P. endodontalis ATCC 35406 (Argenta, Poznań, Poland) and P. gingivalis A7436 (laboratory collection) strains were grown on Schaedler blood agar (ABA) plates (Argenta) at 37°C for 6 and 4 days, respectively, under anaerobic conditions (80% N₂, 10% H₂ and 10% CO₂) (Whitley A35 anaerobic workstation; Bingley, UK). Then, bacteria were used to inoculate a liquid basal medium (BM) composed of 3% trypticase soy broth (Becton Dickinson, Sparks, MD, USA) and 0.5% yeast extract (Biomaxima, Lublin, Poland), supplemented with 0.05% L-cysteine (Carl Roth, Karlsruhe, Germany), 0.5 mg/l menadione (Sigma-Aldrich, St. Louis, MO, USA) (BM medium), and 7.7 µM hemin chloride (Fluka, Munich, Germany) to allow bacteria to grow in optimal, iron- and heme-rich conditions (Hm medium). Alternatively, to starve bacteria of iron and heme, no heme source was added and the medium was supplemented with 160 µM 2,2-dipyridyl (Sigma-Aldrich) to complex free iron (DIP medium). The optical density at 600 nm (OD₆₀₀) at the beginning of the liquid culture was at least 0.5 and 0.2 for P. endodontalis and P. gingivalis, respectively.

To analyze growth curves, bacteria were cultured for two passages in BM medium without the addition of a heme source. BM medium (BM alone) or medium supplemented with 1.25 or 5 µM human hemoglobin (Hb), BM medium supplemented with 5 µM heme and 5 µM human serum albumin (HSA), 5 µM HmuY^Pg, 5 µM HmuY^Pe, or 5 µM HmuY^Tf were inoculated with P. gingivalis or P. endodontalis at starting OD₆₀₀ equal to 0.2 or 0.5, respectively. Media supplemented with proteins and heme were preincubated for 16 hours at 4°C before use, to allow saturation of proteins with heme.

Escherichia coli ER2566 strain (New England Biolabs, Ipswich, MA, USA) was grown under standard aerobic conditions.

The recombinant HmuY^Pe, lacking the predicted signal peptide (MKTRFFLALIATSLVLGVASCRP), was overexpressed and purified using affinity chromatography. Briefly, the hmuY^Pe gene from P. endodontalis (GenBank accession number: POREN0001_0444) was PCR amplified using primers listed in Supplementary Table S1 and cloned into XcmI and BamHI restriction sites of a pTriEx-4 plasmid (Sigma-Aldrich) using NEBuilder HiFi DNA Assembly (New England Biolabs), resulting in the plasmid encoding HmuY^Pe protein with an N-terminal 6×His tag and the site recognized by Factor Xa.

To generate plasmids encoding HmuY^Pe variants with amino acid substitutions, the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA) and primers listed in Supplementary Table S1 were used.

HmuY^Pe protein and its site-directed mutagenesis variants were overexpressed in E. coli after induction with 0.5 mM isopropyl-ß-D-thiogalacto-pyranoside (IPTG; Carl-Roth) at 16°C for 16 hours. Proteins were purified using the soluble fraction of E. coli cell lysates and TALON Superflow resin according to the manufacturer’s instructions (Sigma-Aldrich), using 25 mM HEPES buffer, pH 7.8, supplemented with 300 mM NaCl. For the elution step, 25 mM Tris/HCl buffer, pH 7.6, supplemented with 80 mM NaCl and 150 mM imidazole (Carl-Roth) was used. To obtain un-tagged proteins, the buffer was exchanged for 25 mM Tris/HCl, pH 7.6, supplemented with 80 mM NaCl. The 6×His tag was removed using Factor Xa (New England Biolabs) and filtration through an Amicon Ultra-4 Centrifugal Ultracel-10KDa filter unit (Millipore). If necessary, another step of affinity chromatography with nickel-immobilized resin (Ni-NTA; New England Biolabs) was used to remove the 6×His tag before protein concentration ( Supplementary Figure S2 ).

