Unraveling the Dichotomy of Enigmatic Serine Protease HtrA2

Mitochondrial high-temperature requirement protease A2 (HtrA2) is an integral member of the HtrA family of serine proteases that are evolutionarily conserved from prokaryotes to humans. Involvement in manifold intricate cellular networks and diverse pathophysiological functions make HtrA2 the most enigmatic moonlighting protease amongst the human HtrAs. Despite perpetuating the oligomeric architecture and overall structural fold of its homologs that comprises serine protease and regulatory PDZ domains, subtle conformational alterations and dynamic enzymatic regulation through the distinct allosteric mode of action lead to its functional diversity. This mitochondrial protease upon maturation, exposes its one-of-a-kind N-terminal tetrapeptide (AVPS) motif that binds and subsequently cleaves Inhibitor of Apoptosis Proteins (IAPs) thus promoting cell death, and posing as an important molecule for therapeutic intervention. Interestingly, unlike its other human counterparts, HtrA2 has also been implicated in maintaining the mitochondrial integrity through a bi-functional chaperone-protease activity, the on-off switch of which is yet to be identified. Furthermore, its ability to activate a wide repertoire of substrates through both its N- and C-terminal regions presumably has calibrated its association with several cellular pathways and hence diseases including neurodegenerative disorders and cancer. Therefore, the exclusive structural attributes of HtrA2 that involve multimodal activation, intermolecular PDZ-protease crosstalk, and an allosterically-modulated trimeric active-site ensemble have enabled the protease to evolve across species and partake functions that are fine-tuned for maintaining cellular homeostasis and mitochondrial proteome quality control in humans. These unique features along with its multitasking potential make HtrA2 a promising therapeutic target both in cancer and neurodegeneration.

signal sequences, insulin-like growth factor-binding domains, and serine protease inhibitor domains (human HtrA1, HtrA3, and HtrA4) implicating intra-and inter-species functional divergence. Furthermore, their catalytic activity that can be allosterically tuned through an intricate rheostatic on/off switch as well as the modulatory protein-protein interaction domain(s) aka PDZ, has garnered much attention for their immense translational possibilities.
Interestingly, unlike eukaryotes and bacteria, archaean genomes are devoid of HtrA homologs (Koonin and Aravind, 2002). Although, all sequenced Nematoda genomes including the model organism Caenorhabditis elegans lack HtrA-like genes, they do encode PDZ-containing proteins (Koonin and Aravind, 2002) thus underscoring the functional relevance of this regulatory domain in various cellular pathways. While bacterial HtrAs have been demonstrated to be involved in protein quality control processes such as protein folding, stress response, and degradation of misfolded cell envelope proteins , this function is manifested in their mammalian counterparts through the elimination of misfolded proteins including growth factors, regulation of cell proliferation, migration and apoptosis (Grau et al., 2005;Hou et al., 2005;Kapri-Pardes et al., 2007;Moisoi et al., 2009).
Among the four human HtrAs (HtrA1-4) that have been identified to date, HtrA2 has been most widely studied due to its enigmatic structural characteristics and profound functional relevance. While HtrA2 is found in the mitochondrial intermembrane space (IMS), its paralogs HtrA1, 3, and 4 are mostly found in the secretory process. Despite similar overall structural signature and conserved protease and PDZ domain architecture, these enzymes show a significant divergence in their N-terminal regions that might be essential for catering to their distinct functional properties For example, the N-terminal regions of HtrA1, 3, and 4 include secretory signals, along with insulin-like growth factor binding motifs and Kazal-type S protease inhibitor domains, while HtrA2 contains a mitochondrial localization signal ( Figures 1A,B).
