Mechanism of Fine-tuning pH Sensors in Proprotein Convertases

Background: Propeptides regulate the organelle-specific, pH-dependent activation of proprotein convertases. Results: A histidine residue pair in the propeptide cooperatively defines the activation pH for proprotein convertase 1/3. Conclusion: Different spatial distributions of histidine residues modulate activation pH of proprotein convertases. Significance: These results help our understanding of how protease paralogues evolved to regulate activation within specific cellular compartments. The propeptides of proprotein convertases (PCs) regulate activation of cognate protease domains by sensing pH of their organellar compartments as they transit the secretory pathway. Earlier experimental work identified a conserved histidine-encoded pH sensor within the propeptide of the canonical PC, furin. To date, whether protonation of this conserved histidine is solely responsible for PC activation has remained unclear because of the observation that various PC paralogues are activated at different organellar pH values. To ascertain additional determinants of PC activation, we analyzed PC1/3, a paralogue of furin that is activated at a pH of ∼5.4. Using biophysical, biochemical, and cell-based methods, we mimicked the protonation status of various histidines within the propeptide of PC1/3 and examined how such alterations can modulate pH-dependent protease activation. Our results indicate that whereas the conserved histidine plays a crucial role in pH sensing and activation of this protease an additional histidine acts as a “gatekeeper” that fine-tunes the sensitivity of the PC1/3 propeptide to facilitate the release inhibition at higher proton concentrations when compared with furin. Coupled with earlier analyses that highlighted the enrichment of the amino acid histidine within propeptides of secreted eukaryotic proteases, our work elucidates how secreted proteases have evolved to exploit the pH of the secretory pathway by altering the spatial juxtaposition of titratable groups to regulate their activity in a spatiotemporal fashion.