HmuY^Pg protein from P. gingivalis and HmuY^Tf protein from T. forsythia were overexpressed and purified as described previously (Slezak et al., 2020; Śmiga et al., 2023a). To determine protein concentration, empirical molar absorption coefficients for HmuY^Pg (ϵ₂₈₀ = 36.86 mM⁻¹ cm⁻¹) (Wojtowicz et al., 2009b) and HmuY^Tf (ϵ₂₈₀ = 26.32 mM⁻¹ cm⁻¹) (Bielecki et al., 2018) were used. For P. endodontalis HmuY^Pe, the empirical molar absorption coefficient was determined in this study (ϵ₂₈₀ = 35.56 mM⁻¹ cm⁻¹), as described by others (Eakanunkul et al., 2005).

Protein samples, P. endodontalis and P. gingivalis whole cell lysates, or samples prepared from the whole bacterial cultures were analyzed by SDS-PAGE. Samples were separated on 12% polyacrylamide gels and the proteins were visualized with Coomassie Brilliant Blue G-250 (CBB G-250) or were transferred onto nitrocellulose membranes (Millipore, Billerica, MA, USA). 20 or 100 ng of HmuY^Pe, HmuY^Pg, or HmuY^Tf proteins in 5 µl were applied onto nitrocellulose membranes for dot blotting. Western blotting and dot blotting were performed as described before (Śmiga et al., 2015; Śmiga et al., 2023a). Briefly, membranes were incubated with rabbit anti-HmuY^Pg or rabbit anti-HmuY^Tf polyclonal antibodies (1:10,000; GenScript USA Inc.). Subsequently, goat anti-rabbit IgG antibodies conjugated with horseradish peroxidase (1:10,000; Sigma-Aldrich) were applied. Chemiluminescence staining (Perkin Elmer, Waltham, MA, USA) and ChemiDoc Imaging System (Bio-Rad Laboratories, Hercules, CA, USA) were used to visualize proteins.

Analysis of heme binding was performed as described previously (Śmiga et al., 2023b). Briefly, ~8 mg of hemin chloride (Pol-Aura, Morąg, Poland) was dissolved in 0.1 M NaOH, and its concentration was determined using empirical molar absorption coefficient (ϵ₃₈₅ = 58.5 mM⁻¹ cm⁻¹). Proteins at 5 µM concentration were prepared in 20 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl (PBS). Protein samples were titrated with heme, and protein-heme complexes were monitored by UV-visible absorbance (250–700 nm) spectroscopy with a double-beam Jasco V-750 spectrophotometer (Jasco GmbH, Pfungstadt, Germany). The reduced conditions were formed by the addition of sodium dithionite (Sigma-Aldrich) to the final 10 mM concentration and mineral oil overlay of the sample (Sigma-Aldrich).

Heme sequestration was examined by mixing holo- (protein-heme complex) and apo-protein (protein alone) in PBS and monitoring heme transfer by UV-visible absorbance spectroscopy or PAGE, as described before (Smalley et al., 2011; Śmiga et al., 2023b). Except for methemoglobin (metHb; Sigma-Aldrich), protein-heme complexes were prepared by mixing protein and heme at a 1:1.2 molar ratio and incubation at room temperature for 1 hour. To remove unbound heme, the solutions were passed through Zeba Spin desalting columns (Thermo Fisher, Scientific, Waltham, MA, USA).

1.25 µM metHb or 5 µM other holo-proteins were mixed with 5 µM apo-proteins. The heme sequestration process was monitored over time by recording UV-visible absorbance spectra under oxidizing and reducing conditions. Alternatively, 20 µM holo-HmuY^Tf protein or 10 µM other holo-proteins were mixed with 10 µM apo-proteins and incubated for 30 minutes at 37°C. To 30 µl of the sample, 10 µl of 0.4 M Tris/HCl buffer, pH 6.8, supplemented with 40% glycerol and 0.08% bromophenol blue was added, and 25 µl of the sample was immediately loaded on the 13.5% PAGE-separating gel prepared without SDS. After electrophoresis, the heme-containing complexes were first stained using 5 mM 3,3′,5,5′-tetramethylbenzidine (TMB) prepared in 100 mM Tris/HCl buffer, pH 7.5, supplemented with 140 mM NaCl and 0.05% H₂O₂ (TMB-H₂O₂ staining) up to 30 minutes. Subsequently, all proteins were stained with CBB G-250.

To determine the influence of iron and heme on P. endodontalis gene expression, bacteria were cultured in Hm medium or DIP medium for one or two 24-hour passages. Bacteria from 1 ml of culture were centrifuged and used for RNA isolation with the Total RNA Mini Kit (A&A Biotechnology, Gdańsk, Poland). Genomic DNA contamination was removed using the Clean-up RNA concentrator Kit (A&A Biotechnology). RNA was used to generate cDNA with a LunaScript RT SuperMix Kit (New England Biolabs). qPCR was performed using SensiFAST SYBR no-ROX Kit (Bioline, London, UK) and LightCycler 96 (Roche, Basel, Switzerland). PCR program comprised initial denaturation at 95°C for 120 seconds, 35 cycles of denaturation at 95°C for 5 seconds, primers annealing at 60°C for 10 seconds, and extension at 72°C for 15 seconds. After PCR, the melting curves were generated for PCR quality control. Relative change in gene expression was calculated using LightCycler 96 software (Roche) and 16S rRNA as a reference gene. All analyses were carried out in 4 biological repetitions. All primers used are listed in Supplementary Table S1 .

The total proteolytic activity of whole bacterial cultures was measured using azocasein (Sigma-Aldrich) as a substrate. Briefly, to 40 μl of 1.5% azocasein solution in reaction buffer (20 mM Tris/HCl buffer, pH 7.5, supplemented with 150 mM NaCl, 5 mM CaCl₂, 0.05% Tween 20, and 10 mM L-cysteine hydrochloride freshly neutralized with NaOH), 10 μl of total P. endodontalis or P. gingivalis overnight culture (grown for 24 hours) was added. Samples were incubated for 30 minutes at 37°C, and the reaction was stopped by adding 200 μl of 5% TCA (Sigma-Aldrich). Precipitated, undigested azocasein was separated by centrifugation (2000×g, 10 min). 100 μl of the supernatant was mixed with 60 μl of 0.5 M NaOH and the absorbance at 450 nm (A₄₅₀) was measured using a GloMax Discover plate reader (Promega, Madison, WI, USA). The final results were presented as a change of A₄₅₀ over 60 minutes caused by 1 ml of culture exhibiting OD₆₀₀ equal to 1 (ΔA₄₅₀/60 min/ml).

HSA (Sigma-Aldrich), metHb (Sigma-Aldrich), hemopexin (Hpx; Sigma-Aldrich), P. gingivalis HmuY^Pg, P. endodontalis HmuY^Pe, and T. forsythia HmuY^Tf were used to analyze their susceptibility to proteolysis performed by proteases produced by P. endodontalis. Briefly, Hm medium was supplemented with 2 µM proteins and the bacterial cultures were started with the initial OD₆₀₀ equal to 0.5. Samples were collected over time and examined using SDS-PAGE and staining with CBB-G250. As a control, the fresh BM+Hm medium or P. gingivalis culture started at OD₆₀₀ equal to 0.2 was used.

All experiments were performed independently at least in two biological replicates and at least in three technical repetitions each. The numerical values are represented as mean ± standard deviation (mean ± SD) or mean ± standard error (mean ± SE). All statistical analyses were done using unpaired Student’s t-test with GraphPad software (GraphPad Prism 8.0 Inc., San Diego, CA, USA).

Search for protein sequences was performed in the GenBank database using PSI-BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Protein amino acid sequences were compared with The Clustal Omega (Madeira et al., 2022) and Jalview (Waterhouse et al., 2009). Protein similarity and identity were analyzed using Sequence Manipulation Suite: Ident and Sim (Stothard, 2000). Solved (RCSB Protein Data Bank; https://www.rcsb.pdb) and predicted with AlphaFold Protein Structure Database (Jumper et al., 2021; Varadi et al., 2022) three-dimensional protein structures were visualized with UCSF Chimera (Pettersen et al., 2004). Heme binding in HmuY^Pe was predicted with EDock (Zhang et al., 2020). Prediction of P. endodontalis hmu operon genes was performed using an Operon mapper (https://biocomputo.ibt.unam.mx/operon_mapper/).

P. endodontalis ATCC 35406 hmu operon organization is similar to that found in P. gingivalis, including one HmuY homolog ( Figure 1A ). HmuY^Pe protein is closely related to HmuY^Pg with ~50% amino acid sequence identity, as well as 21–22% identity and 32% similarity to other characterized HmuY proteins produced by human oral pathogens from the Bacteroidota phylum ( Figures 1B, C ). Other proteins of the Hmu system in P. endodontalis show high homology to those found in P. gingivalis with amino acid sequence identity as follows: HmuR ~49%, HmuS ~ 65%, HmuT ~50%, HmuU ~68%, and HmuV ~79%. Interestingly, in P. endodontalis an additional gene (GenBank accession number: POREN0001_0445) is located upstream of the hmuY^Pe gene, encoding a protein containing T9SS type A sorting domain (T9SS) ( Figure 1A ). Using theoretical and experimental approaches, we confirmed that the POREN0001_0445 gene is a part of the hmu operon ( Supplementary Figure S1A ). Proteins homologous to the protein encoded by the POREN0001_0445 gene are found in other bacteria including some Porphyromonas species and P. intermedia. However, in contrast to P. endodontalis, they are not encoded as a part of operons but as orphan genes (data not shown). Despite the low amino acid sequence identity between analyzed homologs of the POREN0001_0445 protein ( Supplementary Figure S1B ), their predicted tertiary structure is highly similar ( Supplementary Figure S1C ). They form a β-barrel-like structure, which may suggest their role in transporting an unknown molecule through the outer membrane.

Analysis of the overall three-dimensional experimentally solved (P. gingivalis HmuY^Pg and T. forsythia HmuY^Tf) or theoretically predicted (other selected Porphyromonas species) protein structures revealed that HmuY proteins identified in the Porphyromonas species are highly similar to the P. gingivalis HmuY^Pg ( Figure 1D ), with the highest similarities observed in the core region (Śmiga et al., 2023a). As in so far characterized HmuY proteins (Olczak et al., 2024), the main differences in HmuY proteins identified in Porphyromonas species are visible in the structure of heme-binding pockets and differ mainly in the size of the entrance of the heme-binding pocket ( Figure 1D ). The data shown in Figure 1 allowed us to verify the spatial location of predicted amino acids involved in heme binding in HmuY homologs in analyzed Porphyromonas species. To date, the characterization of HmuY proteins has included the HmuY^Pg protein from the P. gingivalis, which uses two histidines for heme-iron coordination, and HmuY homologs which use two methionines in this process (for example T. forsythia HmuY^Tf, P. intermedia HmuY^Pi-1 and HmuY^Pi-2). Our theoretical analyses showed that the majority of HmuY proteins identified in Porphyromonas species most likely coordinate heme-iron using a histidine-methionine pair or two methionines ( Figures 1C, D ). Only proteins closely related to the HmuY^Pg, namely HmuY^Pgu from P. gulae and HmuY^Plo from P. loveana may coordinate heme iron with two histidines ( Figures 1C, D ).

To confirm the heme binding ability of the HmuY^Pe, the protein was overexpressed in E. coli and purified using chromatographic methods ( Supplementary Figures S2A , S2B ). Purified and concentrated protein samples exhibited a reddish color (data not shown) similar to the HmuY^Pg sample (Bielecki et al., 2018, 2020). The UV-visible spectroscopic analysis demonstrated that under both oxidizing and reducing conditions, the HmuY^Pe sample exhibited spectra similar to those of the purified HmuY^Pg sample ( Figures 2A, B ), which indicated heme binding. To confirm this, the HmuY^Pe protein was saturated with heme [Fe(III)heme], and heme excess was removed by desalting. A UV-Vis absorbance spectrum of the sample was characterized by maxima in both the Soret band (414 nm) and Q bands (530 nm and 559 nm) ( Figure 2C ). Reduction of heme in this sample [Fe(II)heme] resulted in a red shift in the Soret band (425 nm), and a shift of the Q bands to 528 nm and 558 nm, which became more intense and resolved ( Figure 2D ). These spectra were similar to those obtained for the HmuY^Pg-heme complex ( Figures 2E, F ).

Heme binding strength was analyzed by determination of the HmuY^Pe-heme complex dissociation constant (K_d ) ( Figure 3A ) by titration of apo-protein with increasing concentrations of heme ( Supplementary Figure S3 ). K_d of HmuY^Pe-heme complex was lower under reducing (4.10 × 10^-8 M) than oxidizing (2.45 × 10^-7 M) ( Figure 3A ) conditions, thus being similar to HmuY homologs coordinating heme-iron with two methionine residues, e.g., T. forsythia HmuY^Tf, than to P. gingivalis HmuY^Pg (Bielecki et al., 2018).

To experimentally confirm heme-iron coordinating amino acids, we prepared HmuY^Pe site-directed mutagenesis protein variants with selected histidines or methionines replaced by an alanine (M123A, H128A, H132A, and M163A), chosen based on the comparative analysis of amino acid sequences ( Figure 1C ). Particular single mutagenesis variants were overexpressed in E. coli and purified at similar levels as the unmodified protein ( Supplementary Figure S2C ). All protein variants were stable after purification, which suggests that amino acid replacements did not influence the tertiary protein structure. UV-visible absorbance spectra of HmuY^Pe-heme complexes showed changes mostly in the case of H128A and M163A protein variants, suggesting that these amino acids could be engaged in heme-iron coordination ( Figure 3B ). The determination of K_d showed that the H128A and M163A variants are characterized by a lower affinity for heme than an unmodified protein ( Figure 3A ). A slightly lower affinity of heme binding under reducing conditions in the case of the M123A variant ( Figure 3A ) may suggest local structural changes in the loop engaged in heme-iron coordination and/or supportive role in heme binding as it was shown for HmuY^Pg M136 (Wojtowicz et al., 2009b; Bielecki et al., 2018). This finding was confirmed by our theoretical analysis using the modeled HmuY^Pe structure and its comparison to the experimentally solved HmuY^Pg-heme structure ( Figure 3C ).

To compare the strength of heme binding to the HmuY^Pe with other HmuY proteins, we used a competitive analysis with HmuY^Pg and HmuY^Tf. For this purpose, we employed UV-visible spectroscopy and native electrophoresis (PAGE) ( Figure 4 ). HmuY^Pe protein was unable to sequester heme complexed with HmuY^Pg ( Figures 4A–C ). At the same time, HmuY^Pg was able to capture heme from the HmuY^Pe-heme complex under both oxidizing and reducing conditions ( Figures 4D–F ). In the case of HmuY^Tf, we observed efficient heme sequestration of heme from the HmuY^Tf-heme complex by HmuY^Pe under oxidizing and reducing conditions, while HmuY^Tf was not able to capture heme from the HmuY^Pe complex ( Figures 4G–L ). In contrast to the spectroscopic method, we could visualize neither HmuY^Pe-heme nor HmuY^Tf-heme complexes with PAGE because they were hardly visible after TMB-H₂O₂ staining ( Figures 4A, D, G, J ). This effect could be caused by oxidizing conditions applied in this experiment, resulting in weak heme binding or heme release.

Further, we analyzed the ability of HmuY^Pe to directly sequester heme from host hemoproteins, which serve in vivo as the main heme source. Under oxidizing conditions, the affinity of HSA and HmuY^Pe to heme was similar because while using HSA-heme or HmuY^Pe-heme complex with apo-form of the counterpart, we observed an equilibrium in heme binding between the proteins ( Figures 5A–F ). Although PAGE results are not clear ( Figures 5A, D ), using the spectroscopic method we observed the spectrum shift in two experimental settings ( Figures 5B, E ). Under reducing conditions, HmuY^Pe was able to capture heme bound to HSA, whereas apo-HSA could not sequester heme from HmuY^Pe ( Figures 5C, F ). In the case of Hpx, the results are inconclusive due to the low quality of HmuY^Pe TMB-H₂O₂ staining ( Figures 5G, J ). In the case of these proteins, equilibrium in Hpx-heme and HmuY^Pe-heme complexes could also occur, which was confirmed by changes in the intensity of the analyzed UV-visible absorbance spectra over time ( Figures 5H, I, K, L ). However, since the absorbance maxima of both proteins in complex with heme are similar, we cannot draw firm conclusions. In contrast to HmuY^Pg (Smalley et al., 2011; Byrne et al., 2013; Śmiga et al., 2021), HmuY^Pe was unable to capture the heme associated with metHb ( Supplementary Figures S4A , S4B ). Moreover, heme sequestration was not observed after the chemical reduction of metHb ( Supplementary Figure S4C ). As a control, we used HmuY^Pg, which can directly sequester heme complexed with HSA ( Figure 5A ; Supplementary Figure S5 ) (Smalley et al., 2011; Sieminska et al., 2021), Hpx ( Figure 5G ) (Bielecki et al., 2018), and metHb ( Supplementary Figure S4A ) (Smalley et al., 2011).

P. gingivalis can use a variety of heme sources thanks to the direct sequestration of heme from hemoproteins by HmuY^Pg or its synergistic cooperation with highly active proteases – gingipains (Smalley and Olczak, 2017; Śmiga et al., 2023b; Olczak et al., 2024). Although P. endodontalis does not encode gingipain homologs, it produces proteases, albeit less active compared to P. gingivalis proteases ( Figure 6A ). Nevertheless, they can efficiently degrade metHb in the P. endodontalis culture ( Figure 6B ). None of the additionally analyzed proteins were degraded by P. endodontalis proteases, including human hemoproteins and representatives of HmuY homologs produced by other human pathogens ( Figure 6B ). These results are consistent with P. endodontalis growing in the planktonic form with different heme sources, the best achieved in the presence of free heme or metHb ( Supplementary Figure S6 ; Figure 6C ).

Similar to other HmuY proteins that were characterized previously (Bielecki et al., 2018, 2020; Sieminska et al., 2021; Antonyuk et al., 2023), transcript encoding HmuY^Pe was produced at higher levels when bacteria were grown in iron- and heme-depleted conditions, formed as shown in Figure 7A . However, the increase of the hmuY^Pe gene expression was only up to 6 times, whereas for hmuY^Pg the expression increased up to several hundred times (Bielecki et al., 2018, 2020), resulting in a significant increase in produced HmuY^Pg protein ( Figure 7B ) (Olczak et al., 2010). This dissimilarity may result from weaker growth of P. endodontalis under laboratory conditions, especially under iron and heme starvation ( Supplementary Figure S6 ).

Finally, we aimed to find whether, despite 50% amino acid sequence identity ( Figure 1 ) and almost identical tertiary structures of HmuY^Pe and HmuY^Pg ( Figure 3C ), HmuY^Pe could potentially serve as a specific marker for P. endodontalis. We analyzed the purified HmuY^Pe protein, both in the form of native and denatured protein and showed cross-reactivity neither with IgG antibodies raised against HmuY^Pg nor HmuY^Tf ( Supplementary Figure S7 ).