HtrA2, with a pyramid-shaped trimeric ensemble, is unique among its peers being the only known mitochondrial protease with a PDZ domain that identifies exposed hydrophobic regions of misfolded proteins (Li et al., 2002;Clausen et al., 2011;Singh et al., 2011). Furthermore, with the triggering of apoptotic signal, mature HtrA2 gets released from the mitochondrial IMS into the cytosol at the expense of its first 133 amino acid residues ( Figure 1B). This series of events exposes an N-terminal tetrapeptide motif (AVPS) that binds to the Inhibitor of Apoptosis Proteins (IAPs) and abate their inhibition on caspases thus promoting apoptosis. Furthermore, HtrA2 is known to participate in apoptosis through both caspasedependent and independent pathways, the latter through its serine protease activity (Hegde et al., 2002;Martins et al., 2002;Verhagen et al., 2002). Apart from its prominent role as a proapoptotic molecule, its involvement in neurodegenerative disorders has also been established through a missense mutation (Ser276Cys) in transgenic mice that exhibited motor neuron degeneration 2 (mnd2) implicating a Parkinsonian phenotype in humans (Jones et al., 2003). Further functional and clinical studies established HtrA2's involvement in several neurodegenerative disorders (Inagaki et al., 2008;Kang et al., 2013;Wagh and Bose, 2018;Bose et al., 2021).

STRUCTURAL FEATURES OF HtrA2
Several efforts over the past decade have been made to capture the structural complexity of this proapoptotic enzyme from various perspectives. Shi and co-workers first provided the snapshots of the inactive (S306A) substrate-unbound form of mature HtrA2 in three-dimensional space (Li et al., 2002). The structural data showcased a trimeric pyramidal architecture with the short N-terminal regions upholding the oligomeric ensemble through van der Waals interactions, while three PDZ domains at the base encapsulated the active-sites of the protease domains. The protease domain that embeds a hydrophobic active-site pocket with the catalytic triad (Ser306, His198, and Asp228) forms a compact structural fold comprising seven α-helices and 19 β-strands. Surrounded by several regulatory and specificity loops, this domain is positioned deep within the oligomer at 25 Å above the base of the pyramid ( Figure 1C) suggesting the requirement of substantial conformational changes for substrate binding and subsequent cleavage. The core of the pyramid is flanked by the regulatory PDZ domains that recognize and bind to the C-terminal region of their interacting partners. This is achieved through the canonical PDZ binding groove (YIGV) that is integrated into the PDZprotease domain interface. The structural study also demonstrates that several non-covalent interactions in the substrate-unbound state keep the protease domain in its 'closed' conformation, through inhibitory interference from the surrounding PDZ domains.
Although, this structure provided an excellent overview of the HtrA2 structure, this substrate-unbound form of the protease failed to explain the underlying dynamics of its mode of activation. Most importantly, the model's inability to enumerate the necessity to have a trimeric structure for its enzymatic functions as well as the mode of its distal allosteric regulation, impelled scientists to unravel the minutiae of its interactions from a more physiological as well as quantitative perspectives.

ACTIVE SITE CONFORMATION AND MULTIPLE ACTIVATION MECHANISMS OF HtrA2
The pre-defined conserved domains of HtrA2, along with its regulatory (L1, L3, and LD) and specificity (L3-that accommodates specificity pocket) loops contribute to the activation mechanism of HtrA2 through multiple regulatory nodes ( Figure 1C). Since these dynamic loops were mostly unresolved in the crystal structure, several efforts were made to investigate the multimodal allosteric regulation of the protease as well as understand the intricacies of HtrA2-mediated substrate cleavage (Martins et al., 2003;Jarzab et al., 2016). Because the Frontiers in Molecular Biosciences | www.frontiersin.org February 2022 | Volume 9 | Article 824846 allosteric binding partners are also predominantly its substrates (such as IAPS, GRIM-19, and Dusp-9), therefore the stepwise concerted allosteric mechanism either individually or in collaboration with different activation pathways could not be unequivocally determined using discrete peptide libraries. To circumvent the problem, Bose and co-workers utilized enzymology and biophysical approaches to understand the intricate coordination between the protease domain and other regions of the protein using full-length binding partners and/or substrates. Using β-casein, the generic substrate of serine proteases, Chaganti et al., revisited the pre-existing model of HtrA2 activation and propounded a new hypothesis that relies on inter-molecular protease-PDZ crosstalk for initial substrate binding at the PDZ domain and its subsequent cleavage (Chaganti et al., 2013). This study identified interaction between the PDZ domain of one monomer with the serine protease domain of an adjacent one, which led to the rearrangement of H65 of the catalytic triad in a way to form a proper oxyanion hole. This series of inter-molecular making and breaking of bonds unequivocally demonstrated the requirement of the trimeric architecture for its allosteric propagation and activation by capturing the dynamics of the PDZ-and temperature-mediated activation process. Singh et al., built upon the previous studies on N-terminal mediated activation of HtrA2 (Verhagen et al., 2002) and described the global conformational plasticity and subtle conformational reorientations in the loop regions surrounding the active-site to be involved in this process. Interestingly, using quantitative enzyme kinetics studies, they further demonstrated that the N-terminal mediated activation might also be regulated by PDZ-bound allosteric modulators and vice-versa (Singh et al., 2011;Singh et al., 2014) to bring the protease to the most competent catalytic state.
Although these studies provided a holistic understanding of HtrA's mode of activation through three distinct yet non-exclusive modes, they did not provide the stoichiometric contribution of the PDZ-protease communication in a step-by-step manner. Using molecular dynamics, protein engineering, structural and chemical biology approaches, two different groups Toyama et al., 2021) distinctly established the trans-mediated Upon apoptotic trigger, the mitochondrial localization signal (133 residues) from the N-terminus gets cleaved exposing a tetrapeptide IAP-binding motif (IBM/AVPS), and concomitantly releasing the protease into the cytosol. Subsequent substrate binding at N-and/or C-termini leads to allosteric protease activation and substrate cleavage as described in the text. (C) The three-dimensional trimeric model adopted from the crystal structure (PDB ID: 1LCY) of HtrA2 highlighting the hidden catalytic triad (rainbow spheres) 25 Å above base of the pyramid (left side) while a single monomer has been zoomed into for describing the loops (yellow) and domains (N-terminal region: light purple, SPD: pink, PDZ: orange); the catalytic site has been shown in the inset (right side). L1, L2, L3 and LD are loops; SPD is serine protease domain and Linker represents the flexible region between the SPD and PDZ domain.
Frontiers in Molecular Biosciences | www.frontiersin.org February 2022 | Volume 9 | Article 824846 PDZ-protease collaboration that espouses a unique reciprocative mechanism where the distal PDZ reorients the active site of the adjacent monomer and attunes it for catalysis through a precise synergistic relay of information. This multitiered regulation of HtrA2 activation might be critical toward prevention of untimely proteolysis as well as accurately controlling its involvement in different pathophysiological pathways such as apoptosis, protein quality-control, cancer, arthritis, and neurodegeneration, where it cleaves a wide spectrum of substrates in different subcellular locations. This is substantiated by the identification and characterization of protein-protein interactions involving HtrA2 and its substrates such as Inhibitor of Apoptosis Proteins (IAPs), hematopoietic cell-specific protein-1-(HS1)-associated protein X-1 (Hax-1), Dual-specificity phosphatase-9 (DUSP-9), a gene associated with retinoic and interferon-induced mortality-19 protein  and Phosphoprotein enriched in astrocytes-15 (Pea-15) (Chaganti et al., 2019;Acharya et al., 2020;Kummari et al., 2021) that unlike other HtrAs are interestingly not restricted to the C-terminal PDZ domains. The holistic enumeration of HtrA2's activation network has been vividly illustrated in Figure 2 and the mechanism is elaborated in the figure legend. This chain of ground-breaking revelations on the reciprocity of its structural dynamism and multifarious physiological as well as disease-associated functions as discussed below have opened up avenues to regulate HtrA2 functions at various check-points toward devising customized therapeutic strategies.

IS HtrA2 A CHAPERONE?
The neurodegenerative phenotype of mice lacking HtrA2 or harboring the enzymatically inactive mnd2mutant (S276C) . This conformation might be important in some scenarios for preparing the basal protease to readily bind substrates of distinct cellular pathways either/both at the N-terminal or/and the PDZ domains (shown by arrows). (C) Represents N-terminal mediated allosteric activation of HtrA2, where the basal state (I) binds to the N-terminal binding partners (such as IAPs) of HtrA2 that leads to favorable conformational alterations in the distal protease and PDZ domains leading to an active state (E N ' state or N-terminal mediated activated state) as shown in step (V). This conformational state can further be modulated through temperature and/or PDZ domains leading to the most active protease (E act** state) as shown in (VI). This unique model shows intricate crosstalk among distinct activation networks that might or might not be mutually exclusive depending upon specific cellular requirements.
Frontiers in Molecular Biosciences | www.frontiersin.org February 2022 | Volume 9 | Article 824846 implies that HtrA2 protease activity protects neuronal mitochondria (Jones et al., 2003;Martins et al., 2004). It was earlier speculated that HtrA2 monitors and regulates protein folding in the mitochondria in a way DegP does in the bacterial periplasm. Further studies demonstrated that unfolded protein response (UPR) induced by tunicamycin or heat shock  as well as etoposide-activatedp53 stress pathwayupregulated expression of HtrA2 protease (Jin et al., 2003). Alike DegP, HtrA2 is also activated by elevated temperatures (Martins et al., 2003). Moreover, both HtrA2 and DegP prefer aliphatic Val or Ile in P1 position for substrate recognition and cleavage (Kolmar et al., 1996). Despite these similarities, HtrA2 shares strikingly higher structural and functional traits with DegS, which argue against its DegP-like chaperoning function, and hints a bearing closer to DegS. In particular, HtrA2 is protease-active at room temperature (Savopoulos et al., 2000), while DegP is activated only at elevated temperatures (Spiess et al., 1999). In addition, while DegP with two PDZ domains, folds into a higherorder hexagonal cage (Krojer et al., 2002), the trimeric HtrA2 and DegS (sans the additional PDZ and the necessary longer LA loop) are unable to prevent the entry of correctly folded proteins into the proteolytic sites Kim et al., 2003;Kim and Kim, 2005;Clausen et al., 2011) thus creating certain equivocacies toward defining its role as a chaperone. Interestingly, the identification of presenilin and amyloid precursor protein as natural substrates of HtrA2 Gupta et al., 2004) necessitates further studies to resolve the ambiguities surrounding HtrA's role in unfolded protein aggregation and quality control.

ROLE OF HtrA2 IN APOPTOSIS
HtrA2 was the first to be identified as an IAP binding protein (Hegde et al., 2002). Its functional similarity with second mitochondria-derived activator of caspase (Smac)/direct IAP binding protein with low pI (DIABLO) established its role as a proapoptotic molecule (Martins et al., 2002;Suzuki et al., 2004). HtrA2, which resides in the mitochondrial IMS is released into the cytosol after the separation of its 133-residue mitochondrial localization signal. This exposes an N-terminal IAP-binding motif (IBM) comprising a tetrapeptide 'AVPS' that is recognized as a binding site for IAPs. Unlike Smac, HtrA2 also cleaves IAPs and hence irrevocably relieves their inhibition on caspases (caspases-3,-7, and -9), thus promoting apoptosis Yang et al., 2003). Conservation of the IBM motif is found across species, where its Drosophila ortholog with two IBM motifs attracts DIAP1, enabling its removal by the serine protease activity (Challa et al., 2007). Likewise, the rhesus monkey and rodent orthologs of the protease have maintained the IBM motif suggesting evolutionary diversification of HtrA2 functions in higher organisms (Vande Walle et al., 2008). Interestingly, the two IAP-related proteins in C. elegans do not appear to be involved in apoptosis regulation (Fraser et al., 1999;Speliotes et al., 2000), suggesting that IAP proteins and the appearance of IAP antagonists like HtrA2 and Drosophila Reaper, Hid, and Grim are recent additions to the apoptotic molecular repertoire. Although human HtrA2 and its evolutionary paralogs bind and degrade many IAP family members, XIAP is found as the most effective amongst them as it engages a second interaction surface that permits strong caspase inhibition (Eckelman et al., 2006). However, to inhibit caspase activation, cIAP1,cIAP2, and XIAP target bound caspases for ubiquitinmediated proteasomal degradation (Vaux and Silke, 2005) thus necessitating HtrA2 to cleave all of them. Apart from N-terminal mediated apoptosis, HtrA2 binds important molecules of the apoptotic pathway through its regulatory C-terminal PDZ domain. The binding of substrates to the hydrophobic YIGV groove allosterically activates the protease for substrate binding and subsequent catalysis. Furthermore, binding to mitochondrial substrates at the early apoptotic stage such as GRIM-19 and Hax-1 might be important toward attuning the mature protease for its proapoptotic functions before it enters the cytoplasm (Cilenti et al., 2004;Ma et al., 2007;Chaganti et al., 2019;Kummari et al., 2021) where it binds several antiapoptotic proteins including IAPs and death effector domain (DED) containing Pea-15 (Trencia et al., 2004). HtrA2 is also capable of inducing caspase-independent apoptosis via its serine protease activity by cleaving several critical cellular molecules such as cytoskeletal proteins (actin, α-/β-tubulin, and vimentin) that are important for upholding cellular integrity (Vande Walle et al., 2007). KIAA1967 and KIAA0251are two newly identified proteins of the apoptotic pathway that have been found to be substrates of HtrA2 (Vande Walle et al., 2007). A caspase-generated cleavage fragment of KIAA1967 was demonstrated to cause mitochondrial clustering and matrix condensation in apoptotic HeLa cells (Sundararajan et al., 2005), whereas KIAA0251 interacts with the endoplasmic reticulum (ER) membrane protein Bap29, a component known to be required for caspase-8 activation in the ER (Breckenridge et al., 2002). Taken together, the substrates found and verified for HtrA2, reveal that this protease is involved in the apoptotic process at the cytoskeleton, translation initiation complex, and organelle dismantling levels.
Multiple modes of activation and a variety of substrates in different subcellular locations make HtrA2 omnipresent in the apoptotic pathway. Furthermore, distal N-/C-termini and heatmediated positive allosteric modulation as well as negative regulation of its proapoptotic functions through phosphorylation at Ser212 (Yang et al., 2016) re-instate its enigmatic role in the cell death network. However, the lack of definitive in vivo models of HtrA2's contribution toward apoptotic pathway might be limited by the number of identified natural substrates to date as well as due to redundancy in its functions in the cell, which requires further investigations.

HtrA2 IN NEURODEGENERATIVE DISORDERS AND CANCER
The first report on HtrA2's involvement in neurodegeneration came into existence with the identification of its interaction with Alzheimer's disease-associated protein, presenilin-1 . This was later substantiated by a homozygous loss-of-Frontiers in Molecular Biosciences | www.frontiersin.org February 2022 | Volume 9 | Article 824846 function mutation (S276C) identified as motor neurodegeneration 2 (mnd2) in mice (Jones et al., 2003), which was further bolstered by the development of homozygous HTRA2 knock-out mice exhibiting Parkinsonian phenotype (Martins et al., 2004) thus assigning HTRA2 gene the PARK13(Parkinson's disease 13) locus (Strauss et al., 2005;Abou-Sleiman et al., 2006). These critical inputs led to the initiation of several clinical studies involving PD cohorts from various populations across the globe to identify the involvement of HTRA2 and its mutations in PD progression and pathogenesis. However, the data obtained were quite contrasting. For example, a Germany-based clinical study that demonstrated heterozygous G399S and A141S mutations (Strauss et al., 2005;Bogaerts et al., 2008), was later impugned by another study from North America (Simon-Sanchez and Singleton, 2008). However, in vivo studies in transgenic mice harboring the G399S mutation (Casadei et al., 2016) and several other independent clinical investigations on non-overlapping rare HTRA2 mutations in Asian and European populations, re-established the correlation between HTRA2 gene and PD risk (Bogaerts et al., 2008;Lin et al., 2011;Wang et al., 2011). Furthermore, to delve into the loss of enzymatic activity of S276C mutation in human HtrA2 and correlate it with PD if any, X-ray crystallographic studies of the mutant were performed to understand the structural correlates of this functional repercussion (Wagh and Bose, 2018). The study provided a structural snapshot of the mutant at an atomic resolution where the inactivity was found to be conferred by loss of water-mediated H-bond between residues S276 and I270 on regulatory L2 and LD loops respectively; however, no clinical study could identify S276C mutation in any PD patient. Recently, another patient-derived research in the Indian population identified a rare likely-pathogenic mutation (T242M), which is critical for altering mitochondrial homeostasis due to loss of GSK-3β-mediated phosphorylation on HtrA2 leading to uncontrolled cell death with PD phenotype . Moreover, another contemporary study demonstrates a connection between neuronal death and selective downregulation revealing its link with Huntington's disease (Inagaki et al., 2008).
Despite these crucial discoveries, several contradictory reports challenge the establishment of HtrA2's role in neurodegeneration. This apparent anomaly in these studies might be due to a lack of focus on close interconnections among several parameters that include alterations in HTRA2, mitochondrial functional aberrations, and neurodegeneration. Therefore, future research endeavors encompassing both genetic and epigenetic interactions underlying the complex pathophysiological network of neurodegenerative disorders might provide a more comprehensive picture of HTRA2's association with these diseases.
While the involvement of HtrA1 in cancer is quite prevalent, there have been only a few direct reports of HtrA2's association with oncogenesis. HtrA2 has been found to be widely expressed in several cancer cell lines where over-expression triggered cell death (Suzuki et al., 2001;Martins et al., 2002). Biopsy sample analyses of specific cancers exhibited altered expression of HtrA2 suggesting its role in those cancers. For example, the level of the protease was found to be substantially less in endometrial and ovarian cancer tissues (Narkiewicz et al., 2008;Narkiewicz et al., 2009). On the other hand, higher HtrA2 expression in prostate tumors implicated its association with the differentiation of prostate cancer cells (Hu et al., 2006). Furthermore, elevated levels of HtrA2 in gastric cancers link it with this malignancy . However, although, the contribution of HtrA2 toward cancer development or regression yet remains to be conclusively elucidated, future studies using multidisciplinary approaches for delineating the HtrA2-associated extensive apoptotic network, and identifying its effect on tumorigenesis might shed more light on this pathophysiological collaboration.

CONCLUDING REMARKS AND FUTURE PERSPECTIVE
Recent progress in the structural and functional characterization of HtrA2 has greatly enhanced our understanding of this fascinating protein. Association of this protease with critical cellular functions such as apoptosis, protein quality control, cell growth, and unfolded protein response implicate it in several diseases including neurodegeneration, arthritis, and cancer. Unfortunately, the complexity of its oligomeric structural constitution and mechanism of activation makes it one of the most complex molecules in the HtrA family of proteases. However, recent advancements in deciphering the multi-layered allosteric modulation of HtrA2 from both structural and functional perspectives provide important cues toward targeting its different functions with specific modulators having desired characteristics.

AUTHOR CONTRIBUTIONS
KB, AC and RB conceived and designed the contents of the review. Manuscript preparation was done by AC, RB and KB. All authors read and approved the manuscript.