The propeptides of proprotein convertases (PCs) regulate activation of cognate protease domains by sensing pH of their organellar compartments as they transit the secretory pathway. Earlier experimental work identified a conserved histidine-encoded pH sensor within the propeptide of the canonical PC, furin. To date, whether protonation of this conserved histidine is solely responsible for PC activation has remained unclear because of the observation that various PC paralogues are activated at different organellar pH values. To ascertain additional determinants of PC activation, we analyzed PC1/3, a paralogue of furin that is activated at a pH of ϳ5.4. Using biophysical, biochemical, and cell-based methods, we mimicked the protonation status of various histidines within the propeptide of PC1/3 and examined how such alterations can modulate pH-dependent protease activation. Our results indicate that whereas the conserved histidine plays a crucial role in pH sensing and activation of this protease an additional histidine acts as a "gatekeeper" that fine-tunes the sensitivity of the PC1/3 propeptide to facilitate the release inhibition at higher proton concentrations when compared with furin. Coupled with earlier analyses that highlighted the enrichment of the amino acid histidine within propeptides of secreted eukaryotic proteases, our work elucidates how secreted proteases have evolved to exploit the pH of the secretory pathway by altering the spatial juxtaposition of titratable groups to regulate their activity in a spatiotemporal fashion.
Eukaryotic cells have evolved an elegant series of membranous compartments to regulate protein synthesis, folding, activation, sorting, and export (1,2). Beginning with their entry into the endoplasmic reticulum (ER) 2 as they exit the ribosome, proteins transit these secretory pathway compartments on their journey to their ultimate destination. Just as each group of compartments has mechanisms for maintaining precise pH and calcium balance (3,4), each family of proteins has co-evolved means to sense the environmental changes to ensure optimal organismal homeostasis. How eukaryotic proteins have diverged from their prokaryotic ancestors to exploit the unique organellar environment of the secretory pathway for biological function remains a fundamental question in cell biology (5).
Proprotein convertases (PCs) are eukaryotic members of the ubiquitous superfamily of subtilases (6). Comprising nine serine endoproteases (PC1/3, PC2, furin, PC4, PACE4, PC5/PC6, PC7/LPC/PC8, SKI/S1P, and NARC1/PCSK9), PCs are responsible for the conversion of a diverse range of precursor substrates to their active forms and thus play a central role in maintenance of physiologic homeostasis within cells and tissues (5,7). Furin, the most thoroughly characterized PC, is constitutively expressed in virtually all tissues and catalyzes the maturation of a diverse repertoire of hormones, enzymes, and receptor precursors within the secretory pathway (6). Not surprisingly, misregulation of furin results in both hyper-and hypoactivity and has been associated with cancer invasiveness and metastasis (8 -13), susceptibility to viral and parasitic infection (14), and increased severity of cardiovascular disease (15,16). Additionally, furin-deficient mice die at embryonic day 11 due to cardiac defects resulting from failed chorioallantoic fusion, axial rotation, and ventral closure (17). PC1/3 (also known as PC1 or PC3) along with PC2 is a neuroendocrine convertase responsible for the processing of several critical metabolic regulators, including insulin, glucagon, and proopiomelanocortin (18). Mice with knock-out of pcsk1 or pcsk2, the genes encoding PC1/3 and PC2 respectively, remain viable despite hormonal and/or neuroendocrine deficiencies (19 -21). Consistent with this, several studies characterizing mutations in PC1/3 have demonstrated an altered substrate processing that may underlie obesity, type II diabetes mellitus, and endocrine derangements (21)(22)(23)(24)(25). Further support for the correlation between polymorphisms in PC1/3 and metabolic disease is offered by reports of both individual patients (25)(26)(27)(28)(29) and epidemiologic studies (20).
As unregulated protease activity can have devastating consequences on organismal homeostasis (30), PCs, like all subtilases, are synthesized as zymogens and are activated in a precise spatiotemporal fashion by intramolecular proteolysis (31). A fundamental question has long been how the timing of activation is encoded and recognized by the protease. Our understanding of how this regulation is achieved is based on studies of profurin (32)(33)(34)(35). This furin precursor contains an 83-residue N-terminal propeptide that is requisite for folding of its cognate catalytic domain in the ER (36 -38) after completion of which the propeptide gets cleaved but subsequently remains associated as an inhibitor until reaching the trans-Golgi network. The necessity of furin to reach the trans-Golgi network before becoming active (39) coupled with the understanding of the unique environments of each compartment of the secretory pathway (40) argues for the presence of an encoded sensor that recognizes and responds to specific environmental signals (32). In fact, studies demonstrate that it is the mildly acidic pH of the trans-Golgi network (pH ϳ 6.5) that triggers the release and degradation of the furin propeptide (PRO FUR ) and thus release of inhibition of the mature protease domain of furin (MAT FUR ) (37,38). Interestingly, although the PC1/3 precursor, pro-PC1/3, transits the secretory pathway in much the same way as the furin precursor, our previous work indicates that PC1/3 requires a lower pH for its activation (pH ϳ 5.5) and thus is likely activated in a later compartment (32).
Organellar pH can alter the protonation status of charged residues, thus altering the structure and stability of the protein both on a local and global scale (41). Given that the pH range of interest in the case of the secretory pathway and PC activation falls within the physiologic range of ϳ7.4 in the ER to ϳ5.4 in the mature secretory granules, basic residues such as arginine (pK a ϳ 12.5) or acidic residues such as glutamate (pK a ϳ 4.2) are less likely to undergo altered protonation and be candidates for pH sensors; however, the imidazole ring of histidine has a pK a of ϳ6.0, making it ideally situated to respond to pH changes in this range. Histidine is used as a pH sensor in a variety of biological molecules, including hemoglobin and class II MHC (42)(43)(44), and is enriched within the propeptides of eukaryotic proteases that transit the secretory pathway in contrast to both cytosolic proteases and prokaryotic orthologues (33,34). Indeed, we identified a histidine (His 69 ) within PRO FUR that regulates the compartment-specific activation of MAT FUR (45). This histidine is in a loop adjacent to the secondary cleavage site nestled in a solvent-accessible pocket lined by hydrophobic residues. At the near-neutral pH of the ER, the deprotonated histidine acts as a hydrophobic residue, stabilizing the packing within this pocket and keeping the pH-sensitive loop protected against cleavage. Upon entry into the trans-Golgi network, the histidine is exposed to a ϳ10-fold higher proton concentration and thus is protonated; as a result, the imidazole side chain becomes polar, disrupting the packing and driving a local conformational change that exposes the secondary cleav-age site, allowing for rapid degradation and release of PRO FUR from MAT FUR (35).
Propeptide domains alone encode sufficient information for regulating the organelle-specific pH-dependent activation of cognate protease domains (32). Thus, when the propeptides of furin and PC1/3 are swapped, the pH-dependent protease activation is transferred in a propeptide-dictated manner both in vitro and in cells. Interestingly, the histidine pH sensor identified in PRO FUR is absolutely conserved within the propeptides of all PCs, including PRO PC1/3 . Despite their structural similarity and the presence of the conserved histidine throughout the family of PCs, it remains unclear how the pH sensitivity of other PCs is encoded.
In this study, we asked whether the pH-sensing histidine is conserved in PRO PC1/3 and why does MAT PC1/3 not become active at the same pH as furin? We present biochemical and structural data that suggest that the conserved histidine residue plays a critical role in pH-dependent activation similar to what has been established in furin; however, an additional histidine residue modulates this pH sensitivity such that a more acidic environment is required for PC1/3 activation.

Materials and Methods
Cell Culture and Transient Transfection-COS7 cells were maintained in Dulbecco's modified Eagle's high glucose medium (Hyclone) containing 10% (v/v) fetal bovine serum and 1% (v/v) penicillin-streptomycin. