once internalized within the cell cytosol, efficiently generate reactive oxygen species (ROS) upon visible light excitation, without affecting cell viability. Interestingly, light-activated ROS generation deterministically triggers modulation of intracellular calcium ion flux, successfully controlled at the single cell level. In perspective, the capability of polymer NPs to produce ROS and to modulate Ca2<sup>+</sup> dynamics by illumination on-demand, at non-toxic levels, may open the path to the study of biological processes with a gene-less approach and unprecedented spatio-temporal resolution, as well as to the development of new biotechnology tools for cell optical modulation.

Keywords: conjugated polymer, bio-organic electronics, cell optical stimulation, Ca2<sup>+</sup> imaging, reactive oxygen species, light, organic semiconductor, photomodulation

### INTRODUCTION

By and large, cells are essentially transparent and do not directly respond to light. Visible light is, however, a vital drive for animal behavior and species development. Accordingly, natural evolution resulted in the appearance of photoreceptors cells containing light-sensitive opsin proteins that provide a broad range of photo-induced functions (Morshedian and Fain, 2017). Photoreceptors detect light and trigger a phototransduction chain that ends in a biochemical signaling. Since long

ago, scientists have explored possibilities to artificially induce light sensitivity in living tissues, in order to control physiological functions by photons. This does not simply mean copying nature in order to restore lost photo-functions in damaged organs, but it opens up a wealth of new applications in research, diagnostics and therapy. Light-induced cell control has indeed a number of advantages with respect to more traditional electrical stimulation, such as space and time resolution, reduced invasiveness and selectivity. Optogenetics is one major approach (Deisseroth, 2011). Unfortunately, in spite of the important record of experimental results, it requires the viral transfer of the sample (cell or organ) with exogenous DNA, which raises important safety constraints for application in human subjects and is still considered to be a critical limitation.

For this reason, there is a number of attempts to induce light sensitivity by a gene-less approach. The direct heating of water with a near IR laser radiation (Shapiro et al., 2012), the use of organic and inorganic devices such as photocapacitors and photodetectors (Goda and Colicos, 2006; Farah et al., 2013; Antognazza et al., 2015; Martino et al., 2015; Feyen et al., 2016; Maya-Vetencourt et al., 2017), and the introduction of conducting or semiconducting nanoparticles (Colombo et al., 2016) are among reported strategies. The latter approach is quite interesting, because it allows an easy sensitization of the target tissue and, in principle, it can select specific sites within cell sub-compartments. Nano-scaled inorganic semiconductors, such as quantum dots, are widely studied for the peculiar optical properties induced by the reduced dimensionality. Indeed, they have a huge potential for biological applications, being their dimensions close to the size of biomolecules as proteins and nucleic acids. Although there are many studies on surface functionalization being carried out, the cytotoxicity of QDs still remains an issue, halting their possible biomedical applications (Medintz et al., 2005). Despite this, they offer a strongly selective and sensitive tool for biosensing (Howes et al., 2014) and fluorescence imaging (Bruni et al., 2017). The plasmonic resonance of metallic nanoparticles has been also successfully exploited to gain efficient photo-thermal processes upon NIR radiation (Yang et al., 2017), but in this case high photo-excitation densities are usually required.

Here, we focus our attention on organic semiconducting nanoparticles (NPs), based on poly(3-hexylthiophene) (P3HT). In vitro use of conjugated polymer NPs has been widely reported in the literature (Tuncel and Demir, 2010; Feng et al., 2013). However, the exploitation of excellent NPs biocompatibility properties has been mainly limited, so far, to imaging and drug delivery applications (Feng et al., 2010). Conversely, their possible use as functional, light-sensitive actuators was not intensively investigated (Zangoli et al., 2017). Only very recently, we demonstrated an optical modulation effect of P3HT NPs on animal model of Hydra vulgaris. P3HT NPs act on the animal both at a macroscopic level, by inducing a behavioral change, and from a molecular point of view, by causing enhanced expression of opsin proteins (Tortiglione et al., 2017).

The goal of the present work is to explore the photoinduced coupling mechanism between P3HT NPs and living cells. Interestingly, we demonstrate that upon illumination polymer NPs both increase the intracellular production of reactive oxygen species (ROS) and modulate Ca2<sup>+</sup> dynamics. ROS, such as the superoxide anion O<sup>−</sup> 2 and the hydrogen peroxide H2O2, are widely involved in both physiological and pathophysiological processes (Halliwell, 2015). ROS overproduction is highly detrimental since it induces non-specific damage of proteins, lipids, and DNA (Finkel and Holbrook, 2000), causing oxidative stress (Fu et al., 2014) and leading in some cases to irreversible cell damage and apoptosis (Martindale and Holbrook, 2002). Conversely, endogenously produced ROS are essential to life (Trachootham et al., 2009), being involved in different biological functions: among others, signal transduction (Kamata and Hirata, 1999), neurotransmission (Newsholme et al., 2010), blood pressure modulation (Lee and Griendling, 2008), immune system control (Glassman, 2011), and metabolism regulation (Dröge, 2002). These versatile functions of ROS are finely modulated by their amount, duration and localization, and are currently the object of intensive investigation. Alteration in physiological ROS production leads to important pathological conditions, including autoimmune, cardiovascular, and neurodegenerative diseases, so that several ROS-responsive drug delivery systems are under investigation (Saravanakumar et al., 2017). Overall, controlled release of ROS can have highly beneficial therapeutic effects (Bergamini et al., 2004). However, suitable, fully biocompatible, precise, externally modulated and minimally invasive tools are almost completely missing at the moment. Optical modulation represents an interesting route to achieve it (Wang et al., 2012; Waiskopf et al., 2016).

Not invasive ways to modulate calcium ions flow are attracting considerable attention as well, given their key role in cell metabolism and inter-cell communication processes. Several attempts have been reported to modulate Ca2<sup>+</sup> by external mediators, for instance by using magnetic NPs (Huang et al., 2010; Stanley et al., 2012; Bar-Shir et al., 2014; Lee et al., 2014; Chen et al., 2015; Wheeler et al., 2016), metallic NPs (Nakatsuji et al., 2015), organic molecules and polymers (Stein et al., 2013; Lyu et al., 2016; Takano et al., 2016) and other carbon-based materials (Miyako et al., 2014; Chechetka et al., 2016).

Interestingly, there is a close interplay between ROS and Ca2<sup>+</sup> dynamics (Chakraborti et al., 1999; Brookes et al., 2004; Görlach et al., 2015), whose intensive investigation may benefit from the availability of not-invasive, locally confined and reversible tools.

Here, we show that photoexcitation of P3HT NPs internalized within Human Embryonic Kidney (HEK-293) cells leads to an increase of ROS level, associated to a variation of the intracellular Ca2<sup>+</sup> dynamics. In perspective, this work may pave the way to the development of new tools for local, optical modulation of the cell physiological activity.

### MATERIALS AND METHODS

### Cell Cultures

HEK-293 cells were grown in Dulbecco's modified Eagles' medium (D-MEM, Sigma Aldrich) with 10% fetal bovine serum (FBS, Sigma Aldrich), supplemented with 2 mM glutamine (Sigma Aldrich), 100 U/ml streptomycin (Sigma Aldrich) and 100 U/ml penicillin (Sigma Aldrich). Cells were kept in T-75 culture flasks and maintained in incubator at 37◦C in a humidified atmosphere with 5% CO2. After reaching 80–90% of confluence, cells were detached by incubation with 0.5% trypsin-0.2% EDTA (Sigma Aldrich) for 5 min and plated for experiments. To promote cell adhesion, a layer of 2 mg/ml fibronectin (from bovine plasma, Sigma Aldrich) in phosphate buffer saline (PBS, Sigma Aldrich) was deposited on the surface of the glass coverslips and incubated for 30 min. After rinsing the fibronectin with PBS, cells were plated in their culture medium and eventually treated with polymer NPs.

### Preparation and Characterization of P3HT NPs Dispersions

Sterile P3HT and Polystyrene (PS) NPs were prepared by the reprecipitation method working under a laminar flow hood. P3HT (regio-regular, Sigma Aldrich) was dissolved in tetrahydrofuran (THF, Sigma Aldrich) and the resulting polymer solution was added drop-wise to sterilized water under magnetic stirring. The resulting colloidal dispersion was put in a dialysis sack and subjected to dialysis against 2 l sterilized water overnight to remove the residual organic solvent. Afterwards, the water colloidal suspension was centrifuged at different rates for 10 min, from 2,000 to 8,000 rpm, separating every time the supernatant from the precipitate, to obtain a wide range of samples with different particle dimensions (from 100 to 600 nm) and different optical density (OD). The NPs used in this work have an average hydrodynamic radius of 237 ± 82 nm, a polydispersity index of 0.12. The ODs of the dispersion administered to the cell cultures are 0.05, 0.2, 0.4. NPs are dissolved in a volume of 1 ml of culture medium 24 h before measurements. Dynamic Light Scattering (DLS) measurements were performed with a Nanobrook Omni Particle Size Analyzer, with a wavelength of 659 nm in back-scattering mode. Particles were dispersed in distilled water or cell culturing medium during analysis. Measurements were taken at 25◦C. Zeta potential evaluation was performed using Smoluchowski equation. The optical absorption spectra were acquired by using Perkin Elmer Lambda 1050 UV/Vis/NIR Spectrophotometer and, for NPs upload estimation, TECAN Spark 10M Plate reader. The photoemission spectra were recorded by using HORIBA Jobin Yvon NanoLogTM spectrofluorimeter.

### Scanning Electron Microscopy (SEM)

Cells treated with P3HT NPs and grown on an indium-tin oxide (ITO) covered coverslip were fixed with Glutaraldehyde (Sigma Aldrich) 2.5% for 20 min. Then cells were dehydrated with incubation in solutions at increasing concentration of ethanol (Sigma Aldrich). Images were acquired with the Scanning Electron Microscope MIRA3 TESCAN.

### Transmission Electron Microscopy (TEM)

HEK293 cells were incubated with P3HT NPs for 24 h and were processed following a published protocol (Marotta et al., 2013). Briefly the cells after incubation were fixed for 1 h in 1.2% glutaraldehyde in 0.1 M sodium cacodylate buffer (Sigma Aldrich, pH 7.4), post fixed in 1% osmium tetroxide (Sigma Aldrich) in the same buffer and en bloc stained with 1% uranyl acetate aqueous solution. The cells were then dehydrated in a graded ethanol series and embedded in epoxy resin (Epon 812, TAAB). Semi-thin and thin sections of the embedded cell monolayer were cut with an ultramicrotome (UC6, Leica) equipped with a diamond knife (Diatome). Projection images were acquired using a JEOL JEM 1011 transmission electron microscope operating at at 100 kV of acceleration voltage and recorded with a 2 Mp charge-coupled device (CCD) camera (Gatan Orius SC100).

### Immunocytochemical Studies

Cells treated with P3HT NPs grown on fibronectin-coated glass coverslips were washed twice with PBS and fixed for 20 min at RT in 4% paraformaldehyde (Sigma Aldrich) and 4% sucrose (Sigma Aldrich) in 0.12 M sodium phosphate buffer, pH 7.4. Labeling with phalloidin and fluorescein Isothiocyanate (Sigma Aldrich) was applied in GDB buffer (30 mM phosphate buffer, pH 7.4, containing 0.2% gelatin, 0.5% Triton X-100 and 0.8 M NaCl (Sigma Aldrich) for 2 h at RT.

### Viability Assay

In order to preliminarily evaluate the NPs cytotoxicity, the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] (Sigma Aldrich) assay was performed on HEK-293 cells. Cells were seeded in 12 wells plates at a density of 2 × 10<sup>4</sup> cells/well, together with the NPs. Cell proliferation was evaluated after 24, 48, 72, and 96 h incubation. For each time point, the growing medium was replaced with RPMI without phenol red (Sigma Aldrich) containing 0.5 mg/ml of MTT. The samples were incubated again for 3 h at 37◦C in dark. Formazan salt produced by cells through reduction of MTT was then solubilized with 200 µl of ethanol and the absorbance was read at 560 and 690 nm. The proliferation cell rate was calculated as the difference in absorbance at 560 and 690 nm. Photocytoxicity of NPs- treated cells subjected to optical excitation has been evaluated by Fluorescein Diacetate (FDA, Sigma Aldrich) assay. FDA is a non-fluorescent molecule, which is hydrolyzed to fluorescent fluorescein in viable cells only. Different cohorts of HEK-293 cells (treated/untreated with P3HT NPs, subjected/not subjected to photoexcitation) were incubated for 5 min with FDA 5µM. After careful wash outs of the excess FDA from the extracellular medium and repeated rinses with physiological saline solution, representative images were acquired (excitation/emission wavelengths, 490/520 nm, integration time 100 ms) using an inverted fluorescence microscope (Nikon Eclipse Ti; 20× objective), equipped with an Analog-WDM Camera (Cool Snap Myo, Photometrics). Photocytotoxicity tests have been carried out in different experimental conditions, namely: (i) at different NPs doses, corresponding to OD values in the range 0–0.5, at 2 DIV/24 h after illumination. The photoexcitation protocol used in FDA assay is the same employed in ROS detection experiments (see below); (ii) at fixed NP dose, corresponding to OD = 0.2, at different time points after the photoexcitation protocol, up to 4 DIV/72 h after illumination. Specific care was taken in order to image, at different time points, exactly the same sample area previously subjected to optical excitation. All experiments were carried out in relevant control samples as well (untreated and/or not illuminated cell cultures).

### Chronoamperometry Measurements

Photocurrent measurement was carried out by using a potentiostat/galvanostat (Autolab, PGSTAT 302N). The electrochemical cell was in three-electrode configuration and was divided into two compartments, connected by a saline bridge. One compartment contained the reference and the counter electrodes, a saturated KCl Ag/AgCl and Pt wire respectively, immersed in the pure electrolyte (phosphate buffer 10 mM, pH 7). The second compartment contained the Indium-Tin-Oxide (XynYan Technology, 15 nm thickness, sheet resistance 15 ohm/sq) working electrode and the P3HT NPs dissolved in the same electrolyte solution. Chronoamperometry measurements were performed at the electrochemical equilibrium, by applying a potential equal to the open circuit potential (OCP), and by changing the applied voltage in the range (OCP - 300 mV) ≤ V ≤ OCP vs. Ag/AgCl. A continuous light source (Thorlabs LED M470L3-C5, 470 nm central emission wavelength) was used for the photoexcitation, with a power density of 2.7 mW/mm<sup>2</sup> . The photocurrent data at fixed wavelengths were obtained by exciting the NPs dispersion with six CW LED light sources in the visible spectrum. Peak emission wavelengths and corresponding power densities are: λ<sup>1</sup> = 470 nm, PD1 = 3.45 mW/mm<sup>2</sup> ; λ<sup>2</sup> = 505 nm, PD2 = 1.69 mW/mm<sup>2</sup> , λ<sup>3</sup> = 530 nm, PD3 = 1.36 mW/mm<sup>2</sup> , λ<sup>4</sup> = 617 nm, PD4 = 2.02 mW/mm<sup>2</sup> , λ<sup>5</sup> = 627 nm, PD5 = 2.01 mW/mm<sup>2</sup> , λ<sup>6</sup> = 655 nm, PD6 = 2.21 mW/mm<sup>2</sup> ).

### Reactive Oxygen Species (ROS) Detection

Dichlorofluorescein diacetate (H2DCF–DA, Sigma Aldrich), 2-[6-(4′ -amino)-phenoxy-3H-xanthen-3-on-9-yl] benzoic acid (APF, Sigma Aldrich) and 2-[6-(4′ -Hydroxy)-phenoxy-3Hxanthen-3-on-9-yl] benzoic acid (HPF, Sigma Aldrich) were employed for intracellular detection of Reactive Oxygen Species (ROS). Sterile P3HT and PS NPs were administered to the cells at the plating step, by diluting them in the extracellular medium up to the desired concentration, corresponding to OD = 0.05, 0.2, 0.4. At 1DIV HEK-293 cells already showed uniform coverage of the substrate (about 40% confluence) and NPs were efficiently internalized within the cell cytosol. Excess, non-internalized NPs were removed by subsequent, repeated rinses of the extracellular medium. NPs treated cells and control, untreated cells were photo-excited by illuminating each sample for 2 min with a LED system (Lumencor Spectra X light engine, λ = 540 nm, P<sup>D</sup> = 95 mW/mm<sup>2</sup> ). Subsequently, cells were incubated with the ROS probes for different times according to the employed probe (tH2DCF−DA = 30 min, tAPF = tHPF = 40 min). After careful wash-out of the excess probe from the extracellular medium, the fluorescence of the probes was recorded (excitation/emission wavelengths, 490/520 nm; integration time, 70 ms for H2DCF– DA and 500 ms for HPF e APF) with an inverted microscope (Nikon Eclipse Ti), equipped with an Analog-WDM Camera (Cool Snap Myo, Photometrics). The same protocol was employed for control, untreated cells. Variation of fluorescence intensity was evaluated over Regions of Interest covering single cells areas, and reported values represents the average over multiple cells. See figure captions and SI for additional details about statistical analysis. Image processing was carried out with ImageJ and subsequently analyzed with Origin 8.0.

## Intracellular Ca2<sup>+</sup> Measurements

HEK-293 cultures were loaded for 30 min at 37◦C with 1µM Fluo-4 (Life Technologies) in extracellular solution (5 mM HEPES; 135 mM NaCl, 5,4 mM KCl, 1 mM MgCl2, 1,8 mM CaCl2, 10 mM glucose, all purchased from Sigma Aldrich; pH 7.4) and Ca2<sup>+</sup> free extracellular solution (10 mM HEPES; 150 mM NaCl, 3 mM MgCl2, 5 mM EGTA, all purchased from Sigma Aldrich; pH 7.4). Then, cells were washed for 10 min with pre-warmed extracellular solution before recordings. Cells treated with P3HT and PS NPs, as well as untreated samples, were illuminated for 3 min (emission peak wavelength, 485 nm; photoexcitation density, 2.1 mW/mm<sup>2</sup> and 14.6 mW/mm<sup>2</sup> , spot size 0.9 mm<sup>2</sup> ) by using a LED source light (Lumencor Spectra X light engine). Videos were collected with an inverted microscope Nikon Eclipse Ti equipped with an Analog-WDM Camera (Cool Snap Myo, Photometrics). Variation of fluorescence intensity was evaluated over Regions of Interest covering single cells areas, and reported values represents the average over multiple cells. See Figure captions and SI for additional details about statistical analysis. Image processing was carried out with ImageJ and subsequently analyzed with Origin 8.0. ROS inhibition was obtained by administration of N–Acetyl–Cysteine (NAC, Sigma Aldrich), dissolved in extracellular solution (with and without Ca2+) at a final concentration of 10µM.

### Statistical Analysis

See SI section.

## RESULTS

### Optical and Microscopy Characterization of Polymer P3HT Nanoparticles (P3HT NPs)

Polymer NPs, entirely constituted by poly(3-hexylthiophene), have been synthesized by the re-precipitation method in sterile conditions, as previously described (Zucchetti et al., 2016). P3HT NPs have an average hydrodynamic diameter of 237 ± 82 nm and a polydispersity index of 0.12, as measured by dynamic light scattering (**Figure S1A**). They also show excellent colloidal stability, being characterized by a zeta-potential value of −35 ± 8 mV. Recent literature has firmly established the importance of studying in detail the interaction between polymer-coated NPs and proteins, which largely affects cellular uptake and cytotoxicity (Hühn et al., 2013). Here, colloidal stability within the cell culturing medium (D-MEM, without the pH indicator phenol red) was evaluated over time, by carrying out both DLS measurements (i.e., by assessing modifications of the hydrodynamic diameter, **Figure S1B**), UV-Vis absorption and photoluminescence spectroscopy (**Figures S1C,D**). Three different NPs dispersions have been considered, corresponding to optical density values of 0.2, 0.4, and 0.5. In all cases, NPs dispersions do not show significant aggregation effects within the cell culturing medium, up to 48 h. The optical absorption and emission spectra of P3HT NPs aqueous dispersion are reported in **Figures 1A,B**, respectively (inset, P3HT chemical structure). It has been recently demonstrated that the optical properties of P3HT NPs are well preserved within the cell cytosol environment (Zucchetti et al., 2016). Preliminary studies of P3HT NPs toxicity both in Human Embryonic Kidney (HEK-293) cells (Zucchetti et al., 2016) and in Hydra Vulgaris animal models (Tortiglione et al., 2017) have been previously carried out in dark condition, and no clear adverse effects were reported. In this work, P3HT NPs were administered to the HEK-293 cell culturing medium at the cell plating step, and their cytotoxicity was preliminarily evaluated by measuring cellular proliferation through the MTT method. The assay relies on the capability of living cells to metabolize/reduce a water-soluble tetrazolium salt (3-(4′ ,5′ -dimethylthiazol-2′ -yl)- 2,5-diphenyl-2H-tetrazolium hydrobromide, MTT) into a waterinsoluble formazan product, which has a characteristic purple color. The optical absorption of formazan is thus considered a clear indicator of cell viability and capability to proliferate, being proportional to the number of living cells. **Figure S2** shows the formazan optical absorption recorded in NPs treated and -untreated cells at four different time points, 1, 2, 3, and 4 days after incubation (DIV). The presence of NPs

FIGURE 1 | Light sensitive P3HT NPs and HEK-293 living cells. (A,B) Optical absorption and fluorescence (PL) spectra of P3HT NPs in aqueous dispersion. P3HT chemical structure is also shown in the inset. (C–F) Top view scanning electron microscopy (SEM) images of HEK-293 cells treated with P3HT NPs, at increasing magnification. Scale bar, 1µm.

leads in general to a reduced cell proliferation with respect to untreated samples; the absorption percentage variation between treated and untreated cell cultures is about 45% at DIV 1. This difference, however, shows a partial recovery over time, and amounts at 20%, from DIV 2 onwards, up to DIV 4. Scanning electron microscopy (SEM) images (**Figures 1C–F**) document the very well defined spherical geometry and the good surface smoothness of NPs. The limited presence of aggregates is a further index of the high stability properties within the cellular environment of P3HT NPs, despite the absence of any surfactant agent during the fabrication protocol. P3HT NPs can be easily visualized by fluorescence imaging, thanks to their intrinsic photoluminescence peaking in the red spectral region (**Figure 1B**). Effective P3HT NPs internalization within HEK-293 cell cytosol was previously assessed by confocal laser scanning microscopy (Zucchetti et al., 2016). Here, P3HT NPs cellular uptake was quantitatively estimated by using two different methods, based on fluorescence imaging (**Figure S3**) and variation of the optical absorption spectrum of the NPs dispersion before and after 24 h administration to the cell cultures (**Figure S4**). In both cases, three initial concentrations of P3HT NPs dispersions were considered, corresponding to optical density values of 0.05, 0.2, and 0.4. In the latter two cases, effective NPs upload is detected, in the order of few hundreds of NPs /cell, in agreement with previously reported CLSM data. Conversely, for the lowest considered concentration of the dispersion, a considerably lower fraction of NPs undergoes efficient internalization within the cell cytosol (see details in the SI section). Finally, for control experiments, we used light-insensitive polystyrene NPs (PS NPs). These NPs have dimensions and zeta-potential values (hydrodynamic diameter, 225 ± 72 nm; zeta-potential,−41.8 ± 1.1 mV) similar to P3HT NPs, but do not have remarkable optoelectronic properties: they have negligible absorption in the blue-green spectral range, do not generate electric charges upon optical excitation and are electrically insulating materials, thus representing a good model for control, optically and electrically inert NPs. Detailed characterization of PS NPs properties and colloidal stability within the cell culturing medium over time are reported in the SI section (**Figure S5**). Transmission electron microscopy (TEM) analysis corroborates SEM and fluorescence imaging showing the presence of endosome-like compartments inside the cell cytoplasm. These compartments are filled with electron-dense nanoparticles compatible in shape and size with the P3HT NPs (**Figure S11**).

### P3HT NPs Photo-Electrochemical Characterization

The photo-electrochemical activity of P3HT polymer thin films in an aqueous saline environment has been reported in a number of recent works (Lanzarini et al., 2012; Bellani et al., 2015; Giron et al., 2016; Tullii et al., 2017). The observed photocurrent generation has been attributed to photoelectrochemical reactions, leading to oxygen and, less efficiently, to hydrogen reduction. In the present case, P3HT NPs are dispersed in a phosphate buffer solution (10 mM, pH 7, optical density of 2). Their capability to photo-generate charges has been initially assessed. We carried out chrono-amperometry and spectral responsivity measurements on the NPs dispersion in a three electrode configuration, employing a planar ITO slab, a saturated KCl Ag/AgCl and a Pt wire as the working (WE), the reference (RE) and the counter electrode (CE), respectively (**Figure 2A**). Visible light is incident from the glass substrate of the ITO slab. In order to avoid electrode contamination by P3HT NPs, the WE, immersed in the dispersion, was separated from the RE and CE compartment by a saline bridge. Prior to chrono-amperometry measurements, the open circuit potential (OCP) value was estimated, being about 150 mV. Photocurrent has been measured at different excitation wavelengths by using six LED light sources (**Figure 2B**, red diamond symbols). The spectral dependence shows a symbatic behavior, closely resembling the optical absorption spectrum of the P3HT thin film (red dashed line). This indicates that the recorded current is undoubtedly due to charge photogeneration by polymer NPs and allows to exclude spurious reduction/oxidation effects. Chronoamperometry measurements have been then carried out, by applying fixed bias V in the range (OCP-300 mV) ≤ V ≤ OCP, with steps of 100 mV amplitude (**Figure 2C**). Upon light excitation (470 nm peak excitation wavelength, 3 min stimuli duration, 2.7 mW/mm<sup>2</sup> photoexcitation density), a photocurrent signal is recorded, with increasing amplitude at increasingly negative bias. The external quantum efficiency, defined as the ratio between the number of electrons flowing in the external circuit and the number of incident photons per unit time at a fixed wavelength (470 nm in our case), is very low, in the order of 10−<sup>6</sup> , four orders of magnitude lower than the value obtained in hybrid solid/liquid devices based on bulk heterojunction polymer thin films, namely P3HT doped with the fullerene derivative electron acceptor phenyl-C61-butyric acid methyl ester (Lanzarini et al., 2012; Bellani et al., 2015). Still, these results unequivocally demonstrate that P3HT NPs can photo-generate free charges, which interact with the surrounding electrolyte and sustain photo-electrochemical reactions. Photocurrent dynamics recorded by using P3HT polymer samples prepared in different ways (namely, P3HT thin films spin-coated from solution onto the ITO WE, P3HT NPs thin films casted from dispersion and ITO WE previously subjected to P3HT NPs exposure) all show negative photocurrent signal, like in the case of NPs dispersion (**Figure S6**), thus indicating similar interfacial charge transfer mechanisms at the ITO electrode. In particular, in accordance with the observed photocurrent sign, photogenerated holes (carried by the ionized NPs or NP+) are efficiently transferred to the WE, while electrons react with species present in solution. In analogy with results obtained with polymer thin films (Tullii et al., 2017), we believe that the predominant photoelectrochemical reaction at NPs surface is reversible oxygen reduction, which leads to the creation of reactive oxygen species, such as O<sup>−</sup> 2 . This is fully confirmed by photocurrent measurements with P3HT NPs dispersion in de-oxygenated condition, and after partial re-oxygenation (**Figure 2D**).

The energetic alignment of HOMO and LUMO levels of P3HT polymer (∼5 and ∼3.5 eV, respectively) with respect to the ITO work function (∼4.5 eV for oxygen-plasma treated substrates,

well aligned with the polymer HOMO level) and to the 1 electron oxygen reduction reaction O2/O**.**<sup>−</sup> 2 (– 0.33 V vs. SHE at pH 7, corresponding to ∼4.1 V in the absolute scale, well aligned with the P3HT LUMO level) are in agreement with this picture. P3HT NPs work function has been experimentally evaluated by Kelvin Probe measurements, obtaining values slightly lower than P3HT polymer thin films [4.5 vs. 4.8 eV, respectively, in line with recent literature (Di Maria et al., 2017)], which may further facilitate the occurrence of oxygen reduction reactions.

temporal window corresponding to optical excitation.

### Production of Reactive Oxygen Species (ROS)

Photocurrent measurements demonstrate the occurrence of photo-electrochemical reactions, possibly involving reactive oxygen species (ROS) as intermediates and/or final products.

We experimentally compared the modulation of ROS production within the cell cytosol, in NPs treated and control samples, either in dark or subjected to photoexcitation. To this goal, we use intracellular ROS fluorescent probes, namely 2,7-dichlorodihydrofluorescein diacetate (H2DCF-DA), 2-[6-(4′ -amino)-phenoxy-3H-xanthen-3-on-9-yl] benzoic acid (APF) and 2-[6-(4′ -Hydroxy)-phenoxy-3H-xanthen-3-on-9-yl] benzoic acid (HPF). H2DCF-DA has been widely employed for the detection of ROS in several cell models, due to its capability to rapidly diffuse through the cellular membrane (Wang and Joseph, 1999). Cellular esterase first hydrolyzes the non-fluorescent H2DCF-DA to H2DCF, which is then oxidized by ROS and originates the fluorescent compound 2,7-dichlorofluorescein (DCF). Three different samples cohorts have been taken into account: non treated cells, cells treated with P3HT NPs, and cells treated with electrochemically inert, light-insensitive PS NPs. For each experimental group, two different conditions have been considered, cells subjected to illumination (2 min, 95 mW/mm<sup>2</sup> photoexcitation density) prior to ROS detection analysis, and cells kept into dark conditions. We found that cells treated with NPs and subjected to illumination protocol show a statistically significant increase in ROS production, as compared to control untreated cells, and/or to treated cells in dark condition (**Figure 3A**). Since H2DCF-DA is sensitive to a large variety of different ROS, including H2O2, HO**.** , ROO**.** , we also used APF and HPF, which exhibit strong fluorescence enhancement upon Odearylation, as induced by reactive species. APF and HPF provide complementary information with respect to H2DCF-DA because they are sensitive to HO**.** , ONOO−, and HOCl, but not to H2O<sup>2</sup> (Gomes et al., 2005). In addition, they show much higher resistance to auto-photo-oxidation, being more suitable for the comparative evaluation of the effects of visible light illumination. APF and HPF data, shown in **Figures 3B,C** respectively, confirm that significant production of ROS occurs only in the case of cells treated with polymer NPs and previously subjected to the illumination protocol. **Figures 3D–F** show representative images of the fluorescence enhancement obtained in HEK-293 cells treated with P3HT NPs and illuminated for 2 min prior to ROS detection tests with H2DCF-DA (**Figure 3D**), APF (**Figure 3E**), and HPF (**Figure 3F**).

## Modulation of Intracellular Ca2<sup>+</sup> Dynamics

Recent literature highlighted the mutual crosstalk between ROS and Ca2<sup>+</sup> dynamics. It has been reported that the cell redox state can efficiently modulate the activity of a variety of Ca2<sup>+</sup> channels, and, on the other hand, Ca2<sup>+</sup> signaling actively contributes to control the production of ROS<sup>47</sup> .

To determine if the observed ROS enhancement leads to specific modulation of calcium dynamics, we monitored the intracellular Ca2<sup>+</sup> ion concentration by using Fluo-4 as a calcium indicator. We employed two different light photoexcitation densities, namely 2.1 and 14.6 mW/mm<sup>2</sup> .

Taking advantage of the partial spectral overlap between the light absorption of the polymer NPs and the excitation spectrum of the Fluo-4 calcium indicator, we used the same illumination protocol both to optically excite the P3HT NPs and to monitor Ca2<sup>+</sup> dynamics at the same time. In all considered cases continuous illumination for 3 min has been employed, coherently with protocols used in photoelectrochemical measurements and ROS detection analysis.

While optical stimulation at low excitation density (2.1 mW/mm<sup>2</sup> ) did not lead to significant differences among considered experimental conditions (**Figures S7A,D,G**), higher intensity (14.6 mW/mm<sup>2</sup> ) optical excitation of P3HT NPs treated cells led to significant activation of Ca2<sup>+</sup> dynamics (**Figure 4**). **Figure 4A** shows 3 representative Ca2<sup>+</sup> transient dynamics for each considered case; the whole ensemble of registered curves (n ≥ 100) is reported in **Figures S7B,E,H**. Ca2<sup>+</sup> peak amplitude shows a relative increase of about 60% both with respect to untreated cells and to cells treated with PS NPs. No significant changes in Ca2<sup>+</sup> dynamics were detected in positive

FIGURE 3 | Intracellular ROS production by P3HT NPs optical excitation. (A–C) Fluorescence variation intensity measured in control (gray), P3HT NPs–treated (red), and PS NPs treated (blue) cells, due to ROS sensitive intracellular probes (H2DCF-DA, A, APF, B, and HPF, C). Data are represented as mean ± SE values, over statistical samples of n = 34 (control light), n = 81 (control dark), n = 49 (P3HT NPs light), n = 102 (P3HT NPs dark), n = 109 (PS NPs light), n = 105 (PS NPs dark) for H2DCF-DA; n = 57 (control light), n = 26 (control dark), n = 90 (P3HT NPs light), n = 63 (P3HT NPs dark), n = 70 (PS NPs light), n = 69 (PS NPs dark) for APF; n = 110 (control light), n = 118 (control dark), n = 137 (P3HT NPs light), n = 120 (P3HT NPs dark), n = 147 (PS NPs light), n = 120 (PS NPs dark) for HPF, where n represents the number of cells over three different experiments for each condition. Statistical significance has been evaluated by one-way ANOVA analysis followed by post-hoc Tukey test. P-values of the test are assigned as follows: \*\*\* for p < 0.001. Statistically non-significant results are not marked in the figures. All details of statistical analysis are reported in the SI section. (D-F) Representative images of ROS-induced fluorescence response of H2DCF-DA (D), APF (E), HPF (F) in P3HT NPs treated cells. Fluorescence images have been acquired after carrying out the protocol for NPs excitation, as described in detail in the Methods section. Scale bar, 100µm.

and negative control experiments, i.e., in illuminated, untreated cells and illuminated cells treated with inert, light-insensitive PS NPs (**Figure 4B**). Representative fluorescence images at different time points are reported in Figures S12A–H, to give a direct reply of the reported curves.

The Ca2<sup>+</sup> sources involved in the stimulation, e.g., intracellular stores and/or extracellular environment, were investigated in Ca2+-free conditions, and in the presence of 5 mM of the Ca2<sup>+</sup> chelator ethylene glycol tetraacetic acid, EGTA. **Figure 4D** shows 3 representative curves for each case. The whole data set is reported in **Figures S8A,D,G** at lower intensity photoexcitation and **Figures S8B,E,H** at a higher intensity photoexcitation. Representative fluorescence images at different time points are reported in **Figures S12I–R**, to give a direct reply of the reported curves. **Figure 4E** shows that the relative increase of fluorescence intensity exhibited by P3HT NPs-treated cells in absence of extracellular Ca2<sup>+</sup> is not significantly different from the one obtained in standard conditions (1F = 1.92 and 1F = 1.85, respectively). This indicates that Ca2<sup>+</sup> influx through the plasma membrane does not play a significant role and that the main origin of activation is in intracellular Ca2<sup>+</sup> stores.

Based on results reported so far, we conjecture that lightevoked production of ROS in the cell cytosol acts as the main source of the observed intracellular Ca2<sup>+</sup> increase. In order to verify this hypothesis, after cell cultures illumination, we supplemented cell cultures with a well-known ROS inhibitor, N-Acetyl-L-cysteine (NAC), and thereafter we carried out Ca2<sup>+</sup> fluorescence imaging experiments, reported in **Figures 4C,F** (complete data sets are shown in panels **C,F,I** of **Figures S7, S8**). Interestingly, Ca2<sup>+</sup> dynamics were significantly silenced, in the presence of extracellular Ca2<sup>+</sup> as well as in Ca2<sup>+</sup> free extracellular solution.

### Dependence of ROS Production and Ca2<sup>+</sup> Modulation on P3HT NPs Concentration

We investigated the effect of different polymer NPs dispersion concentrations on ROS production and modulation of Ca2<sup>+</sup> dynamics (**Figure 5**). We compared three concentrations of NPs, corresponding to optical densities of 0.05, 0.2, and 0.4. **Figures 5A–C** show the effect of different NPs doses on ROS production, as evaluated by using H2DCF-DA, HPF, and APF probes, respectively. In all the considered cases, the increased number of internalized P3HT NPs does not influence, per se, the production of ROS, which is not statistically different from the values recorded in untreated, control samples (dark condition, gray histograms). Conversely, upon visible light excitation, NPs concentration strongly affects ROS evolution (light condition, red histograms). We notice that the lowest NP concentration, corresponding to an initial optical density

FIGURE 4 | Optical modulation of intracellular Ca2<sup>+</sup> dynamics. (A) Representative Ca2<sup>+</sup> dynamics recorded in control (gray solid lines), P3HT NPs treated cells (red solid lines) and PS NPs treated cells (blue solid lines). (B) Fluorescence relative increase of the Fluo-4 calcium indicator. Data are shown as mean ± SE, over statistical sample sets of n = 71, n = 68, n = 55 cells (for control, P3HT NPs and PS-NPs samples, three experiments for each condition, respectively). Statistical analysis has been carried out by one-way ANOVA, followed by post-hoc Tukey test. P-values of the test are assigned as follows: \*\*\* for p < 0.001. Statistically non-significant results are not marked in the figure. See SI section for all details of statistical analysis. (C) Effect of the specific ROS inhibitor (NAC) on Ca2<sup>+</sup> dynamics. Data have been multiplied by a 10<sup>3</sup> factor, to make them visible over the same y-axis scale used in (A). (D–F) Representative Ca2<sup>+</sup> dynamics, Fluo-4 fluorescence relative increase and effect of NAC ROS inhibitor, in absence of extracellular Ca2<sup>+</sup> ions. In panel E, data are shown as mean ± SE, over statistical sample sets of n = 29, n = 60, n = 37 cells (for control, P3HT NPs and PS-NPs samples, respectively).

of the dispersion of 0.05, does not lead to a substantial increase of ROS production, thus representing a lower limit for obtaining sizable physiological effects. Conversely, as expected, an increasingly higher number of NPs leads to enhanced photogeneration of charges, thus determining an overall, sizable increase of ROS production. It is worth noting how the enhancement of ROS in dependence on NPs concentration is more pronounced in the case of the H2DCF-DA probe, being the latter sensitive to several species. Conversely, the response of APF and HPF probes presents a smoother

FIGURE 5 | Effect of P3HT NPs dispersion concentration on ROS generation by using different probes (A, H2DCF-DA; <sup>B</sup>, HPF; C,D, APF) and Ca2<sup>+</sup> modulation (E,F). Data report the mean values over populations, three experiments for each condition, of NH2DCF−DA\_Light\_OD0.4 = 66, NH2DCF−DA\_Dark\_OD0.4 = 40, NH2DCF−DA\_Light\_OD0.2 = 46, NH2DCF−DA\_Dark\_OD0.2 = 40, NH2DCF−DA\_Light\_OD0.05 = 48, NH2DCF−DA\_Dark\_OD0.05 = 30, NH2DCF−DA\_Light\_untreated = 42, NH2DCF−DA\_Dark\_untreated = 30, NHPF\_Light\_OD0.4 = 48, NHPF\_Dark\_OD0.4 = 46, NHPF\_Light\_OD0.2 = 48, NHPF\_Dark\_OD0.2 = 43, NHPF\_Light\_OD0.05 = 54, NHPF\_Dark\_OD0.05 = 52, NHPF\_Light\_untreated = 47, NHPF\_Dark\_untreated = 39, NAPF\_Light\_OD0.4 = 56, NAPF\_Dark\_OD0.4 = 56, NAPF\_Light\_OD0.2 = 69, NAPF\_Dark\_OD0.2 = 69, NAPF\_Light\_OD0.05 = 48, NAPF\_Dark\_OD0.05 = 48, NAPF\_Light\_untreated = 42, NAPF\_Dark\_untreated = 42, NFLUO−4\_OD0.4 = 43, NFLUO−4\_OD0.2 = 51, NFLUO−4\_OD0.05 = 60, NFLUO−4\_untreated = 56. (D) Shows the effect on ROS production of repeated light stimulation events, each one of 2 min overall duration, as evidenced by using the APF probe. Statistical analysis has been carried out by one-way ANOVA, followed by post-hoc Tukey test. P-values of the test are assigned as follows: \*\*\*for p < 0.001, \*\*for p < 0.005. Statistically non-significant results are marked as n.s. in the figures. See SI section for all details of statistical analysis.

FIGURE 6 | used for ROS production (CW LED light illumination, peak emission wavelength λ = 540 nm, closely matching the polymer absorption spectrum; photoexcitation density P = 95 mW/mm<sup>2</sup> ; overall duration of the photoexcitation protocol, 2 min), 24 h after NPs administration and upon careful rinsing of the cell growth medium (B, D, F, H, J). All other samples have been subjected exactly to the same preparation protocol, but they did not undergo photoexcitation (A, C, E, G, I). Viable cells are finally detected by FDA staining. Representative images have been acquired at 2 DIV. Scale bar, 100µm.

dependence on NPs concentration with respect to H2DCF-DA, probably due to their higher selectivity. Interestingly, **Figure 5D** shows the temporal dynamics of ROS evolution upon a photostimulation pattern (2 min photoexcitation followed by 10 s in dark, repeated five times), as detected by the APF probe. After three photoexcitation events, the ROS production reaches a saturation concentration. We conclude that the main effect is achieved within the first minutes of photostimulation. We carried out the same experiment using H2DCF-DA and HPF probes, but unfortunately, these probes undergo photo-oxidation and bleaching phenomena upon prolonged illumination, which makes them unreliable for this kind of study.

Ca2<sup>+</sup> dynamics as well are heavily affected, upon visible light excitation, by the concentration of P3HT NPs**.** Intracellular Ca2<sup>+</sup> dynamics and averaged maximum fluorescence variation at different NPs concentrations and two photoexcitation densities (2.1 and 14.6 mW/mm<sup>2</sup> ) are shown in **Figures 5E,F,** respectively. In line with results obtained for ROS production, the lowest employed NPs concentration, corresponding to an initial optical density of the dispersion of 0.05, does not lead to sizable effects. The Ca2<sup>+</sup> response upon photostimulation density of 2.1 mW/mm<sup>2</sup> is fully comparable to the physiological Ca2<sup>+</sup> concentration, as obtained in control samples. Conversely, effective NPs excitation, with a photoexcitation density of 14.6 mW/mm<sup>2</sup> , leads to a pronounced modulation of Ca2<sup>+</sup> dynamics. Overall, these data show that both ROS and Ca2<sup>+</sup> increase strongly depends on the concentration of P3HT NPs, and further demonstrate the key-role of polymer charge generation upon visible light.

### Photo-Toxicity Assays

We evaluated the cytophototoxicity effects eventually determined by optical excitation of polymer NPs, as compared with samples treated with NPs but not subjected to the illumination protocol, i.e., in which no ROS generation was optically activated. We considered five NPs doses, corresponding to OD in the range 0–0.5 (0, 0.05, 0.2, 0.4, and 0.5 OD values). Photoexcitation was carried out 24 h after cell plating/NPs administration, by using the same protocol employed for inducing ROS generation (see **Figure 5** and related methods). In all cases (i.e., for both illuminated and non-illuminated samples) cell viability was evaluated 48 h after plating and NPs administration. **Figure 6** displays representative images obtained by fluorescein diacetate (FDA) assay. FDA is a non-fluorescent molecule, which is hydrolyzed to fluorescent fluorescein in live cells only. We notice that cells not subjected to the illumination protocol show good viability at all considered NPs doses (**Figures 6A,C,E,G,I**). In the case of illuminated samples (**Figures 6B,D,F,H,J**), photoexcitation of P3HT NPs leads to substantial toxicity effects at the two highest NPs concentrations considered here (**Figures 5H,J**). Conversely, at P3HT NPs doses corresponding to 0–0.2 OD values no remarkable toxicity effect is detected. By combining these results with data reported in **Figure 5**, we identify the intermediate NPs concentration, corresponding to OD 0.2, as the most suitable one, leading to effective ROS production and reliable Ca2<sup>+</sup> ions modulation, but with no detrimental effects on cell viability, as evaluated 24 h after photoexcitation.

However, it is known that ROS can also lead to delayed toxicity effects, visible only after few days after production. Thus, we carried out viability tests at different time points, up to 72 h after NPs photoexcitation (at 4 DIV) and related ROS generation. **Figure 7** shows representative images of viable cells stained by FDA. The comparison among 4 different sample cohorts (NPs treated/untreated and illuminated/non-illuminated) does not show significant differences in the staining of viable cells. Most importantly, no remarkable signs of delayed photocytotoxicity effects are evidenced.

### DISCUSSION

In this work, we study the coupling mechanism at the abiotic/biotic interface between polymer NPs and living cells upon photoexcitation with visible light. Our NPs are good candidates for light-driven cell actuation. Indeed, functional effects on HEK-293 cells electrophysiology and in Hydra Vulgaris have been recently demonstrated (Tortiglione et al., 2017; Zangoli et al., 2017). Here, experimental data demonstrate that optical excitation of polymer NPs enhances the physiological ROS production and elicits intracellular Ca2<sup>+</sup> dynamics. We propose that this is occurring to a suitable extent to cause physiological effects and yet it is below the phototoxic regime.

First, we observe that upon photoexcitation with visible light P3HT NPs suspended into an aqueous electrolyte generate a photocurrent signal. We assign this to redox processes, accordingly to what occurs in P3HT films when used as the lightsensitive working electrode in photo-electrochemical cells. The same process is invoked to explain the observed generation of ROS in the cytosol compartment within living cells.

Several inorganic NPs, including metallic NPs (Wang et al., 2017), inorganic nanocrystals (Ipe et al., 2005; Wang et al., 2015) and magnetic NPs, have been reported to have sizable effects on the ROS population, thus representing an interesting strategy for photothermal therapies (Ye and Chen, 2016; Dayem et al., 2017). Among carbon-based materials, carbon nanotubes, graphene and polymer nano-bioconjugates can also sustain photo-thermal stimulation (Miyako et al., 2014; Chechetka et al., 2016; Lyu et al., 2016; Li et al., 2017). Yet, there is a fundamental difference to be taken into account when comparing those

OD, panel lines A–D respectively), and in presence or absence of polymer photoexcitation (± light, panel lines A–D, respectively), at fixed time points after plating and NPs administration. At 1 DIV, the same illumination protocol used for ROS generation was applied to samples treated by light (CW LED light illumination, peak emission wavelength λ = 540 nm, closely matching the polymer absorption spectrum; photoexcitation density P = 95 mW/mm<sup>2</sup> ; overall duration of the photoexcitation protocol, 2 min). Viable cells were then stained at fixed time points, namely 6, 24, 48, 72 h after the illumination protocol, corresponding to a maximum of 4 DIV (from left to right panels, respectively). Scale bar, 100µm.

systems with our P3HT NPs: for both inorganic- and organicbased systems reported so far, NIR-mediated photo-thermal heating is the main phenomenon, indirectly leading to ROS enhancement. Conversely, in our case, significant thermal effects can be ruled out. In fact, for a single NP upon light excitation density of I<sup>0</sup> = 100 mW/mm<sup>2</sup> the expected temperature increase is 1T = I <sup>0</sup> π R 2 NP / (4 <sup>π</sup> K r), where RNP <sup>=</sup> 240 nm is the average NP radius, K = 0.6 W/ mK is the thermal conductivity of water and r is the distance from the NP center. At the particle surface (r = RNP), the expected 1T is in the order of 10 mK, a value considered too low for provoking sizable effects. In order to directly test the effect of NPs-mediated thermal increase also on an experimental ground, we carried out electrophysiology experiments (whole cell patch clamp configuration) in HEK-293 cells stably transfected with temperature-sensitive channels (TRPV1), a system which intrinsically bears enhanced sensibility to temperature variations. Upon photoexcitation of P3HT NPs internalized within the cell cytosol we could not observe any significant effect (**Figure S9**) mediated by TRPV1 channel response. This confirms the numerical evaluation and rules out the thermal effect as a primary result of photoexcitation.

We conclude that the observed changes in ROS concentration and Ca2<sup>+</sup> dynamics are most plausibly caused by a direct photocatalytic effect. The dependence of the cell response on internalized NPs concentration and light excitation protocol further supports this hypothesis (**Figure 5**).

It is important to notice again that as soon as the delicate balance of ROS production is altered, or the intracellular levels of antioxidants is lowered, ROS can become highly harmful, causing oxidative stress which, in some cases, leads to irreversible cell damage (Martindale and Holbrook, 2002). Based on a detailed dose-response analysis and cell viability tests, we identify a suitable NPs concentration, corresponding to 0.2 OD value of the colloidal dispersion as an optimal compromise between reliable functional photomodulation of ROS generation and intracellular Ca2<sup>+</sup> ions increase, on one side, and occurrence of cytophototoxicity effects on the other side (**Figure 6**). In these conditions, and within the temporal window considered in the present work, up to 4 DIV (compatibly with the use of in vitro HEK-293 cell model), we do not observe sizable toxicity effects within treated cells, as documented by fluorescence imaging before and immediately after the photoexcitation protocol (Figure S10), as well as at different time points after photoexcitation, up to 72 h (**Figure 7**).

The proposed interaction mechanisms leading to ROS enhancement are depicted in **Figure 8**, based on the well-known P3HT photo-physical scenario. P3HT belongs to the class of polymeric organic semiconductors. It has a carbon conjugated backbone that supports π-electron delocalization. In solution, as isolated chains, such polymers behave like large organic molecules. The main photo-excited species are singlet and triplet states, the latter mostly produced by intersystem crossing from the initial singlet. Due to the extended conjugation, a fraction of the initial singlet states may dissociate into charge pairs. Those are very short-lived (<1 ps) in isolated chains, but in a solid, where inter-chain coupling may be non-negligible, they can separate on different chains and survive up to the milliseconds' timescale. In presence of oxygen (or any other dopants) this scenario dramatically changes. The singlet state quenches at the dopant, by energy or charge transfer. In the latter case, this generates fairly mobile holes and trapped electrons, a behavior often referred to as p-type. P3HT has a remarkably small triplet photogeneration yield (below 0.1%) (Jiang et al., 2002). This has important implications for biophotonics, since it substantially reduces the photo-toxicity of the material, cutting down singlet oxygen generation, the major killing path in photodynamic therapy. Accordingly, we think that optical excitation of the polymer NPs leads to singlet and polaron (charged) states. Singlets or negative polarons react with the oxygen dissolved in the aqueous dispersion (**Figure 2**) or in the cytosol environment, first reducing oxygen. The superoxide further evolves leading to a generation of different ROS (**Figure 3**). Because we use CW light, the steady state population of the singlet state is small (the lifetime is in the order of 1 ns). For this reason, the polaron-mediated path seems the most plausible one.

Polymer-mediated production of ROS deterministically triggers an increase in the intracellular Ca2<sup>+</sup> concentration, as demonstrated by the fact that upon ROS inhibition by a selective pharmacological agent (NAC), variations in Ca2<sup>+</sup> dynamics are completely suppressed (**Figure 4**). Moreover, we evidence that the main source of intracellular Ca2<sup>+</sup> increase resides within the cytosol, thus indicating a crucial role in the Ca2<sup>+</sup> release by mitochondria and/or the endoplasmic reticulum, being these organelles mostly involved in Ca2<sup>+</sup> dynamics. Further studies would be needed, however, to exactly identify the organelle involved in the photo-transduction effect triggered by illuminated P3HT NPs, and also to assess if the Ca2<sup>+</sup> pathway on the cell membrane is eventually changed.

Both Ca2<sup>+</sup> ions and ROS represent prominent signaling pathways able to finely tune several cell functions (Chakraborti et al., 1999). Importantly, dysfunctions of the delicate equilibrium in Ca <sup>2</sup>+/ROS balance have been reported to have serious

increase the SOD (superoxide dismutase) concentration. Increased Ca2<sup>+</sup> release from intracellular sources is also observed.

implications in various disorders, including, among others, neurodegenerative diseases such as Parkinson's and Alzheimer's disease, inflammatory diseases, metabolic diseases, and ischemic events (Chinopoulos and Adam-Vizi, 2006; Chaudhari et al., 2014; Sharma and Nehru, 2015; Zucchetti et al., 2016). In perspective, our results represent a novel, minimally invasive and locally confined tool for efficient modulation of ROS and intracellular Ca2<sup>+</sup> concentration.

### AUTHOR CONTRIBUTIONS

CB carried out all the experimental measurements, with help from IA (electrochemical characterization, ROS measurements, Ca2<sup>+</sup> Dynamics), GT (electrochemical characterization). EZ fabricated P3HT NPs, DD carried out TEM experiments. MZ and FD synthesized P3HT for the NPs. MA and GL planned and supervised the research. All authors contributed to manuscript drafting and approved the final version of the manuscript.

### REFERENCES


### FUNDING

The work was supported by a national grant from Fondazione Cariplo (project ON-IRIS 2013–0738).

### ACKNOWLEDGMENTS

The authors are grateful to Dr. F. Lodola for help in patch clamp electrophysiology measurements on HEK-293 cells transfected with TRPV1 channel, and to Dr. G. Barbarella from Mediteknology SRL for useful discussions and manuscript reading.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbioe. 2018.00114/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Bossio, Abdel Aziz, Tullii, Zucchetti, Debellis, Zangoli, Di Maria, Lanzani and Antognazza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Photogenerated Electrical Fields for Biomedical Applications

Giuseppina Polino† , Claudia Lubrano† , Giuseppe Ciccone† and Francesca Santoro\*

Center for Advanced Biomaterials for Healthcare, Istituto Italiano di Tecnologia, Naples, Italy

The application of electrical engineering principles to biology represents the main issue of bioelectronics, focusing on interfacing of electronics with biological systems. In particular, it includes many applications that take advantage of the peculiar optoelectronic and mechanical properties of organic or inorganic semiconductors, from sensing of biomolecules to functional substrates for cellular growth. Among these, technologies for interacting with bioelectrical signals in living systems exploiting the electrical field of biomedical devices have attracted considerable attention. In this review, we present an overview of principal applications of phototransduction for the stimulation of electrogenic and non-electrogenic cells focusing on photovoltaic-based platforms.

Edited by: Gianni Ciofani, Politecnico di Torino, Italy

### Reviewed by:

Ilaria Elena Palamà, Istituto di Nanotecnologia (NANOTEC), Italy Erhan Zor, Necmettin Erbakan University, Turkey

### \*Correspondence:

Francesca Santoro francesca.santoro@iit.it

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 23 September 2018 Accepted: 23 October 2018 Published: 09 November 2018

#### Citation:

Polino G, Lubrano C, Ciccone G and Santoro F (2018) Photogenerated Electrical Fields for Biomedical Applications. Front. Bioeng. Biotechnol. 6:167. doi: 10.3389/fbioe.2018.00167 Keywords: bioelectronics, photovoltaics, tissue engineering, biointerfaces, electrical stimulation

## INTRODUCTION

Bioelectronics devices are meant to probe and stimulate biological entities through electricity by adopting smart biocompatible materials which are also conductive (Liao et al., 2015). In particular, traditional inorganic conductors, semiconductors and, more recently introduced, conjugated polymers had find major application in the development of bioelectronic platforms (Zhang and Lieber, 2016).

Electrical fields can be generated either intrinsically or upon transduction as, for instance, in photovoltaic materials, where light can induce the generation of relevant currents through the bulk of a photoconductive material.The materials in which this phenomenon happens can be inorganic, fully organic, or hybrid mixtures.

In particular, electrical fields can be exploited to locally stimulate cells and tissue inducing responses of different nature, for instance, altering the electrophysiological activity of electrogenic cells or modulate certain cellular processes and functionalities, i.e., polarity, proliferation, differentiation, in non-electrogenic cells (Blau, 2013; Pennacchio et al., 2018).

For example, silicon-based devices are widely employed for electrical interfaces both at the micro and nanometer scale with neural tissues (Thukral et al., 2018) and, furthermore, have gained major interest in the niche of photovoltaic-based platforms for retinal implants (Di Maria et al., 2018). However, these materials exhibit some limitations in terms of flexibility and stiffness in general, which make the cell-device coupling still not optimal.

In the last decade, new generation of hybrid or fully organic, highly biocompatible and functionally self-powered prostheses have been developed to treat, among various applications, blindness proposing a cutting-edge paradigm to interface phototransductive materials with biological cells (Khraiche et al., 2013; Maya-Vetencourt et al., 2017; Benfenati and Lanzani, 2018; Ferlauto et al., 2018). In addition, photovoltaic platforms have been recently proposed to directly interface non-electrogenic cells, such as fibroblasts, to trigger their proliferation and open up the possibility to use such materials as optimal candidate for wound healing purposes (Jin et al., 2013). Here, we present an overview of the main applications photovoltaic-based platforms have found in the recent years to mimic biological component with similar basic functionalities, i.e., retina, and in addition we aim to highlight the important potential of those materials have for cell stimulation in general.

### ELECTRICAL PHOTOTRANSDUCTION IN DEVICES FOR BIOMEDICAL APPLICATIONS

Electrical phototransduction is the process through which photons are converted into electrical signals. For instance, inorganic semiconductors are exploited for their capacity to convert light into electrons flow by means of their P-N junction (Richter et al., 2018). However, their rigid nature, together with their poor biocompatibility has generated a major attention toward their organic counterparts for biomedical applications. Organic photovoltaic devices are able to deliver photocurrents thanks to the generation of electrical fields when light is absorbed in their photoactive layers, composed of a donor, and an acceptor semiconducting material. The donor material donates electrons transporting holes while the acceptor material withdraws electrons and further transports them (Polino et al., 2015).

In both inorganic and organic cases, charges are finally transported to an electrode, which can be in contact with a biological cell or a tissue. Starting from the carrier transport at the electrode, an electrical field is generated between the device and the biological matter and this bioelectronic coupling can modulate biological processes at different matter of scale and cellular architectures (Pennacchio et al., 2018).

The effects of this interface on cells can be investigated by observing capacitive, chemical and thermal mechanisms involving the cell membrane (Di Maria et al., 2018). In particular, platforms based on electrical photo-transduction have aroused the interest of many research groups with the purposes of recording and stimulating single cells, cell networks or tissues (Sim et al., 2014; Chenais et al., 2017; Jeong et al., 2017).

For example, in the work of Ghezzi et al. (2011), first attempts have successfully shown that stimulation of primary neurons via light absorption in a P3HT-based biointerface is possible. In particular, the bioelectronic interface consisted in a very thin film (≈150 nm) of a P3HT:PCBM blend deposited on an ITO-coated glass substrate. One step further was accomplished by the same group using a platform for neuronal stimulation which consisted of an active layer only in the n-type conjugated polymer P3HT (Ghezzi et al., 2013). This is just an example of how photo-transduction can take place in 2D materials and how this phenomenon can be exploited for generating electrical fields to trigger certain processes at the membrane of cells. However, it is interesting to note that even bioelectronic platforms are moving fast toward more biomimetic approaches inspired by those of tissue engineering (Pennacchio et al., 2018).

In this context, first attempts to create 3D photovoltaic platforms for skin regeneration have been carried and the local photo-transduction mechanism of 3D meshes is still under discussion (Jin et al., 2014). Here, we explore the use of light photo-trasduction effect on single cells particularly focusing on tuning cellular behavior and act as cell-instructive platforms for various applications such as platforms for immunorecognition, or as activation/inhibition of the activity of electrogenic and non-electrogenic cells.

## ELECTROGENIC CELLS

Platforms based on transduction of photons in to current/voltage generation have found major applications in the modulation of the electrical activity in electrogenic cells. In fact, these cells, such as cardiomyocytes or neuronal cells, are capable of changing their membrane potential upon stimuli generating fast-changing events (action potentials). Thus, using certain electrical fields can be a suitable approach to trigger action potential inhibition or activation (Love et al., 2018).

Moreover, photovoltaic platforms has found a niche in the design of implantable devices for restoring lost functionalities in the retina (Benfenati and Lanzani, 2018; Di Maria et al., 2018).

Numata and co-workers reported the first example of photoinhibition of ion transport in PC-12 cells by using chargeseparation molecules ferrocene (Fc)–porphyrin (P)–C60 linked triads. This compound was delivered close to the plasma membrane using drug carriers and, after light stimulation a depolarization of the membrane potential and an inhibition of potassium channels was observed. This result suggests thatmore sophisticated molecules can lead to the control of firingneuronal cells (Numata et al., 2012).

Abdullaeva and co-workers, with the intent to create organicbased artificial photoreceptors, developed a photoactive layer consisting of an anilino-squaraine donor blended with a fullerene acceptor as support for N2A cells (neuronal model cell line) growth. They supposed that during the pulse stimulation there is an accumulation of negative charge carriers at the photoconductor–electrolyte interface (**Figure 1A**), which results into cell depolarization. When the light is turned off instead, a rapid hyperpolarization of the cell membrane was detected (Abdullaeva et al., 2016). Similarly, induced photocapacitance can be exploited for modulation of the membrane potential in cells.

In fact, Martino and co-workers observed a similar effect on human embryonic kidney (HEK-293) cells growth on a conjugated polymer poly(3-hexylthiophene) (P3HT). The local heating of the material produced an increase in the ion transport through membrane channels, causing a decrease of the membrane resistance (Martino et al., 2015). Recently, Glowacki and co-workers have proposed an approach for neural photostimulation employing an electrolytic photocapacitor (**Figure 1B**) built with a trilayer of metal and p–n semiconducting organic nanocrystals (Rand et al., 2018).

Moving toward the coupling of photovoltaic platforms with tissues, retina implants have found major interest in the last years (Benfenati and Lanzani, 2018). Photovoltaic devices which are foldable and flexible have been developed for wide-field epiretinal

prosthesis. These devices are capable of stimulating wireless retinal ganglion cells. The material stack included poly(3,4 ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) as anode and Poly(3-hexylthiophene) (P3HT) blend based as photoactive layer (**Figure 1C**) (Maya-Vetencourt et al., 2017; Ferlauto et al., 2018).

Retinal and subretinal inorganic prostheses were developed to restore sight in patients blinded by retinal degeneration by stimulating the inner retinal neurons using 3D pillar electrodes (**Figure 1D**) which enhance the cell-chip coupling and the integration the implant in the target tissue (Mathieson et al., 2012; Flores et al., 2018). In a recent work done by Ho et al. (2018a) a photodiode array made of boron-doped silicon-oninsulator (SOI) wafers was used to perform in vivo and in vitro measurements, in which the retina is placed between the recording array (ganglion cell side) and the photodiode array (photoreceptor side). In a recent work, transparent extracellular microelectrode arrays (MEA) were used to characterize the spatial and temporal response properties of retinal ganglion cells (RGCs) to photovoltaic stimulation in the healthy and degenerate rat retina (Ho et al., 2018b). It was demonstrated that using silicon photodiodes to build photovoltaic pixels arrays and it is possible to convert signals into patterns of current to stimulate the inner retinal neurons for wireless neural stimulation in translucent tissues (Boinagrov et al., 2016).

### NON-ELECTROGENIC CELLS

Beside electrogenic cells and tissues, photoelectrical devices can be used also for promoting cellular proliferation, for instance, in tissue regeneration applications.

One successful approach in tissue engineering is the use of photovoltaic platforms as novel treatments for angiogenesis. In this approach, a solar cell generates an electrical field promoting the formation of capillaries and arterioles at the ischemic region, attenuating muscle necrosis and fibrosis and promoting the secretion of angiogenic growth factors and the migration of mesenchymal stem cells (MSCs), myoblasts, endothelial progenitor cells, and endothelial cells in in vitro experiments (**Figure 2A**) (Jeong et al., 2017).

Many efforts are carried out on different types of cells like adipose-derived stem cells (ASC) and endothelial cells, evaluating the influence of pulsed LED light of three different wavelengths using continuous LLLT stimulation (Rohringer et al., 2017). In Couto et al., it was shown that a treatment of skin lesions with coherent and incoherent light sources (laser and LED, respectively) enables the wound healing process in elderly rats, with good results in terms of collagen deposition, fibroblasts proliferation and inflammatory cellular response (Couto et al., 2017). In fact, other studies have been developed on fibroblasts treatment using laser or led therapy to inhibit keloid fibroblasts after irradiation (Couto et al., 2017; Lee et al., 2017). In the end, different studies demonstrated the importance of therapy with coherent light in the visible spectrum to optimize the process of tissue repair, and some studies suggests that comparable effectiveness could be reached with the lower costs non-coherent light in the same spectral region.

Another interesting approach involved the fabrication of nanoweb substrates made of poly(3-hexylthiophene) (P3HT) to enhance the neurogenesis of human fetal neural stem cells (hfNSCs) and control their behavior via optoelectrical stimulation (**Figure 2B**) (Yang et al., 2017). Low-level laser therapy (LLLT) based on low-level laser or lightemitting diodes (LEDs) was used for tissue regeneration in mouse neural stem cell on 3D printed scaffolds (Zhu et al., 2017).

An interesting novelty regards the possibility to treat diseases delivering cellsto an injured or diseased organ/tissue. Usually this is accomplished by collecting cells through a digestion process with enzymes which can lead to Undesired enzymatic residuces. In orer to overcome this, cells have been cultured onsilicon based photovoltaic surfaces have been released through light stimulation as presented in the work of Bhuyan et al. (2016).

In a recent work done by Diring et al. (2017), photoactive zirconium-based metal organic framework (MOF) particles embedded in a polymer matrix were employed as growth substrate for HeLa cells. The substrate was able to release CO upon light irradiation, and its subsequent cellular uptake was monitored using a fluorescent probe.

In this context, one example, which also attempted to recapitulate 3D tissue-like environment with particular focus on regenerative medicine, was presented by Jin and co-workers. They reported on the photocurrent stimulation effect on cells with particular focus on understanding the combined effect of direct light interaction together with the responseto the induced electromagnetic field. In fact, porphyrins are activated by photons interactions while calcium ions translocate through the voltagegated channels at the membrane due to the electromagnetic field. Overall, this results in an increase of cytosolic Ca2<sup>+</sup> which might trigger protein pathways responsible for cell proliferation (Jin et al., 2011) (**Figure 2C**). Furthermore, they showed how to combine 3D structures with photoactive materials in order to achieve skin regeneration. In particular, they used human dermal fibroblasts (HDFs) and prepared a photosensitive nanofibrous scaffolds by electrospinning using Poly(3-hexylthiophene) (P3HT) and Polycaprolactone (PCL) as base materials to fabricate the nanofibers. Here, they further discussed how the cytosolic Ca2+ increase would led to the activation of the low molecular weight protein calmodulin, which causes the activation of several key intracellular processes leading to cell division (Tomlinson et al., 1984) in combination with direct light activation of protein kinase C. As a result, these two effects led to an increase in cell proliferation (Jin et al., 2013, 2014) (**Figure 2D**). Furthermore, Jin and co-workers designed a new photosensitive semiconductive polymer, PDBTT (poly (N,N-bis

### REFERENCES


(2-octyldodecyl)-3,6-di(thiophen-2-yl)-2,5-dihydropyrrolo[3,4 c]pyrrole- 1,4-dione-alt-thieno[3,2-b]thiophene), with maximum absorbance at 600 nm. This polymer with nanofibers of PCL incorporated, was electrospun and tested to study the proliferative effect of PCL/PDBTT nanofibers on HDFs under red LED illumination (Jin et al., 2017). Another interesting study showed how a scaffold made of electrospun fiber of P3HT in combination with a typical polymer used for tissue engineering, poly(L-lactic acid)-co-poly-(e-caprolactone) could be successfully used as photosensitive platforms for enhancing fibroblast proliferation and thus as good precursor of a wound dressing solution (Jin et al., 2014).

### CONCLUSION

In this review, we reported the last advances in electrical phototransduction applied to biomedical applications. In particular, we focused the attention on the possible effects of photoelectrical stimulation driven on cells and tissues. The aim was to report the behavior of the biological system at the biotic-abiotic interface considering various approaches for enhancement/inhibition of electrogenic and non-electrogenic cells functionalities. In particular, we highlighted how the use of photovoltaic platforms had found major application in designing retinal prosthetics. Moreover, organic photovoltaic materials are used for building new bio-compliant platforms and, in perspective, represent the foremost class of material for cell-chip coupling. We foresee an increasing application of these materials also for tissue engineering purposes as preliminarily proposed for wound healing. Furthermore, the few examples of photovoltaic-based scaffolds paved the way to advanced platforms which need to be developed toward more biomimetic 3D environments.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

models and performance. IEEE Trans. Biomed. Circuits Syst. 10, 85–97. doi: 10.1109/TBCAS.2014.2376528


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Polino, Lubrano, Ciccone and Santoro. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Targeting Inflammation With Nanosized Drug Delivery Platforms in Cardiovascular Diseases: Immune Cell Modulation in Atherosclerosis

Antonio Cervadoro<sup>1</sup> , Roberto Palomba<sup>2</sup> , Giuseppe Vergaro3,4, Roberta Cecchi 1,5 , Luca Menichetti <sup>6</sup> , Paolo Decuzzi <sup>2</sup> , Michele Emdin3,4 and Stefano Luin1,7 \*

<sup>1</sup> NEST Laboratory, Scuola Normale Superiore, Pisa, Italy, <sup>2</sup> Laboratory of Nanotechnology for Precision Medicine, Fondazione Istituto Italiano di Tecnologia, Genova, Italy, <sup>3</sup> Division of Cardiology and Cardiovascular Medicine, Fondazione Toscana Gabriele Monasterio, Pisa, Italy, <sup>4</sup> Institute of Life Sciences, Scuola Superiore Sant'Anna, Pisa, Italy, <sup>5</sup> Center for Nanotechnology Innovation (CNI@NEST), Istituto Italiano di Tecnologia, Pisa, Italy, <sup>6</sup> CNR Institute of Clinical Physiology (IFC), Pisa, Italy, <sup>7</sup> NEST Laboratory, Istituto Nanoscienze, CNR, Pisa, Italy

Edited by: Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Ciro Chiappini, King's College London, United Kingdom Vincenzo Calcagno, Universität Leipzig, Germany Stefano Leporatti, Istituto di Nanotecnologia (NANOTEC), Italy

> \*Correspondence: Stefano Luin s.luin@sns.it

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 14 October 2018 Accepted: 06 November 2018 Published: 27 November 2018

#### Citation:

Cervadoro A, Palomba R, Vergaro G, Cecchi R, Menichetti L, Decuzzi P, Emdin M and Luin S (2018) Targeting Inflammation With Nanosized Drug Delivery Platforms in Cardiovascular Diseases: Immune Cell Modulation in Atherosclerosis. Front. Bioeng. Biotechnol. 6:177. doi: 10.3389/fbioe.2018.00177 Atherosclerosis (AS) is a disorder of large and medium-sized arteries; it consists in the formation of lipid-rich plaques in the intima and inner media, whose pathophysiology is mostly driven by inflammation. Currently available interventions and therapies for treating atherosclerosis are not always completely effective; side effects associated with treatments, mainly caused by immunodepression for anti-inflammatory molecules, limit the systemic administration of these and other drugs. Given the high degree of freedom in the design of nanoconstructs, in the last decades researchers have put high effort in the development of nanoparticles (NPs) formulations specifically designed for either drug delivery, visualization of atherosclerotic plaques, or possibly the combination of both these and other functionalities. Here we will present the state of the art of these subjects, the knowledge of which is necessary to rationally address the use of NPs for prevention, diagnosis, and/or treatment of AS. We will analyse the work that has been done on: (a) understanding the role of the immune system and inflammation in cardiovascular diseases, (b) the pathological and biochemical principles in atherosclerotic plaque formation, (c) the latest advances in the use of NPs for the recognition and treatment of cardiovascular diseases, (d) the cellular and animal models useful to study the interactions of NPs with the immune system cells.

Keywords: atherosclerosis, inflammatory diseases, smart nanomaterials, drug delivery, nanomedicine, imaging and theranostics, immune cells, cardiovascular diseases

### ATHEROSCLEROSIS AND INFLAMMATION

Atherosclerotic disease, or simply atherosclerosis (AS), is initially characterized by the formation of fatty streaks, with the accumulation of lipids [primarily cholesterol, but also triglycerides (Goldberg, 2018)] in the intima and inner media of arterial wall, especially in regions with abnormal flow patterns (Chistiakov et al., 2017). Fatty streaks may then evolve into soft, lipid-rich plaques, and eventually into thick cap-calcified lesions and/or unstable plaques characterized by inflammatory infiltration and sustained oxidative processes. Stable fibrocalcific atheroma is characterized by calcium deposits, small lipid deposits, slight lumen reduction, and, often, by poor functional relevance. Vulnerable atheroma, which is more prone to rupture, is characterized by a large lipid-rich necrotic core, thin fibrous cap (<65µm), neovascularization, spotty calcifications, inflammatory cells, and positive remodeling (Moore and Tabas, 2011; Hansson et al., 2015). The possibility of plaque regression has been reviewed in Chistiakov et al. (2017) focusing on animal models, and in Fisher (2016) also considering the dynamic changes in lipid and immune cells distributions.

AS is currently considered an inflammatory disorder, characterized by the infiltration into sub-endothelial space of various immune cells (ICs), especially circulating monocytes (MCs) that subsequently differentiate into macrophages (M8s) and then into foam cells (FCs), going along with plaque formation and evolution (Moore et al., 2013; Zhang et al., 2017). Such simplified picture is complicated by the heterogeneity of the cells within and close to lesions during evolution [MCs, M8s, and FCs, but also neutrophils, dendritic cells (Butcher and Galkina, 2012), T cells (Taleb, 2016), and possibly others]. In the cited reviews particular attention is given to the markers of the various IC phenotypes, although cells can present more than one function and can express a continuum of markers of different subtypes (Butcher and Galkina, 2012).

A key example regards M8s, the characterizing cells in AS: their different subtypes can have antithetic roles. Initially classified only as classically activated M1 (pro-inflammatory) or alternatively activated M2 (anti-inflammatory), evidences have brought to the definitions of additional subtypes (e.g., Mox, M4, Mhem) and even "sub-subtypes" (e.g., M2a, M2b, M2c); there could actually be a continuum of specializations, and M8s could even convert into each other (Butcher and Galkina, 2012; Leitinger and Schulman, 2013). It must be noted that also smooth muscle cells (SMCs) can assume some of the functions usually assigned to M8s (efferocytosis, internalization of lipids or cholesterols), and may transform in foam cells (Chistiakov et al., 2017).

In any case, cells with phenotypes closer to Mox or M1 cells are the most active in internalizing lipids (particularly cholesterol), especially if agglomerated within oxidized low density lipoprotein (oxLDL). Upon this process, efferocytosis efficiency of cells is reduced, they transform into foam cells, and are easily subjected to imbalances in cholesterol influx/efflux process and hypoxia. The final outcome is most often apoptosis or necrosis, which resolve in local accumulation of the lipid content, the major constituent of the inner (possibly necrotic) core of the atherosclerotic plaque (Moore and Tabas, 2011). At the same time, the cell's residual components promote further inflammation signals and generation of oxidative species, in a persistent cycle of recruitment of intimal macrophages and their polarization toward pro-inflammatory subsets (Libby et al., 2014). Moreover, both monocytes and macrophages have been shown to contribute to the increase of the gamma-glutamyl transferase enzyme (γGT) in AS lesions, increasing the oxidative character of these zones (Pucci et al., 2014; Belcastro et al., 2015).

This self-sustained state of inflammation (similar to what happens in chronic wound environments) is characteristic of the vulnerable plaque. The presence of activated macrophages leads to the secretion of matrix-components degrading enzymes (such as matrix metalloproteinases-MMPs), inhibits the production of collagen by the SMCs, and finally contributes to vascular wall reshaping. These mechanisms are useful in early lesions, supporting resolution of disrupted endothelial layers and avoiding accumulation of toxic species, as well as for the reduction of atherosclerotic plaques; but, at the same time, they induce fragility in late plaques, with the possible final consequence of plaque rupture (Libby et al., 2014; Hansson et al., 2015; Martinez and White, 2018). Inflammation also modulates the clinical consequences of the thrombotic complications of AS, while its inhibition could attenuate progression, mitigate the risk of plaque rupture, and even promote regression of AS (Awan and Genest, 2015; Bäck and Hansson, 2015; Bäck et al., 2015; Kamaly et al., 2016).

### ATHEROSCLEROSIS TREATMENTS

Current therapeutic approaches to AS aim at reducing promoting factors, including hypertension, smoke, and dyslipidemias; considered drugs (e.g., inhibitors of cholesterol hepatic synthesis like statins) particularly affect production and transport of cholesterol (or other lipids) to the arterial walls (Coomes et al., 2011; Duivenvoorden et al., 2014; Tsujita et al., 2015). Different approaches have been investigated, such as blocking absorption of cholesterol in the intestine (e.g., by ezetamibe) or controlling its reverse transport in ICs, e.g., by using an agonist for the liver X receptor (LXR) like GW3965, or administering artificial forms of high density lipoproteins (HDL; Duivenvoorden et al., 2014; Tsujita et al., 2015; Pulakazhi Venu et al., 2017; Goldberg, 2018; Mueller et al., 2018).

Controlling inflammation is another promising strategy; in particular, the interleukin-1 pathway has been identified as a possible therapeutic target. Canakinumab, a monoclonal antibody targeting interleukin-1β, was tested in the CANTOS (Canakinumab ANti-inflammatory Thrombosis Outcomes Study) trial (Ridker et al., 2011, 2017). Moreover, the chemotherapeutic drug methotrexate (MTX), largely used as an immunomodulatory and anti-inflammatory (AI) drug, was described to lower the risk for total cardiovascular diseases in patients with chronic inflammation (Popkova et al., 2015; Gomes et al., 2018), and has been observed to inhibit atherogenesis and macrophage migration to the intima in animal models (Bulgarelli et al., 2012). Although its mechanisms of action are not fully clear, this molecule is capable of reducing the secretion of pro-inflammatory cytokines and the expression of adhesion molecules on both immune and endothelial cells (Coomes et al., 2011).

While promising, systemic administration of AI drugs is often limited by a narrow therapeutic index and by severe adverse effects, including bone marrow suppression, neutropenia and immunodepression. Blockade and stimulation of the mechanisms involved in inflammation can hold positive, null, or even detrimental results depending on the phase of the atherogenic process (Aluganti Narasimhulu et al., 2016). Based on these premises, it is authors' belief that a rational modulation of the functions of ICs is promising for an efficient treatment and hopefully comes with fewer adverse effects.

Ideally, it would be desirable to identify an effective strategy for all of the AS phases; e.g., in (Libby et al., 2014) Resolvin E1 is indicated as a mediator that reverses all the advanced lesions associated processes, while contributing to the resolution of the plaque also in earlier phases. Another promising possibility is controlling release or activation of compounds used to alleviate or cure AS symptoms/consequences. Possible approaches include: the use of drugs modified with a glutamyl in order to exploit the increased concentration of γGT in AS lesions (Belcastro et al., 2015); developing methods exploiting high cholesterol concentrations; tackling pathways leading the transformation process from monocyte to pro-inflammatory M8s and finally to foam cells (Rousselle et al., 2013).

Due to their ambivalent role of inducing inflammation and regulating tissue regeneration, macrophages have been the first candidates addressed in order to control AS. Vannella and Wynn presented an interesting review describing macrophages behavior in different tissues, the mediators of the various mechanisms involved, and finally several possible ways to modulate them (Vannella and Wynn, 2017). It would be ideal to force their egress from AS lesions as transformation into foam cells is occurring; to this end, a suggestive idea could be activating or releasing a "macrophage migration-enhancement factor" (Weisbart et al., 1974; Ueno et al., 1997; Nunami et al., 1998), or interfering with cell surface adhesion molecule only in overabundancy of cholesterol or other AS markers, e.g. by stopping the phosphorylation of CD44 (Qin, 2012).

### TOWARD TARGETED THERANOSTICS NPs

As hinted above, spatial and temporal control of drug activity can reduce collateral short and long term effects; moreover, it can result in more convenient administration methods. These premises strongly call for the development of efficient, targeted delivery strategies for AI molecules, just like the ones based on their encapsulation into NPs (Costa Lima and Reis, 2015). Several kinds of NP have been considered for drug delivery; characteristics and production protocols for some of these are reviewed in Allen et al. (2016), Ulbrich et al. (2016), Cheraghi et al. (2017), and Matoba et al. (2017). There are still limitations and drawbacks for the clinical use of many NPs; these could arise from a not-yet perfect control of the final fate of many formulations, since they often accumulate also in the organs of the reticuloendothelial system (RES), from polydispersion and/or poor reproducibility in their preparation, or from the often difficult scale-up and high cost for their production, especially when multifunctional capabilities are added (Cheng et al., 2012; Ulbrich et al., 2016). However, NP physicochemical properties can be finely tailored during their synthesis and this allows optimizing drug loading and NP target specificity (Allen et al., 2016; Pentecost et al., 2016).

In the clinical setting of atherosclerotic disease, NPs loaded with anti-inflammatory drugs can be a powerful tool to hit inflammatory targets at the plaque level, preventing systemic side effects (Jokerst and Gambhir, 2011; Di Mascolo et al., 2013). Drug-loaded NPs targeting macrophages and other immune cells could control their pro-inflammatory activities and thus prevent, attenuate, and possibly reverse related disorders (i.e., increased atherosclerosis, but also altered adipocyte function and insulin resistance). At the same time, NPs could also be loaded with imaging agents allowing the detection of vulnerable atherosclerotic plaques; similar theranostic strategies already showed potential for detection and treatment of other diseases (including cancer and neurodegenerative disorders), exploiting a number of imaging modalities, among which optical imaging (OI), magnetic resonance imaging (MRI), ultrasound and photoacoustic (US-PA), computed tomography (CT), and nuclear imaging based on single photon and positron emission tomography (SPECT, PET; Xie et al., 2010; Kim et al., 2014; Weissleder et al., 2014; Atukorale et al., 2017; Stigliano et al., 2017; Zhang et al., 2017).

There is evidence that NPs can segregate into the plaque in preclinical models of AS via a "passive" targeting mechanism (Weissleder et al., 2014); while such a relative selectivity has not been completely understood, some hypotheses have been proposed. First, the abnormal hemodynamic forces where plaques develop may favor NPs deposition (Hossain et al., 2015); second, the endothelium appears discontinuous with openings for sufficiently small objects (Kim et al., 2014); third, there can be an active role of immune cells in vehiculating or accumulating the nanoparticles in inflamed districts (Moore et al., 2017).

Other strategies focused on specific targeting, e.g., toward inflammatory factors, dysfunctional endothelial cells, or specific macrophage receptors involved in cholesterol transport. Examples are drug carriers functionalization with selectin ligands or antibodies directed to CAMs (sialyl-Lewis X, PSFL-1, ICAM and VICAM ligands, anti-ICAM, anti-VCAM; Robbins et al., 2010) or anti-oxLDL receptor (Li et al., 2010). Li et al. used liposomes decorated with LOX-1 antibodies, Indium (111In) or Gadolinium (Gd), and DiI fluorescence markers to image atherosclerotic plaques in ApoE−/<sup>−</sup> mice. Alternatively, it has been proposed to target collagen IV, which is present on the vascular basement and exposed when vascular permeability increases (Chan et al., 2011; Chen et al., 2013; Kamaly et al., 2016; Meyers et al., 2017). Also common strategies to target macrophages are based on functionalizing nanoparticles with dextran sulfate coating or peptides mimicking low-density lipoproteins (LDL) such as apolipoprotein A1 (ApoA-1), whose receptors (Class A Scavenger Receptor 1 MSR-1, and class B Scavenge Receptor CD36) are expressed on macrophages cell membrane (Canton et al., 2013). Further works focusing on visualization of macrophages distribution using nanoparticles are well reviewed in Weissleder et al. (2014).

Active targeting was achieved also using a "biomimetic" approach: NPs can derive from, or have properties similar to, aggregates like LDL or HDL, naturally accumulating in AS plaques (Allijn et al., 2013; Duivenvoorden et al., 2014; Gomes et al., 2018). More specific tissue or cell targeting can be found testing nanoparticle libraries in-vitro or in-vivo (Kamaly et al., 2016; Tang et al., 2016), but the mechanisms of the found specificity should be understood, also in order to ensure it is preserved downhill of modifications of nanoconstructs or in different biological environments. Libraries are often tested preliminarily in cell cultures, but in-vivo tests are necessary at least to evaluate the impact of the different biological media.

Indeed, upon entry in an organism, NPs are usually coated by a protein corona (PC), changing their biological identity. PC formation (or "opsonisation") is often the first step toward the sequestration of NPs by the RES. Various approaches have been considered to avoid such phenomenon. Recent strategies are based on controlling nanoconstruct stiffness, since it was found that deformable particles are less subject to uptake by macrophages in RES in off-target tissues (Key et al., 2015; Palomba et al., 2018). Other methods exploit different coatings and functionalization developed for controlling the PC, e.g., by using polymers like PEG. This is thought to be an antifouling agent, but it actually seems to modulate the PC (Schöttler et al., 2016); moreover, it has been shown to be immunogenic, requiring the use of additional functionalizing moieties (Mima et al., 2017). Other possible coating molecules are based on peptides: examples comprise zwitterionic ones, for limiting serum-protein adsorption (Ranalli et al., 2017), or aptamer-likes, for enriching the PC with specific molecules present in biological fluids, which act as targeting moieties when properly oriented (Santi et al., 2017). All these approaches easily grant nanoconstructs with extended circulation half-life.

An advantage that will play important role consists in NPs potential multimodality: not only they can contain more than one drug and/or imaging agent, but it could also be possible to implement a trigger for drug release/activation. This could either be intrinsic (provided by the pathological environment) or exogenous (ultrasound, light, oscillating magnetic fields).

Especially under this view, an interesting development can arise from a synergy between nanotechnology and personalized medicine (Mendes et al., 2018). Early screening of the most suited bioactives (Tang et al., 2016; Risum et al., 2017), real time monitoring of local accumulations (Zavaleta et al., 2018), observing early feedbacks to the treatment (Qiao et al., 2017), and predicting patient responses (Sykes et al., 2016) are all possible aspects of personalized medicine. We believe that the rational application of nanotools in all these steps will embody a fundamental role in the upcoming clinical and pre-clinical research (Mura and Couvreur, 2012; Polyak and Ross, 2017; Yordanova et al., 2017).

### NPs FOR AS TREATMENT AND DIAGNOSIS

The use of NPs for treatment of AS and visualization of plaques up to 2015 has been elegantly reviewed in Zhang et al. (2017); we further selected more recent works not reported there (**Table 1**). In this session, we describe some of these and other relevant works.

Often, intrinsic properties of NPs lead to the exploitation of drug delivery strategies together with photothermal- and radiofrequency-mediated triggering effects; e.g., Johnston proposed the possibility to use photothermal destruction of macrophages using iron oxide NPs with thin gold and dextran coating, excited by a laser pulse at 755 nm. In this manuscript the particles were used for MRI and in vitro photothermolysis (Ma et al., 2009). A similar approach has been tested in the NANOM-FIM trial (Kharlamov et al., 2015); here, NPs composed by silica shells containing gold and eventually magnetic nanoparticles were delivered on AS plaques by a bioengineered on-artery patch or using a magnetic navigation system; detonation of NPs using a NIR laser caused a significant final reduction of the total atheroma volume.

A different, interesting strategy consists on preventively acting on the selective recruitment of monocytes from the precursors of pro-inflammatory macrophages M1 (Nakashiro et al., 2016; Matoba et al., 2017). The authors proposed polymeric NPs loaded with Pioglitazone, an agonist of the receptor PPARγ shown to be able to influence macrophage polarization. The formulation was tested in ApoE−/<sup>−</sup> mice fed a high fat diet (HFD) and infused with angiotensin II, promoting inflammation driven by monocytes/macrophages. 2 days post injection the ratio between peripheral pro- and non-inflammatory monocytes subsets decreased substantially, and tissue macrophage polarity was regulated toward the noninflammatory phenotype M2, with consequent suppression of EMMPRIN/MMP pathway and reduction of plaque destabilization risk.

In a similar approach, Stigliano et al. (2017) confirmed the potentiality of MTX in preventive-oriented treatments by loading it into NPs. The authors demonstrated specific accumulation of NPs into macrophages residing within lipid-rich plaques along the arterial tree. In the aortic arch of ApoE−/<sup>−</sup> mice on HFD treated with MTX-NPs, 50% less coverage of plaques was found in comparison to the control group. Importantly, this result was obtained by injecting a dose four times lower than reported in literature (Leite et al., 2015; Gomes et al., 2018).

MTX has also been investigated in synergy with different bioactive compounds. The group of Serrano Jr. (Gomes et al., 2018) tested the combined effect of injecting Paclitaxel (PTX) loaded LDL-mimicking NPs and MTX-loaded NPs. In New Zealand white rabbits on atherogenic diet treated with both the NPs formulations, the regression of the lesion area and the intimal width reduction corresponded to 17 and 63%, respectively, compared to the group treated with PTX-NPs alone. The authors speculate that this athero-regression is mostly due to the macrophage reduction effect and not to an inhibition of SMCs migration into the intima.

Other possible applications of nanomedicine in the treatment of AS are based on gene regulation. Majmudar et al injected siRNA-containing polymeric NPs to silence the expression of C–C chemokine receptor type 2 (CCR-2), a key player in recruiting inflamed monocytes to atherosclerotic plaques; they observed reduced PET signals from <sup>89</sup>Zr-labeled dextran nanoparticles in aortic root when compared to mice treated with an irrelevant siRNA (Majmudar et al., 2013). Another example is the downregulation of the tissue inhibitor of metalloproteinase 3 by the use of miR-712, delivered by cationic lipid NPs targeting VCAM1. The treatment was performed in ApoE−/<sup>−</sup>


Frontiers in Bioengineering and Biotechnology | www.frontiersin.org

diet; ApoE, apolipoprotein

 E; w, week(s); NIRF, near infrared fluorescence;

 IL, interleukin; EMMPRIN, extracellular

 matrix

metalloproteinase

 inducer; Ldlr, low density lipoprotein receptor.


(Continued)

TABLE

2


Selected

reviews.

**121**


mice and was able to significantly attenuate the development of atherosclerotic plaques (Kheirolomoom et al., 2015).

### TOOLS AND IDEAS FOR ADDITIONAL RESEARCH

As reviewed above, fundamental and pre-clinical researches on AI-drugs-loaded NPs against AS are beginning to bloom, and there are also clinical trials for the (separated) use of NPs and AI drugs for regression or consolidation of AS plaques. More fundamental and translational research on the mechanisms underlying their action could inspire and motivate future more efficient clinical trials using AI-drug loaded NPs against AS.

Important future steps for the development of new NP-based theranostics strategies can arise from the study of interaction mechanisms between NPs and immune cells, with particular regard to what fate NPs and their components will face upon site deposition and cellular uptake. Investigations on these directions will remarkably improve the development of NPs rational designs aimed at specific accumulation and cargo smart activation.

Indeed, where, how and in which proportion NPs are internalized in the various subtypes of ICs is still not clear, also due to the different internalization routes observed even in close phenotypes (Lunov et al., 2011), nor it is clear how or if they are exocytosed (Oh and Park, 2014). In particular, are NPs mostly internalized by circulating monocytes that enter the lesions, or by resident macrophages in lesions, where NPs enter because of the enhanced permeability of inflamed endothelium?

Immortalized cell lines can be used for preliminary experiments. The most used are murine macrophage-like RAW 264.7 and J774, and human monocyte-like THP-1 and U937 (Luster et al., 1995; Qin, 2012; Andreu et al., 2017) reviews in particular the use of THP-1, unchanged or differentiated toward a M8 phenotype (especially M1), but also cites other human monocyte/macrophage models. Following these models, even though there are evident parallelisms, care must be taken for intrinsic differences: between cells originating from different organisms (Ingersoll et al., 2010; Matoba et al., 2017; Zhang et al., 2017); between immortalized cell lines and primary cells (Andreu et al., 2017); even between possible different differentiations ICs may undergo during regular cultures. Primary cells should be used in final tests, but in this case the last type of unwanted differentiations are even more critical [e.g., monocytes are extremely sensible and may be activated just by sole adhering on surfaces (Belcastro et al., 2015)]. For these reasons, a thoughtful characterization of their phenotypes (by visual inspection or, better, by markers analysis) should be carried out before every experiment.

Together with the internalization pathway, also the biodistribution of NPs deserves attention; a nice review for both these issues, which considers MC and M8s, can be found in Pentecost et al. (2016), with interesting discussions about dependencies on NP geometry and surface chemistry, targeting, control on M8 phenotype, and imaging. In addition, medium and long-term fates of nano-construct components (in particular its bioactive cargo) should be addressed; these details can indeed impact the efficacy of the researched treatment, but also the long-term effects on the patients' health.

Finally, various animal models have been developed for the different phases of AS, as reviewed in Emini Veseli et al. (2017) and Lee et al. (2017). A recently developed model of controllable and reversible hypercholesterolemia is based on transient knockdown of the hepatic LDL receptor (LDLR) by antisense oligonucleotides in wild type C57BL/6 mice, followed by its rapid restoration (Basu et al., 2018).

### CONCLUSION

Cardiovascular events, such as acute myocardial infarction and stroke, are often associated with erosion/rupture of arterial atherosclerotic plaques and superimposed thrombosis. These can cause arterial-vessel occlusion, downstream ischemia and necrosis, with subsequent heart failure and the whole clinical spectrum of vascular encephalopathies. Being AS an important source of morbidity and mortality, intense research efforts toward precision and personalized medicine in this field are motivated.

There are several reviews for selected aspects of the highly complicated issue of AS and its theranostics with NPs (**Table 2**). Here, we shortly presented the broad background necessary to understand the use of NPs for smart delivery and activation of anti-inflammatory drugs. We tried to focus on how basic science, together with the translational and clinical aspects, can address different portions of this, aiming at providing a better

### REFERENCES


answer for the underlying mechanisms of NPs intra-plaque internalization, as well as for the molecular and cellular processes at the bases of different treatments. We stressed that a more thorough understanding of these would help foresee possible side effects, even long term ones, and also the opportunity to employ similar techniques and (nano)tools to other pathologies.

While single studies can be carried out in specialized groups, a deeper understanding of concepts, mechanisms and applicability of the developed (nano)tools needs a highly interdisciplinary environment, involving e.g. cardiologists, biologists, physicists, chemists, engineers; professionals able to talk with each other, and fostering the formation of multi-faceted scientists.

### AUTHOR CONTRIBUTIONS

SL conceived the idea. SL and AC wrote the manuscript, with major contributions from RP, RC, and GV, starting from texts written in collaboration with LM, ME, and PD. All authors discussed, edited, and contributed to the manuscript.

### FUNDING

SL acknowledges funding from Scuola Normale Superiore (university coordinated project Visualization and theranostic of atherosclerotic PLAQUEs VIA compound NANOparticles, PLAQUEviaNANO, in collaboration with Scuola Superiore Sant'Anna) and Fondazione Pisa (project Nanotechnology for tumor molecular fingerprinting and early diagnosis, RST 148/16).

contributes to b-gamma-glutamyltransferase accumulation inside atherosclerotic plaques. J. Transl. Med. 13:325. doi: 10.1186/s12967-015- 0687-6


permeable endothelium in experimental atherosclerosis. Proc. Natl. Acad. Sci. 111, 1078–1083. doi: 10.1073/pnas.1322725111


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Cervadoro, Palomba, Vergaro, Cecchi, Menichetti, Decuzzi, Emdin and Luin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Frontiers of Medical Micro/Nanorobotics: in vivo Applications and Commercialization Perspectives Toward Clinical Uses

Fernando Soto<sup>1</sup> \* and Robert Chrostowski <sup>2</sup>

*<sup>1</sup> Department of Nanoengineering, University of California, San Diego, La Jolla, CA, United States, <sup>2</sup> Texas Materials Institute, The University of Texas at Austin, Austin, TX, United States*

The field of medical micro/nanorobotics holds considerable promise for advancing medical diagnosis and treatment due to their unique ability to move and perform complex task at small scales. Nevertheless, the grand challenge of the field remains in its successful translation towards widespread patient use. We critically address the frontiers of the current methodologies for *in vivo* applications and discuss the current and foreseeable perspectives of their commercialization. Although no "killer application" that would catalyze rapid commercialization has yet emerged, recent engineering breakthroughs have led to the successful *in vivo* operation of medical micro/nanorobots. We also highlight how standardizing report summaries of micro/nanorobotics is essential not only for increasing the quality of research but also for minimizing investment risk in their potential commercialization. We review current patents and commercialization efforts based on emerging proof-of-concept applications. We expect to inspire future research efforts in the field of micro/nanorobotics toward future medical diagnosis and treatment.

### Edited by:

*Gianni Ciofani, Politecnico di Torino, Italy*

#### Reviewed by:

*Arianna Menciassi, Scuola Sant'Anna di Studi Avanzati, Italy Edoardo Sinibaldi, Fondazione Istituto Italiano di Technologia, Italy*

\*Correspondence:

*Fernando Soto f1soto@ucsd.edu*

#### Specialty section:

*This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology*

Received: *29 August 2018* Accepted: *25 October 2018* Published: *14 November 2018*

#### Citation:

*Soto F and Chrostowski R (2018) Frontiers of Medical Micro/Nanorobotics: in vivo Applications and Commercialization Perspectives Toward Clinical Uses. Front. Bioeng. Biotechnol. 6:170. doi: 10.3389/fbioe.2018.00170* Keywords: nanomedicine, medical translational research, in vivo, commercialization, microrobot, nanorobot

### INTRODUCTION

Imagine a world where robots the size of cells operate inside our body. This might sound like a science fiction story written by Isaac Asimov, or a visionary speech from Richard Feynman; however, it is conceivable that micro/nanorobotics will soon play a prominent role in medicine. (Wang, 2013; Wang et al., 2013; Li et al., 2017b) We use the term medical micro/nanorobots to refer to all nano- to micron-size structures (300 nm−300µm) capable of converting power sources into kinetic energy.Three groups of powered micro/nanorobots are mainly described. Biohybrid systems integrate synthetic nanostructures with motile microorganisms as the engine of the micro/nanorobot. (Kim and Tung, 2015; Ricotti et al., 2017; Bente et al., 2018; Palagi and Fischer, 2018). Chemically powered micro/nanorobots use asymmetric catalytic engines to selectively convert chemical fuels into locomotion (Chen et al., 2016; Nourhani et al., 2017; Schattling et al., 2017). Physically powered nanorobots convert external energy inputs (e.g., magnetic, ultrasound, or light fields) into translational motion based on engine geometry and material designs (Garcia-Gradilla et al., 2013; Dai et al., 2016; Bi et al., 2018; Pal et al., 2018).

The ability of micro/nanorobots to perform different task has been demonstrated at the laboratory scale, reporting their use for diverse proof of concept applications, including targeted cargo delivery (Solovev et al., 2010; Srivastava et al., 2018), fluid mixing (Orozco et al., 2013; Singh et al., 2015) and physical manipulation of micro objects (Cappelleri et al., 2014; Schuerle et al., 2017). The clinical aspirations of medical micro/nanorobots are still beyond the current capabilities of nanotechnology and bioengineering. Nevertheless, recent engineering breakthroughs have led to the successful in vivo operation of medical micro/nanorobots, illustrating initial proofs of concept for biopsy, delivery, healing and retention, as represented by the scheme in **Figure 1A.** While promising, these technological innovations are still challenging to translate into actual clinical therapies due to the safety concerns and the complexity of operating inside the human body.

This review focuses on the recent progress in the in vivo usage of micro/nanorobotics and on the efforts to commercialize and translate laboratory results into clinical applications. While there are several micro/nanorobot reviews addressing power and actuation principles (Sánchez et al., 2014; Teo and Pumera, 2016; Tu et al., 2017; Ren et al., 2018), fabrication procedures (Lin et al., 2015; Wang and Pumera, 2015; Jurado-Sánchez et al., 2017), and applications (Guix et al., 2014; Peng et al., 2017; Kim et al., 2018; Luo et al., 2018; Safdar et al., 2018), none of these reviews have addressed the crucial emerging clinical translations and potential commercial uses. We envision medical micro/nanorobotics as the frontier in treatment and diagnosis, potentially entailing benefits to human health by opening new therapies that are otherwise impossible to achieve.

### In vivo MICRO/NANOROBOTIC APPLICATIONS

The applications of micro/nanorobots for medical purposes in animal models are still limited compared to the large number of in vitro proofs of concept. However, the increase in the number of cumulative in vivo micro/nanorobotic publications (**Figure 1B**) and the high impact factor of the journals in which they are published (**Figure 1C**), both attest to the advancement in medical applications for micro/nanorobots, and to the encouraging level of interest within the scientific community (based on reviewed articles summarized in **Table 1**). Although there are multiple methodologies to power micro/nanorobots, we identify that only biohybrid (20%), chemical (30%), and physical systems (50%) have been used within inside living animals. The in vivo studies are detailed in the following sections based on their clinical aspiration or area of study.

### Delivery of Therapeutic and Imaging Agents for Cancer Therapy

Medical micro/nanorobotics hold great potential to deliver drugs with a higher degree of precision and speed when compared to passive diffusion methods. In general, this direction has been the motivation for in vivo applications (Erkoc et al., 2018). Thus, targeted delivery has benefited by recent in vitro developments in micro/nanorobotic chemotaxis (Peng et al., 2015; Shao et al., 2017) and material research using stimuli triggered drug release (Genchi et al., 2017; Rao et al., 2018). For example, magnetically guided nanorobots were used toward the delivery of fluorouracil medication for reducing tumor growth in a mice model. The released drug was externally triggered, allowing the nanorobotic platform to distribute a high amount of the therapeutic agent in a localized area of the tumor (Hoop et al., 2018).

Biohybrid nanorobots have also been used for targeted delivery of payloads inside living animals. Listeria monocytogenes has been used to deliver attached nanoparticles containing a payload of genes and proteins within a mouse. These payloads were used to monitor gene expression through differences in the luminescence produced within the different mouse organs (Akin et al., 2007). Magnetotactic bacterias, which naturally produce magnetic iron oxide nanoparticles, have been coupled with liposomes loaded with therapeutic payloads in vitro, as shown in **Figure 2A** (Taherkhani et al., 2014). More recently, these modified bacteria were guided using an external magnetic field to deliver the drug-loaded liposomes in vivo to a mouse tumor site (**Figure 2B**) (Felfoul et al., 2016).

The use of fully bio-engineered biohybrid micro/nanorobots without any inorganic/artificial components for carrying and transporting the therapeutic cargo has become possible through recent advances in synthetic biology. The use of genetically engineered bacteria, S. Typhimurium, have been reported to locally produce a therapeutic payload (α-emolysin E, a poreforming toxin) and to trigger the payload's release upon bacterial lysis. A small number of bacteria survive the lysis event, which allows for a continuous and cyclical delivery process controlled by an activator/repressor, synchronized lysis circuit (Din et al., 2016).

## Transport and Release of Cells

Micro/nanorobots have also been used toward delivering stem cells to a damaged location for tissue restoration. Magnetically guided microrobots have been reported toward carrying and delivering live cells to targeted areas in the body. In vivo transport and proliferation of HeLa cells in a nude mouse model demonstrated that the carried cells could be spontaneously released from the microrobot to the surrounding tissues and proliferate as shown in **Figure 3** (Li et al., 2018). These applications demonstrate that micro/nanorobots could serve as platforms for regenerative medicine and cell-based therapy, potentially proving to be especially useful in the later stages of life, when organs and systems start to fail. Moreover, there are still plenty of other promising in vitro applications to be developed for biohybrid microrobots, such as using a helical structure to guide a sperm toward an egg, for assisted fertilization (Magdanz et al., 2017).

### Retention of Payloads in the Gastrointestinal Tract

The aim of medical micro/nanorobotics is not only to deliver therapeutic payloads to a specific site but also, to retain the payloads within site as long as possible. In this direction, Wang's group has proposed the use of biodegradable zinc and magnesium powered microrobots that utilize gastric and intestinal fluids as fuels to promote cargo retention in the

stomach and intestinal tissues (Gao et al., 2015; Esteban-Fernández de Ávila et al., 2017a). This retention platforms have been applied toward pH neutralization of the gastric fluid (Li et al., 2017a) and for the treatment of a bacterial infection (Helicobacter pylori) in the stomach (Esteban-Fernández de Ávila et al., 2017b). This retention of the microrobot could be explained by direct piercing the surrounding tissue, or by an improvement in mass transport and nucleation due to the gas bubbles generated as means of locomotion for the microrobot, in an effect similar to effervescence.

Magnesium-based microrobots have also been designed with built-in delay activation, by using polymeric enteric coatings that activate the microrobot motion based on their thickness or environmental pH conditions. The polymeric coating only dissolves at neutral pH conditions found in the intestinal fluids resulting in localized ignition of the microrobot. The thickness of the coating allows to selectively localize the retention of the microrobots in different target sections of the gastrointestinal tract (the duodenum, jejunum, and ileum) (**Figure 4**) (Li et al., 2016). More recently, microrobots were integrated/loaded inside a pill matrix toward streamlining their administration with existing pharmaceutical protocols (Karshalev et al., 2018). In general, the use of magnesiumbased microrobots could benefit medical applications where autonomy and simplicity are desired. However, these chemicallypropelled microrobots might be limited to operate in largescale areas, such as the digestive system, as their depletion of propellant through bubble generation results in a short lifetime that could create unexpected complications in smaller capillaries.

### Wound Healing

The human body has diverse mechanisms and biological triggers for identifying wounds and repairing them. Nevertheless, these biological mechanisms can fall short when the wound is bleeding profusely, or there are not enough localized coagulant agents in the target region (Das and Baker, 2016). In this direction, medical micro/nanorobotics aim to simulate such systems by using active delivery toward fast and effective wound healing. Chemically-propelled calcium carbonate-based microrobots have been reported for delivering thrombin to halt the bleeding of wounds in the vasculature of mouse and pig models. The distribution mechanism relied on a combination of lateral propulsion, buoyant rise and convection (Baylis et al., 2015). Another reported approach consisted in the use of locomotive microrobots toward laser-based wound sealing. The high temperature generated by the laser-microrobot interaction produced localized collagen denaturation and melting, whereby a subsequent temperature decrease allowed condensation and wound closure (He et al., 2016).

### Biopsy

Other in vivo applications with potential implications in medicine are the ones targeting biopsy/surgery. Diverse in vitro platforms have been proposed toward precision micro/nanoscale surgery but still have not been translated to in vivo models


(Nelson et al., 2010; Xi et al., 2013; Kwan et al., 2015; Soto et al., 2015). Nonetheless, micro/nanorobotics could serve as a complement to current minimally surgical procedures, allowing unprecedented access into diseased tissues for biopsy analysis or therapeutic applications.

Microrobots with star-shaped grippers, which can reach narrow conduits in the body, have been used to excise tissue samples from a pig bile duct (Gultepe et al., 2012). Additionally, an initial proof of concept based on magnetic microrobots has demonstrated controlled navigation inside the eye of a living rabbit. Although this method has not been demonstrated directly for a surgical procedure, a magnetic coil system enabled the precise navigation of the untethered magnetic microrobots in the posterior eye section (Ullrich et al., 2013; Pokki et al., 2016). Ultimately, biopsy applications are fertile ground for further micro/nanorobotics research as the success of these applications depends on the ability of the robot to physically manipulate its environment and on the ability of the robot's controller to retrieve the robot. Both these problems have been studied much less extensively than propulsion.

### Local Mixing for Enhanced Thrombolysis

The potential modes through which a micro/nanorobot can physically manipulate its environment are not limited to merely excising tissue. Another recent trend in in vivo micro/nanorobots applications is the use of mixing effects to promote blood clot dissolution. In this case, magnetically actuated nanorobots loaded with tissue plasminogen activator were intravenously injected.

The vasculature flow drove the nanorobots to the blood clot. Once at their destination, the nanorobots were rotated by an external magnetic field. Their rotation generated local flow mixing which induces an increased interaction of the tissue plasminogen activator molecule with the blood clot interface, resulting in acceleration of thrombolysis (Cheng et al., 2014). More recently, it was demonstrated that nanorobots can target blood clots in mice's brains (Hu et al., 2018).

### Real-Time Imaging

From the reviewed articles, none of the chemically propelled micro/nanorobots are supported with real-time imaging, which introduces a severe limitation for understanding their therapeutic effect. 75% of biohybrid robots are supported with real-time imaging, but fluorescence is the only technique used (Akin et al., 2007; Park et al., 2013; Din et al., 2016). Physical robots are supported with the most diverse type of imaging techniques, but only 60% of articles are supported with real-time imaging. Represented techniques include the use of endoscopy and X-rays to detect microgrippers inside the gastrointestinal tract (Gultepe et al., 2012), use of optical camera to visualize movement inside the eye (Ullrich et al., 2013; Pokki et al., 2016), and fluorescence imaging techniques to track the position of magnetically actuated helical microrobots inside the peritoneal cavity of a mouse (Servant et al., 2015) or subcutaneously (Li et al., 2018). Moreover, a dual imaging approach was used to detect biodegradable magnetic microhelix nanorobots in mice. Fluorescence imaging was used to detect the nanorobot's position inside the subcutaneous tissue, and the intraperitoneal cavity of a mouse and magnetic resonance-based imaging was used to detect the nanorobot's position inside the mouse's stomach, as shown in **Figure 5** (Yan et al., 2017). Future research such address this important parameter, a key aspect for medical micro/nanorobotic use in clinical uses will rely on individual or population tracking, with consideration of tissue background signal (Medina-Sánchez and Schmidt, 2017; Vilela et al., 2018; Wang et al., 2018a).

### Toxicity

Most of the review micro/nanorobot studies used in animal models, provide only qualitative analyses of safety and toxicology based on histological assays. Achieving accurate targeted delivery requires an understanding of how foreign materials accumulate throughout the body and how to minimize the distribution of the administered micro/nanorobots to non-target tissue, establishing their specific effect on health. Each micro/nanorobot design has different safety concerns. Biohybrids could infiltrate and proliferate in undesired ways. Chemically propelled micro/nanorobots might change the local chemical environment which could have a significant effect on the microbiome of the gastrointestinal tract. For physical micro/nanorobots, most of the materials themselves pose a danger, as they are rigid and nondegradable. Despite these differences, researchers could report more relevant parameters to address the toxicity even at this early stage. Potential parameters to be considered include: the number of motors used for treatment (number or grams) indicated the escalation of units/dosage (limit for toxicity and inefficacy), but most importantly the distribution and toxicity of the nanorobot's constitutive materials within the living system.

### Administration and Retrieval

The administration and retrieval strategy of nanorobots is not discussed in length in the reviewed articles, which present as the most common administration method injection (60%), followed by oral administration (30%), catheter (5%), and topical administration (5%). Regarding the retrieval strategy for micro/nanorobots, both biohybrid and chemical systems were considered to be biodegradable by the authors. Although, in some cases, the fate of their synthetic components was not explicitly stated or demonstrated. For physically powered robots, in most cases, the fate of the nanorobot structure was not clearly stated,

only three articles provide a retrieval mechanism and one other is biodegradable. One plausible solution for retrieval consists of the use of a magnetic catheter to enable both the deployment and retrieval of microrobots in clinical practice (Iacovacci et al., 2018). We note that recent in vitro research efforts have described micro/nanorobot systems that are fully biodegradable (Peters et al., 2015; Chen et al., 2016; Yan et al., 2017; Bozuyuk et al., 2018; Wang et al., 2018c).

### MICRO/NANOROBOTS: STEPS TOWARD CLINICAL TRANSLATION

Basic science research is fundamental for creating both breakthrough medical advances and economic growth. Nevertheless, most basic science research will not achieve clinical translation. This challenge arises, in part, from the prolonged time required to go from basic research to clinical trials. Moreover, the conclusion of most trials indicates that most of the new technologies are less effective or more harmful than the current standards of care. This is a lesson to be learned in the micro/nanorobotic community. It is essential to identify strengths/weaknesses and remain objective about their relevance for later development.

The potential of the current generation of micro/nanorobotics is based on their translational motion and mixing capabilities. The idea of "smart" nanorobots that explore their surroundings and respond to environmental stimuli, commonly described in the field, is still beyond the current capabilities of micro/nanorobotics. Crucial questions of environmental manipulation, micro/nanorobot retrieval, and toxicology still need to be addressed. Indeed, we are far away from a killer app, but this should not be considered uncommon or discouraging to current researchers in the field. As previously mentioned, medical technologies have a significant time delay from lab to the clinic/commercialization, in many cases, taking several decades.

Nevertheless, considerable progress has been achieved in the field. It took less than a decade to go from the initial proof

of concept using chemically propelled nanowires in peroxide (Paxton et al., 2004) to the recent explosion of in vivo studies over the past 3 years. In order to reach the inflection point toward clinical translation, it is necessary first to standardize the methodologies and reproducibility of micro/nanorobotics research and to actively consider their commercialization outlook. Scientists without clear monetary value proposition rarely see their discoveries materialize.

(i: control, ii: after 6 h, and iii: after 12 h of administration. Reprinted with permission from Li et al. (2016). Copyright 2016 American Chemical Society.

### Reproducibility and Standardization

Commonly, the performance of micro/nanorobotic medical technology in vitro is reported by measurements of movement, such as their velocity and directionality, or on indirect methods such as measuring the mean squared displacement of tracer

particles or comparing a clinical outcome (e.g., the percentage of dead cancer cells) vs. passive diffusion. Although these methods serve to validate early proof-of-concept studies, they are not good predictors for successful clinical translation.

The variability between currently used in vivo micro/nanorobotic platforms makes it almost impossible to provide a comparative or quantifiable analysis of their efficiency vs. current standards of care. Moreover, there cannot be a clear definition of clinical efficiency when there is no minimum standardized information for reporting micro/nanorobotic research. Therefore, we need clear information standards in reporting experimental methodology, characterization, and results, which will be key for the successful translation of multidisciplinary research into improved health outcomes.

Some journals have implemented reporting summaries as a requisite for submission (Faria et al., 2018), which has shown to increase the quality of reporting for preclinical biomedical research (Han et al., 2017). In general, an experimental finding can be considered to have a stronger case for translation when the experiments supporting it have been carried out using blinding and randomization, and when, most importantly, those findings have been replicated in other laboratories apart from the inventors' group. Currently, no study presented in this review covers these standards. In this direction, here we suggest guidelines (measures that should be reported in parenthesis) for reporting micro/nanorobotic research that should allow better comparisons, both qualitative and quantitative, of micro/nanorobot performance in addition to already established checklists for safety and toxicology. The goal of the checklist is not to propose research trends or criticize current research, as new technologies start without clear standards. Instead, the homogeneity of experimental reporting can facilitate comparison, improve experimental design, and increase reproducibility between various micro/nanorobotic platforms.

The justification of the power source used to provide energy/movement to the micro/nanorobot could help to establish the unique advantage of each method. Material characterization of the micro/nanorobot, including their size distribution (length units), average lifetime of propulsion (time unit), surface charge (zeta potential), and storage lifetime (days), can help to plan into eventual fabrication and transport in medical uses. Moreover, is important to quantify the performance of the nanorobotic platform by evaluating the therapeutic effect based on experiment dependent parameters. For drug delivery applications, an important parameter is the administered and delivered dose (% by mass). For retention studies the number of nanomotors localized in the target region (% of administered population). For biopsy applications, the micro/nanorobot force applied to a specific area of tissue (Pascal). Moreover, the micro/nanorobot therapeutic effect should be compared against the state-of-the-art treatment, not only passive diffusion. Finally, parameters covered through the review includes, imaging (real time localization), toxicity (folllowing establish guidelines), Administration route (Target Area and method of insertion), and retrival region (% of administered population) could streamline their tranlatio into exisitng therapies.

### Intellectual Property and Commercialization

Clinical efficacy alone is not enough to warrant translation to market. The transition of new technologies from lab to market requires extensive research and development costs, overcoming regulatory barriers, and most importantly, having the right commercialization potential. The potential customer for micro/nanorobotics' technology is not the general population or the physicians, but rather pharmaceutical companies and insurance providers. Therefore, micro/nanorobotics research should consider their commercialization opportunities from the research stages, in order to provide tangible therapies that reduce cost, increase throughput efficiency and productivity.

Minimizing the investment risk can be achieved by protecting the intellectual property of the new technological developments (which is one of the most valuable assets for high-tech companies) and establishing a clear business model with identifiable revenue streams. New technological developments are commonly initiated using patents filed by research groups through their university system, which are later licensed to emerging start-up companies or larger corporations. The prohibitive cost of prototyping and clinical trials is initially financed by non-dilute financing, which includes government research and industry grants, and at later stages by investment from venture capital and industry partnerships.

Tracking the filing of relevant patents, and the formation of relevant startups, provides useful insight into the progress made in identifying potential markets, competitors and consumer needs. **Figure 6** traces the accumulated United States applied and granted patents in nanorobotics, while also separating them into the categories of fabrication, actuation, imaging, and application. The patent filing data were collected manually, as keyword searches do not accurately capture the state of nanorobotic technology due to the multiple names used to refer to nanorobots in the literature (micro/nanobot, micro/nanorobot, micro/nanorocket, micro/nanoengines, micro/nanomotor, micro/nano-swimmer, active colloid). Instead, the search was conducted by reviewing the output of the most prolific nanomotor research groups in the field that have published results in animal models and by reviewing relevant Google patents.

The first patents published consisted of the fabrication methods commonly used for parallel mass fabrication using template platforms, consisting of nanorod structures (Stonas et al., 2005; Natan et al., 2007; Fan et al., 2017b; Odell et al., 2018), microcoils (Schmidt et al., 2010; Jeong et al., 2018), tubular structures (Yao et al., 2015; Wang et al., 2016a), emulsions (Percec et al., 2017), biohybrids (Schmidt and Leibniz-Institut Fuer Festkoerper-Und Werkstofffor Schung, 2017), and colloid chains (Duan et al., 2017).

The next wave of patents consists of tracking/imaging nanorobots using magnetic (Martel et al., 2015; Muntwyler et al., 2017; Odell et al., 2017), optical (Benaron et al., 2008), and ultrasound methods (Mattrey et al., 2017). Followed by methods to power micro/nanorobots using magnetic (Solomon and Solomon Res LLC, 2011; Vollmers et al., 2013; Abbott et al., 2014; Fischer et al., 2014; Mahoney et al., 2014; Odell et al., 2016; Sitti et al., 2016; Tasci et al., 2016; Wang, 2017; Weinberg et al., 2017), ultrasound (Wang and Zhang, 2017; Wang et al., 2017), chemical (Paxton et al., 2009; Sen et al., 2017; Tang et al., 2017), and biohybrid propulsion mechanisms (Magdanz et al., 2018; Martel et al., 2018).

The final, most recent wave, has focused on applications for these micro/nanorobots, including for intracellular delivery by nanospearing (Cai et al., 2011) exploration of subterranean geophysical formations and oil retrieval (Kamal et al., 2015), nanomotor-based patterning of surface microstructures (Wang et al., 2016b), transporting, positioning and assembling nanorobots using electric twisers (Chien et al., 2017; Fan et al., 2017a), capturing and isolating of target biomolecules and living organisms using microrobots (Wang et al., 2018b).

There are only a few active companies working toward the commercialization of nanorobots for use in medical applications. Most notably, Nelson's group has spun off two companies. The first was Aeon Scientific, based on the electromagnetic manipulations systems developed to guide magnetic motors. However, they have recently shifted their focus to developing new manipulation systems for catheters. Indeed, it is not surprising that the technology developed around micro/nanorobots could have applications that were never envisioned at the beginning.

The second, Swiss Magnetibox formed in 2014 is based on wireless tools for actuating and imaging micro/nanomachines using magnetic fields. They primarily sell microscopy equipment coupled with magnetic actuation systems and rod shape microrobots toward expanding basic science in micro/nano robotics research. Although these systems are not specifically designed for in vivo use, they provide the necessary tools to test proof of concept applications (Schuerle et al., 2017).

Finally, Weingberg medical physics has developed an image-guided therapy for targeted delivery of magnetic micro/nanorobots that uses ultra-fast MRI to generate magnetic fields for imaging and manipulation of the magnetic nanorobots. They have demonstrated diverse in vitro applications using the mechanical force of the nanorobots to dislodge bacteria biofilm using rotating nanorobots (Mair et al., 2017) and drilling into mice's brain post-mortem (Jafari et al., 2019).

The mass fabrication of nanorobotics is one of the first challenges to be addressed toward tangible translation to market.

Reproducibility and availability of templates will be significant hurdles that must be solved with traditional micro/nanorobotic fabrication techniques. New methods based on 3D printing and two-photon lithography could allow for complex designs such as microhelices or cell carriers. Although, the cost and slow production throughput of these methods may limit the use of such intricate designs. Moreover, new methods of characterization and manipulation of individual nanorobots are necessary to ensure the quality of each batch. In this regard, nano-manipulation systems are ideal for transporting, inspecting and testing nanorobotic designs (Wang et al., 2015; Lu et al., 2017; Meng et al., 2017; Zhang et al., 2018). Once these technical capabilities are well established, the reduction in the cost to reach the market will allow new companies to spring up and provide innovative medical value propositions.

A possible path toward early adoption of micro/nanorobots in the clinic could be achieved by offering micro/nanorobotic technology as a complementary tool to existing medical procedures. For example, their integration with current oral delivery platforms such as pills, allow for dose escalation by increased retention of therapeutic payloads at the mucosae wall. Although special consideration should be placed in therapies where the primary therapeutic effect is based on systemic distribution, as in this case, the dose escalation proposed by micro/nanorobotic platforms is not desired. Moreover, micro/nanorobots used for tissue biopsy and suturing veins could be integrated with a catheter and mili surgical robotics, allowing the larger robots to reach scale ranges where their size normally does not permit them to operate. The possibilities discussed here are driven by the need to improve patients' outcomes after surgical procedures including reduced hospital stay, lower chances of infection, and minimum scarring. We should note that it is not likely that micro/nanorobots will be used for preventive care or as a chronic treatment since the sustained introduction of synthetic objects into the human body might produce unknown repercussion due to possible accumulation.

### REFERENCES


## FUTURE OUTLOOK

The field of medical nanorobotics has achieved considerable advances. However, several issues and challenges must be addressed before micro/nanorobots could have real-world clinical applications. The goal of the in vivo model is not only to evaluate the therapeutic efficiency of the platforms, but to identify the clinical risk, as evaluating off-target effects of nanorobots is as essential as evaluating efficacy. Indeed, there is a discrepancy between the aspirations of medical nanorobotics and reality, as the legacy of science fiction has set the conceptual boundaries of what to expect long before scientists could. The manufacturing of the micro/nanostructure engines must be optimized, with special consideration for material biocompatibility and degradation, to address in vivo safety concerns.

Furthermore, proper standards should be established to clarify the advantages of micro/nanorobot therapies over traditional methods which already fulfill FDA standards. Nonetheless, micro/nanorobotics might potentially improve medical diagnosis and treatment. We should also consider that the designs and aspirations of a small group of scientist and engineers could soon affect the lives of millions of people directly and profoundly, therefore it is essential to consider the economic, social, and ethical implications of the use of medical nanorobotics. These implications are likely to be on par with those of the most significant technological revolutions.

### AUTHOR CONTRIBUTIONS

FS and RC wrote and revised the manuscript and approved it for publication.

### ACKNOWLEDGMENTS

FS acknowledges the funding from the UC MEXUS-CONACYT program. RC acknowledges his funding from the Cockrell School of Engineering at The University of Texas at Austin.


microrobots for implantation: corrosion, cell culture, and in vivo animal tests. J. Biomed. Mater. Res. B Appl. Biomaterials 105, 836–845. doi: 10.1002/jbm.b.33618


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Soto and Chrostowski. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Transdifferentiating Astrocytes Into Neurons Using ASCL1 Functionalized With a Novel Intracellular Protein Delivery Technology

Meghan Robinson<sup>1</sup> , Ian Fraser 1,2, Emily McKee<sup>2</sup> , Kali Scheck <sup>3</sup> , Lillian Chang<sup>4</sup> and Stephanie M. Willerth1,2,5,6,7 \*

<sup>1</sup> Division of Medical Sciences, University of Victoria, Victoria, BC, Canada, <sup>2</sup> Biomedical Engineering Program, University of Victoria, Victoria, BC, Canada, <sup>3</sup> Biology Program, University of Victoria, Victoria, BC, Canada, <sup>4</sup> Biochemistry Program, Bates College, Lewiston, ME, United States, <sup>5</sup> Mechanical Engineering, Faculty of Engineering, University of Victoria, Victoria, BC, Canada, <sup>6</sup> Center for Biomedical Research, Faculty of Engineering, University of Victoria, Victoria, BC, Canada, <sup>7</sup> International Collaboration for Repair Discovery, University of British Columbia, Vancouver, BC, Canada

Edited by:

Gianni Ciofani, Politecnico di Torino, Italy

#### Reviewed by:

Stefania Moscato, Università degli Studi di Pisa, Italy Madoka Suzuki, Osaka University, Japan

> \*Correspondence: Stephanie M. Willerth willerth@uvic.ca

#### Specialty section:

This article was submitted to Nanobiotechnology, a section of the journal Frontiers in Bioengineering and Biotechnology

Received: 12 September 2018 Accepted: 31 October 2018 Published: 21 November 2018

#### Citation:

Robinson M, Fraser I, McKee E, Scheck K, Chang L and Willerth SM (2018) Transdifferentiating Astrocytes Into Neurons Using ASCL1 Functionalized With a Novel Intracellular Protein Delivery Technology. Front. Bioeng. Biotechnol. 6:173. doi: 10.3389/fbioe.2018.00173 Cellular transdifferentiation changes mature cells from one phenotype into another by altering their gene expression patterns. Manipulating expression of transcription factors, proteins that bind to DNA promoter regions, regulates the levels of key developmental genes. Viral delivery of transcription factors can efficiently reprogram somatic cells, but this method possesses undesirable side effects, including mutations leading to oncogenesis. Using protein transduction domains (PTDs) fused to transcription factors to deliver exogenous transcription factors serves as an alternative strategy that avoids the issues associated with DNA integration into the host genome. However, lysosomal degradation and inefficient nuclear localization pose significant barriers when performing PTD-mediated reprogramming. Here, we investigate a novel PTD by placing a secretion signal sequence next to a cleavage inhibition sequence at the end of the target transcription factor–achaete scute homolog 1 (ASCL1), a powerful regulator of neurogenesis, resulting in superior stability and nuclear localization. A fusion protein consisting of the amino acid sequence of ASCL1 transcription factor with this novel PTD added can transdifferentiate cerebral cortex astrocytes into neurons. Additionally, we show that the synergistic action of certain small molecules improves the efficiency of the transdifferentiation process. This study serves as the first step toward developing a clinically relevant in vivo transdifferentiation strategy for converting astrocytes into neurons.

Keywords: reprogramming, transcription factors, neuroscience, small molecules, drug delivery

## INTRODUCTION

Transdifferentiation converts mature cells from one specialized cell type into another by varying gene expression patterns (Tanabe et al. 2015). These expression patterns are controlled by transcription factors, regulatory proteins which bind to DNA promoter regions to activate transcription of key developmental genes. Ectopic expression of these transcription factors in somatic cells can change cell fates without the need for inducing a pluripotent state (Tanabe et al., 2015). This process avoids the risk of tumorigenesis associated with transplanting pluripotent stem cell-derived therapies in vivo, where incomplete differentiation leads to uncontrolled proliferation (Gordeeva and Khaydukov, 2017). The final genetic signatures of these transdifferentiated cells are stable and not restricted to the lineage of the original cell type, allowing cells to be sourced directly from endogenous tissue local to the site of injury (Tanabe et al., 2015). Performing in vivo transdifferentiation would eliminate the need for cell transplantation and immunosuppression depending on the target application.

Ectopic transcription factor expression can be efficiently accomplished through insertion into the genome by viral vectors encoding the amino acid sequence of the desired transcription factor. However, this method is not safe for in vivo use in clinical trials due to the risk of activating oncogenes by random DNA insertion into the host genome (Li et al., 2002; Modlich et al., 2006; Medvedev et al., 2010). Non-integrating vectors have been developed capable of transdifferentiation, but at lower efficiencies. Some wellknown examples of non-integrating methods include vectors composed of single-stranded RNA, such as the Sendai virus, circular DNA, such as plasmid vectors and minicircles, and mRNA vectors, which reside in the cytoplasm where they are translated into proteins. Despite these improvements in non-viral delivery, vector-based transdifferentiation strategies still face challenges such as transgene silencing, inflammation due to the presence of residual bacterial DNA, and poor nuclear uptake due to their physical size (Hardee et al., 2017).

The search for non-integrating, non-viral reprogramming methods has led to the discovery of transdifferentiation using microRNAs and small molecules (Ma et al., 2017). In 2011, it was determined that the microRNAs miR-124 and miR-9/9<sup>∗</sup> were sufficient to induce neurons from fibroblasts (Ambasudhan et al., 2011; Yoo et al., 2011;). Later in 2014 Zhu et al. discovered that viral-mediated expression of octamer binding transcription factor 4 (OCT4) was sufficient to transdifferentiate somatic cells into neural stem cells when combined with small molecules that promote lineage-specific signaling (Zhu et al., 2014). Morphogens and bioactive small molecules can potentially enhance the action of transcription factors by improving chromatin accessibility, targeting key signaling pathways, and regulating metabolism (Lin and Wu, 2015). These results led several groups to identify combinations of small molecules capable of transdifferentiating somatic cells such as fibroblasts, astrocytes and even glioblastoma cells into neurons (Cheng et al., 2014; Hu et al., 2015; Li et al., 2015; Zhang et al., 2015, 2016; Gao et al., 2017; Lee et al., 2018). Although the transdifferentiation methods differ, these small molecule and microRNA protocols share a common thread in that they produce mainly one neuronal subtype–medium spiny neurons in the case of microRNAs, or interneurons in the small molecules methods with the exception of the glioblastoma-transdifferentiated neurons which resulted in immature post-mitotic neurons. The inability to produce the neuronal subtypes which are lost in neurodegenerative disorders like Parkinson's Disease, Alzheimer's Disease, amyotrophic lateral sclerosis (ALS), Huntingdon's Disease, spinal cord injury and blindness, represents a major limitation in current small molecule and microRNA reprogramming protocols.

Researchers have begun to explore protein transduction domains (PTDs), also known as cell penetrating peptides (CPPs), as tools for transdifferentiation to address these limitations (Zahid and Robbins, 2012). These CPPs are small peptides that penetrate the cell membrane to deliver bioactive macromolecules. The first PTD was derived from the transactivator of transcription (TAT) protein sequence of the human immunodeficiency virus (HIV), which binds to a phospholipid component of the inner surface of the cell membrane through a cationic poly-arginine sequence to enable cellular internalization (Yang et al., 2012b; Gordeeva and Khaydukov, 2017). Other examples of PTDs include Penetratin, a cationic arginine-rich sequence taken from the Drosophila protein antennapedia homeotic transcription factor (Antp), and CADY, a sequence combining arginine residues with aromatic tryptophan, whose helical conformation within the cellular membrane favors cellular uptake (Crombez et al., 2009; van den Berg and Dowdy, 2011). PTDs have the potential to be fitted with binding domains to target specific cell populations, making them ideal tools for in vivo reprogramming of endogenous cells (van den Berg and Dowdy, 2011; Zahid and Robbins, 2015). For example, C-end rule (CendR) is an amino acid sequence containing a carboxy (C)-terminal motif which binds to neuropilin-1, a cell surface receptor expressed in epithelial, neuronal, and cancer cells (Teesalu et al., 2009; Pang et al., 2014). In 2014, Hu et al. successfully transdifferentiated neuropilin-1 expressing retinal pigmented epithelial cells into neurons using a C-end rule PTD fused to the transcription factor self-determining region Y-box2 (SOX2) (Hu et al., 2014).

Little work has been done to develop such protocols to date despite the promise of PTDs as a tool for in vivo transdifferentiation. A search using the Pubmed database revealed only one further publication using PTD delivery where Islas et al. transdifferentiated human dermal fibroblasts into immature contractile cardiomyocytes (Islas et al., 2012). They fused TAT-PTDs to two cardiac transcription factors, erythroblastosis virus E26 oncogene homolog 2 (ETS2) and mesoderm posterior 1 (MESP1). Treatment with these TAT-PTDs followed by treatment with cardiac gene-inducing morphogens activin A and BMP2 resulted in successful transdifferentiation of fibroblasts into functional cardiomyocytes. Morphogen treatment significantly improved induction of key cardiac genes, whereas addition of the morphogens alone had no effect, further illustrating the transdifferentiation potential of combining morphogens or bioactive small molecules with transcription factors.

In this work, we developed a protocol using a novel PTD to transdifferentiate somatic cells into interneurons. We hypothesized that the delivery of the neural transcription factor achaete scute homolog 1 (ASCL1) by this novel PTD design in combination with the action of bioactive small molecules would promote neuronal transdifferentiation of astrocytes, the most abundant cell type in the central nervous system. ASCL1 is a transcription factor and master regulator of neurogenesis (Vasconcelos and Castro, 2014). Its ability to bind readily to closed chromatin and promote the appearance of new regions of open chromatin allows coordinated activation of previously silenced genes (Iwafuchi-Doi and Zaret, 2014; Raposo et al., 2015). In 2015, Liu et al. showed that in vivo viral delivery of ASCL1 to mouse dorsal midbrain astrocytes was sufficient for neural transdifferentiation, with 76.8% of astrocytes adopting a neural fate by day 10 (Liu et al., 2015). ASCL1 was also chosen for its versatility. While the action of ASCL1 alone leads to an interneuron fate, it is known to induce dopaminergic, cholinergic or serotonergic neuronal subtypes when combined with dopaminergic factors LMX1B/NURR1 (Addis et al., 2011; Theka et al., 2013), cholinergic factors MYT1L/BYRN2/LHX3/HB9/ISL1/NGN2 (Son et al., 2011), or serotonergic factors FOXA2/LMX1B/FEV (Xu et al., 2016). The PTD employed for these experiments was designed by iProgen Biotech Inc. to take advantage of the cell's innate retrograde transport pathway by integrating a secretion signal peptide and a cleavage inhibition peptide to a cargo protein (Lee et al., 2013). Initially the PTD integrates itself into the cell surface membrane due to its amphipathic nature. Once internalized, the secretion signal peptide preferentially binds to its native binding partner, translocon, a complex of proteins tasked with translocating polypeptides with a targeting signal sequence into the interior space of the endoplasmic reticulum (ER). As the translocon retrogrades back through the ER into the cytoplasm, so does the cargo. This combination of secretion signal peptide and cleavage inhibition peptide, termed IPTD, shows greatly improved delivery of proteins into the cytoplasmic space over conventional PTDs such as TAT and poly-arginine. In 2016, we showed that ASCL1-IPTD could efficiently generate neurons from human induced pluripotent stem cells (hiPSCs) (Robinson et al., 2016). We show here that ASCL1-IPTD in combination with the small molecules LDN193189, SB431542, DAPT and valproic acid can rapidly reprogram astrocytes into mature GABAergic and glutamatergic interneurons with high efficiency, and that this process is unique to the transcription factor ASCL1. We also show that in the absence of these small molecules ASCL1-IPTD cannot transdifferentiate into a neuronal fate, but does generate a divergent cell fate of myoblasts. Because of its inherent versatility in generating other neuronal subtypes, this ASCL1-IPTD protocol serves as a starting point for further development through the functionalization of subtypespecific factors with IPTD to generate subtypes of neurons. Additionally, the use of IPTD allows for future adaption with specific tissue-targeting motifs. This method can be applied as novel treatment for neurodegenerative diseases or spinal cord injury by transdifferentiating astrocytes into specific neuronal populations as a way to restore function.

### MATERIALS AND METHODS

### Cell Culture

Human fetal astrocytes (ScienCell cat#1800) were cultured on poly-L-lysine coated plates in Astrocyte Media (ScienCell). Media was changed every 2 days and cells were passaged 1:6 when 90% confluent using trypsin-EDTA (ScienCell). 10% fetal bovine serum (FBS, Gibco) was added to media for 6 passages (6 weeks) as previously described to purify astrocyte cultures of neural progenitors (Zhang et al., 2015). Astrocytes were incubated for 24 h on an orbital shaker plate at 200 rpm as previously described to purify low passage (P2) astrocytes of neural progenitors (Lian et al., 2016). Mouse embryonic fibroblasts (MEFs) were purchased from MTI GlobalStem and thawed onto poly-L-ornithine (PLO)/laminin (Sigma Aldrich) substrates in high glucose DMEM (Gibco) supplemented with 10% FBS.

### Astrocyte Reprogramming

Astrocytes were grown to confluence before reprogramming. Cells were cultured on poly-L-lysine (PLO)/laminin substrates and media was changed to high glucose Dulbecco's Modified Eagle Medium (DMEM) with 2 mM Glutamax (Gibco), 1% N2/B27 (ThermoFisher Scientific), 1% Human Serum Replacement 3 (Sigma Aldrich), and 1% penicillin/streptomycin (ThermoFisher Scientific) for cellular reprogramming experiments. 15µg/mL ASCL-IPTD (iProgen Biotech) was supplemented for all 12 days, with 50% media changes every 2 days. For days 1-2, 5µM SB431542 (Stemcell Technologies) and 0.25µM LDN193189 (Stemcell Technologies) were added to cultures. For days 3–12 the following small molecules were added: forskolin (F, 10µM, Medchem Express), CHIR99021 (C, 1.5µM, Medchem Express), ISX9 (I, 20µM, Medchem Express), DAPT (D, 5µM, Medchem Express). The combinations tested were C, F, D, I, CF, CI, CD, FI, FD, ID, FCI, FCD, CID, FID, CFID. For days 3–6, 0.5µM valproic acid (Stemcell Technologies) was also added to improve chromatin accessibility.

### Fibroblast Reprogramming

MEFs were seeded 40,000 cells/cm<sup>2</sup> on PLO/laminin and cultured for 24 h in high glucose DMEM + 10% FBS to adhere. Media was then switched to high glucose DMEM, 1% N2/B27, 1% Human Serum Replacement 3, 2 mM GlutaMAX, 1% penicillin/streptomycin, 15µg/mL ASCL1-IPTD and 20 mM forskolin for 14 days, with media changes of 50% every 2nd day.

### Immunocytochemistry

Immunocytochemistry was performed as described previously (Agbay et al., 2018). Cells were fixed with 3.7% formaldehyde/phosphate buffered saline (PBS) solution for 1 h, then permeabilized in 0.1% Triton X-100 (Sigma Aldrich) in PBS for 45 min at 4◦C, then blocked with 5% normal goat serum (NGS, Sigma Aldrich) in PBS for 2 h at 4◦C. Cells were then incubated with the primary antibodies mouse anti-TUJ1 (Millipore), rabbit anti-MYH3 (Abcam), rabbit anti-ASCL1 (Abcam), rabbit anti-DCX (Abcam), rabbit anti-SOX2 (Abcam), rabbit anti-MAP2 (Abcam), rabbit anti-GLUT1 (Abcam), rabbit anti-GAD65/67 (Abcam), mouse anti-NEUN (Novus Biologicals), mouse anti-TH (Stemcell Technologies), or goat anti-CHAT (Millipore), diluted 1:500 in PBS, and incubated overnight at 4◦C in the dark. The next day cells were washed with PBS three times for 15 min each, and incubated with secondary antibodies goat anti-mouse IgG AlexaFluor488 (ThermoFisher Scientific), goat anti-rabbit IgG AlexaFluor568 (ThermoFisher Scientific), or donkey anti-goat AlexaFluor405 (Abcam) diluted 1:200 in PBS for 4 h at 4◦C. Cells were washed again three times as previously described. For cells counterstained with DAPI nucleic acid stain, a 300 nM DAPI solution (Invitrogen) in PBS was added to the cultures after the final wash and incubated for 3 min, followed by rinsing with PBS. Cells were visualized and imaged with a Cytation5 (Biotek) Imaging platform at 20 × magnification. Fluorescently labeled cells were visualized with Biotek Dapi (377/477), GFP (469/525), and Texas Red (586/647) filter cubes, using dichroic mirror-based wavelength selection.

### Flow Cytometry

Flow cytometry was performed as previously described (Montgomery et al., 2015). The cells were stained for SOX2-PE Mouse IgG2A (R&D Systems) with IgG2A-PE isotype control (R&D Systems), TUJ1-PerCP Mouse IgG2A (R&D Systems) with IgG2A-PerCP isotype control (R&D Systems), MAP2-PE Mouse IgG1 (Clone AP20, Milli-Mark) with Mouse IgG1 PE isotype control (R&D Systems), and NEUN-PE Mouse IgG1 (Clone A60) with Mouse IgG1 PE isotype control (R&D Systems). Cells were collected rinsed with PBS 3 times, followed by incubation in fixation/permeabilization buffer (R&D Systems) for 10 min, and 3 more washes with PBS. Next cells were resuspended in permeabalization/wash buffer (R&D Systems) with each antibody, diluted as per manufacturer's instructions, for 45 min at 4◦C in the dark. Cells were then washed in permeabilization/wash buffer 3 times. Data was collected using the Millipore Guava EasyCyte HT flow cytometer. Cells were gated to exclude small debris and clustered cells, and a maximum of 5,000 gated events were collected for each sample. All analysis was completed using GuavaSoft EasyCyte software.

### Quantification of Transdifferentiated Neurons

Cells were first immunostained with the neuronal cytoskeletal marker TUJ1 to quantify transdifferentiated neurons in each group of the small molecule screen. Next, a grid of 5 × 5 images was taken in each of 3 wells for each group (25 images per well, n = 3 biological replicates). Cells were visualized with a Leica DMI 3000B microscope equipped with an XCite Series 120Q fluorescent light source and QImaging RETIGA 2000R camera at 100 × magnification, and images were captured using QCapture Software 2.9.12 with a GFP filter (blue/470, Leica Systems) and a DAPI filter (400, Leica Systems) and dichroic mirrors (500, Leica Systems). Cells were counted manually if they were positive for TUJ1 and possessed a small rounded soma, with at least 2 neuritic processes. Total cells were counted by DAPI stain using ImageJ: images were made binary and a watershed was applied to separate nuclei that were touching by a single pixel width, and each cell was counted as a particle under the Analyze-Particles Menu.

### Statistical Analysis

Statistical analysis was performed by a one-way ANOVA followed by a one-tailed Student's t-test with a 95% confidence level (α = 0.05). Results are presented as the mean ± standard deviation. All experiments were carried out with a biological n = 3.

### RESULTS

### Transdifferentiation Protocol Developmental

We sought to characterize the transdifferentiation strength of ASCL1-IPTD before designing our protocol. We executed a well-known protocol where ectopic expression of ASCL1 alone converts mouse embryonic fibroblasts (MEFs) into neurons, and at lower expression levels, into skeletal myoblasts (Treutlein et al., 2016). The addition of the small molecule forskolin, an activator of cyclic adenosine monophosphate (cAMP) signaling, dramatically improves the efficiency of this protocol and so it was tested in parallel to determine if conversion efficiency could be similarly improved with ASCL1- IPTD (Shi et al., 2016). After 14 days of treatment, skeletal myoblasts were generated instead of neurons as indicated by myosin heavy chain 3 (MYH3) expression, with or without the addition of forskolin (**Figure 1)**. Increasing the concentration of ASCL1-IPTD 5-fold also had no effect in generating neurons (**Supplemental Figure 1**).

We concluded that the strength of ASCL1-IPTD was not meeting the threshold expression level necessary for neuronal transdifferentiation. Accordingly, we adopted the strategy to first lower this threshold by repressing astrocytic gene expression using small molecules to inhibit transforming growth factor β1 (TGF-β1)/SMAD and bone morphogenic protein (BMP)/SMAD signaling pathways, which are jointly responsible for astrocyte commitment and maintenance (Stipursky and Gomes, 2007; Chambers et al., 2009; Yang et al., 2013; Zhang et al., 2015). Human cerebral cortex astrocytes were primed for 2 days followed by screening of a set of small molecules known to have neuronal transdifferentiation potential: CHIR99021, an inhibitor of glycogen synthase kinase 3 (GSK3) signaling; DAPT, an inhibitor of Notch signaling; forskolin, an activator of cAMP signaling; and isoxazole 9 (ISX9), an activator of calcium-mediated signaling (Zhang et al., 2015; Gascón et al., 2016, 2017; Gao et al., 2017). For days 3–6, valproic acid, a histone deacetylase inhibitor, was added to enhance chromatin accessibility (Huangfu et al., 2008). ISX9 induced a significant degree of toxicity in cultures, and the combination of CHIR99021 and ISX9 caused all the cells to detach from the plate (data not shown). All remaining combinations produced some degree of transdifferentiation from astrocytes into neurons within 10 days, with DAPT alone being the most effective enhancer of this process (**Figure 2**).

### The Action of DAPT Does Not Translate to NGN2-Mediated Transdifferentiation

We exchanged ASCL1-IPTD with another functionalized neurogenic transcription factor, NGN2-IPTD to determine if ASCL1 could be replaced with NGN2. NGN2 also serves

FIGURE 1 | Mouse embryonic fibroblasts (MEFs) express the skeletal myoblast gene myosin heavy chain 3 (MYH3) and lack the neuronal gene beta tubulin (TUJ1) after 14 days of exposure to ASCL1-IPTD and the cAMP activator forskolin, indicating that they have adopted a myogenic cell fate rather than a neuronal cell fate. (A) Control. Scale bar is 100µm. (B) High magnification of (A). Scale bar is 10µm. (C) Forskolin only. Scale bar is 100µm. (D) High magnification of (C). Scale bar is 10µm. (E) ASCL1-IPTD only. Scale bar is 100µm. (F) High magnification of (E). Scale bar is 10µm. (G) ASCL1-IPTD with forskolin. Scale bar is 100µm. (H) High magnification of (H). Scale bar is 10µm.

as a master regulator of neural fate, and when delivered virally is sufficient to transdifferentiate astrocytes into neurons with high efficiency (71%) (Parras et al., 2002; Berninger et al., 2007). However, replacing ASCL1-IPTD with NGN2-IPTD in combination with DAPT did not generate neurons, but was seen by immunocytochemistry to generate a population of cells expressing the neuroblast gene doublecortin (DCX), suggesting that ASCL1 and NGN2 are not interchangeable in our protocol **(Figure 3).**

FIGURE 2 | Small molecule screening of combinations of forskolin, ISX9, DAPT and CHIR99021 with ASCL1-IPTD show that DAPT is the most potent inducer of neural transdifferentiation. (A) Cells express beta tubulin (TUJ1), an early neuronal marker, and ASCL1 after 2 days of priming with LDN193189 and SB431542 and ASCL1-IPTD followed by 10 days of treatment with ASCL1-IPTD and DAPT. Scale bar is 100µm. (B) High magnification of (A). Scale bar is 20µm. (C) Cell survival varied in each condition, with the DAPT alone condition being the least toxic by far (n = 3). (D) Neurons counted from each small molecule combination on day 12 (n = 3). (E) Neuronal conversion percentages. (F) Protocol timeline: 2 days of priming with SMAD inhibitors LDN193189 and SB431542 followed by 10 days of small molecule combinations. All 12 days included the addition of ASCL1-IPTD. Statistical analysis was performed by a one-way ANOVA followed by a one-tailed Student's t-test with a 95% confidence level (α = 0.05). Results are presented as the mean ± standard deviation, with a \* representing a statistically significant difference between that condition and the DAPT alone condition.

FIGURE 3 | NGN2-IPTD in place of ASCL1-IPTD does not generate neurons, but does produce cells expressing doublecortin (DCX), a marker for neuroblasts. (A) Control. Scale bar is 100µm. (B) Small molecules only. Scale bar is 100µm. (C) NGN2-IPTD plus small molecules. Scale bar is 100µm. (D) High magnification image of (C). Scale bar is 10µm. Note that TUJ1 is weakly expressed in astrocytes in addition to neurons and can be regarded as background stain in these images.

### Astrocytic Plasticity Dramatically Improves Transdifferentiation of Neurons

Both early astrocytes and astrocytes found at the site of injury possess the ability to form multi-potent neurospheres, making them highly plastic (Yang et al., 2012a; Götz et al., 2015; Sirko et al., 2015; Wang et al., 2015; Michelucci et al., 2016). We investigated the extent to which our protocol might transdifferentiate plastic astrocytes by applying it to early passage (P3) astrocytes. Indeed, analysis by flow cytometry showed that efficiency was dramatically improved to 73.09 ± 5.83% (from 1.81 ± 1.04%) by priming and DAPT exposure alone (**Figures 2G**,**4**,**6**), while the further addition of ASCL-IPTD resulted in abundant neurosphere formation, giving rise to cells expressing the mature neuronal markers MAP2 (8.48 ± 2.29%, vs. 2.56 ± 0.47% without) and NEUN (2.20 ± 1.01% with ASCL1-IPTD vs. 0.47 ± 0.05% without) by day 12 (**Figures 5**, **6**). These neurons were seen by immunocytochemistry to express both glutamatergic and GABAergic neurotransmitter genes, suggesting a mix of excitatory and inhibitory interneurons, while treatment with small molecules alone were negative for neurotransmitter markers (**Figure 5**).

## DISCUSSION

In this work, we show that ASCL1-IPTD combined with the complementary action of SMAD and Notch signaling inhibition serves an efficient inducer of neurons from astrocytes, particularly those which exhibit plasticity. This effect is important since astrocytes become reactive at the site of injury in vivo and acquire enhanced plasticity, enabling them to proliferate and de-differentiate to generate multipotent neurospheres with the potential to self-renew and give rise to astrocytes, neurons or oligodendrocytes (Yang et al., 2012a; Götz et al., 2015; Sirko et al., 2015; Wang et al., 2015; Michelucci et al., 2016). SMAD signaling regulates genes involved in astrocyte commitment and maintenance (Stipursky and Gomes, 2007; Chambers et al., 2009; Yang et al., 2013; Zhang et al., 2015), therefore we treated cells with SMAD inhibitors from days 1 to 2 to disrupt astrocyte commitment. Our screening process identified DAPT to be the most potent enhancer of ASCL1-IPTD in transdifferentiating neurons from astrocytes. Notch signaling maintains neural progenitors in a state of proliferation by repressing genes responsible for cell cycle arrest and neural differentiation such as ASCL1 and NGN2 (Dhanesh et al., 2016), thus inhibition of Notch signaling is a necessary action for the generation of mature neurons.

A unique aspect of our transdifferentiation protocol is the presence of a progenitor stage characterized by neurosphere formation. Neurospheres are 3-D clusters of neural stem cells possessing the characteristics of self-renewal and multipotency (Bez et al., 2003). Since neurospheres were absent from cultures treated with small molecules alone we attribute the progenitor stage in our protocol to the action of ASCL1-IPTD. Furthermore, only neurons arising from neurospheres expressed mature neuronal markers MAP2 and NEUN by day 12. Passing through

(G) Cells are negative for the neurotransmitter tyrosine hydroxylase expressed by dopaminergic neurons TH, the neurotransmitter glutamic acid decarboxylase expressed by inhibitory interneurons GAD65/67, and the neurotransmitter choline acetyltransferase expressed by motor neurons CHAT. Scale bar is 100µm. (H) High magnification of (G). Scale bar is 30µm. Note that some fluorescence is picked up by all the cells producing background fluorescence which can be ignored.

a progenitor stage has two advantages. First, the rapid metabolic transition that takes place during the fate switch from somatic cell to neuron puts enormous stress on the cell, leading to the formation of reactive oxygen species (ROS), known to induce toxicity and affect cell fate regulation, posing a major barrier to transdifferentiation. It stands to reason that an intermediate stage would reduce this oxidative stress, promoting a safer transition between cell fates and improving efficiency (Gascón et al., 2016). Second, the generation of neural stem cells improves the efficiency of the protocol since each neural stem cell can produce

FIGURE 5 | Plastic astrocytes generate neurospheres and mature neurons after 12 days of exposure to ASCL1-IPTD, including 2 days of priming by LDN193189 and SB431542 followed by 10 days of DAPT. (A) Cells express the early neural marker TUJ1 and the neural stem cell marker SOX2. Scale bar is 100µm. (B) High magnification of (A). Scale bar is 30µm. (C) Cells express the mature neural marker MAP2. Scale bar is 100µm. (D) High magnification of (C). Scale bar is 30µm (E) Cells express the mature neural marker NEUN and the neurotransmitter glutamate expressed by excitatory interneurons GLUT. Scale bar is 100µm. (F) High magnification of (E). Scale bar is 30µm. (G) Cells are negative for the neurotransmitter tyrosine hydroxylase expressed by dopaminergic neurons TH, and negative for the neurotransmitter choline acetyltransferase expressed by motor neurons CHAT, but do express the neurotransmitter glutamic acid decarboxylase expressed by inhibitory interneurons GAD65/67. Scale bar is 100µm. (H) High magnification of (G). Scale bar is 30µm. Note that some fluorescence is picked up by all the cells producing background fluorescence which can be ignored.

several neurons. In vivo, the effect of neural stem cells may also be to prolong the activity of neural regeneration, providing neurotrophic support and newly formed neurons for a sustained period of time (Abe, 2000; Lu et al., 2003; Chen et al., 2017).

Our transdifferentiation process was highly efficient in comparison with protocols using viral ASCL1 for similar applications. For instance, viral delivery of 3 transcription factors, ASCL1, BYRN2, and MYT1L, is reported to transdifferentiate MEFs into neurons after 22 days with an efficiency of 15–19% (Vierbuchen et al., 2010), while viral ASCL1 alone is reported to also transdifferentiate human fibroblasts into neurons by day 22, but at lower efficiencies (Chanda et al., 2014). Astrocytes have proven easier to transdifferentiate into neurons, and viral ASCL1 delivered to postnatal mouse astrocytes is reported to convert 77% into neurons by day 21 (Liu et al., 2015), and when delivered to human astrocytes (Sciencell #1800) in combination with the pluripotency transcription factor Nanog homeobox (NANOG) is reported to produce 41% neurons by day 15 (Corti et al., 2012). By comparison, our protocol generated 69% neurons from human astrocytes (Sciencell #1800) by day 12.

Small molecule protocols are reported to generate neurons from astrocytes with similarly high efficiencies, but with considerable cell death in the process, ultimately reducing total efficiency. For instance, Gao et al. used a combination of 6 small molecules (valproic acid, CHIR99021, repsox, forskolin, i-Bet151, and ISX9) to transdifferentiate human adult astrocytes into neurons in 12 days, reporting an efficiency of 70% of surviving cells, or 8% of the initial cell population (Gao et al., 2017). In contrast, Zhang et al. showed that by adding fewer small molecules in a sequential manner, cell death can be avoided in small molecule protocols. In their protocol they used a combination of 9 small molecules (LDN193189, SB431542, TTNPB, Thiazovivin, CHIR99021, DAPT, valproic acid, SAG and purmorphamine) in a step-wise manner to transdifferentiate human astrocytes (ScienCell #1800) into neurons after 14 days, with an efficiency of 67% (Zhang et al., 2015). They used the priming small molecules LDN193189 and SB431542 for days 1-2 similar to our protocol. They found by removing small molecules one at a time that the Notch signal inhibitor DAPT was the most significant enhancer of transdifferentiation, followed closely by the GSK3 signal inhibitor CHIR99021. GSK3 signaling represses canonical Wnt signaling and NGN2 expression, both key regulators of neurogenesis (Hirabayashi et al., 2004; Li et al., 2012; Valvezan and Klein, 2012), thus GSK3 inhibition can be used to promote neurogenesis. Surprisingly, our screening data showed that while either CHIR99021 or DAPT were necessary to enhance transdifferentiation with ASCL1-IPTD, when used together actually decreased transdifferentiation by ∼50%. This suggests that although efficiencies are similar, the signaling pathways involved in our protocol differ substantially from the small molecule protocol.

Cooperation between ASCL1 and Notch signal inhibition can occur naturally in astrocytes in vivo after injury (Magnusson et al., 2014). Magnusson et al. identified Notch inhibition shortly after injury as a trigger for a latent neural transdifferentiaton program in resident astrocytes. The similarities between this latent program and the protocol we developed are striking. In their study, they observed a downregulation of Notch receptors and ligands in astrocytes followed by the appearance of ASCL1 expression after a stroke. These ASCL1-positive astrocytes clustered around the lesion and multiplied, giving rise to small, round DCX-positive neural stem cells by week 2, and eventually to mature NEUN-positive neurons by week 7, some of which expressed nitric oxide synthase (nNOS), a marker for GABAergic striatal interneurons. Based on these similarities, it is possible that our protocol takes advantage of this latent program in astrocytes by first recapitulating the plastic aspects of astrocytes at the site of an injury using SMAD inhibition, followed by the concerted action of ASCL1 upregulation by IPTD and Notch signal inhibition by DAPT. By this reasoning, it can be hypothesized that in vivo application of ASCL1-IPTD may be sufficient to trigger transdifferentiation of astrocytes at a site of injury into functional neurons. Although further work will be required to verify such a strategy, it would lend itself well to in vivo translation using a version of IPTD fitted with a domain to specifically target reactive astrocytes.

### CONCLUSION

Here we have shown a novel and non-viral protocol for efficient transdifferentiation of astrocytes into neurons. This protocol requires the delivery of only one transcription factor, ASCL1, by a novel PTD, and 4 small molecules, the SMAD inhibitors LDN193189 and SB431542, the Notch inhibitor DAPT, and the histone deacetylase inhibitor valproic acid. The efficiency of this protocol is enhanced significantly by the natural plasticity of astrocytes, suggesting its suitability for application in vivo where injuries lead to increased plasticity in reactive astrocytes. Future work will determine if the small molecules are necessary in vivo, and incorporate an astrocyte-targeting motif in the IPTD design. Further work will also adapt this protocol for producing specific neuronal subtypes through cooperation of ASCL1-IPTD with other IPTD-functionalized transcription factors.

### REFERENCES


### AUTHOR CONTRIBUTIONS

MR designed the experiments, analyzed results, and wrote the manuscript. IF assisted in background research, guidance on experimental design and analysis of the small molecule screening experiment. KS assisted in executing the ASCL-IPTD experiments on astrocytes. EM assisted in executing and analysing the NGN2-IPTD experiments on astrocytes. LC assisted in executing the ASCL1-IPTD experiments on MEFs and providing background research. SW provided experimental guidance, provided help writing and editing the final manuscript.

## FUNDING

Funding for this project was provided by a Collaborative Research and Development grant (CRD) and a Discovery Grant (DG) by the Natural Sciences and Engineering Research Council (NSERC), along with the Canada Research Chair (CRC) program.

### ACKNOWLEDGMENTS

We would like to acknowledge NSERC and the CRC for funding this project, and iProgen Biotech for their support by providing the functionalized transcription factors.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbioe. 2018.00173/full#supplementary-material

Supplemental Figure 1 | Mouse embryonic fibroblasts (MEFs) express the skeletal myoblast gene myosin heavy chain 3 (MYH3) and lack the neuronal gene beta tubulin (TUJ1) after 14 days of exposure to ASCL1-IPTD and the cAMP activator forskolin, indicating that they have adopted a myogenic cell fate rather than a neuronal cell fate. (A) Exposure to 3µg/mL ASCL1-IPTD is insufficient for transdifferentiation. Scale bar is 100µm. (B) Addition of forskolin results in transdifferentiation into myoblasts rather than neurons. Scale bar is 100µm. (C) Increasing the concentration of ASCL1-IPTD 5-fold still generates myoblasts rather than neurons. Scale bar is 100µm. (D) Increasing the concentration of ASCL1-IPTD 5-fold and adding forskolin still generates myoblasts rather than neurons. Scale bar is 100µm (E) Control. Scale bar is 100µm (F) Forskolin only control. Scale bar is 100µm.

functional neurons under defined conditions. Cell Stem Cell 9, 113–118. doi: 10.1016/j.stem.2011.07.002


**Conflict of Interest Statement:** This research was in part funded by an NSERC-CRD grant which we hold with iProgen Biotech who owns the PTD technology. This work was conducted in collaboration with iProgen Biotech who provided the functionalized transcription factors used in this work.

Copyright © 2018 Robinson, Fraser, McKee, Scheck, Chang and Willerth. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.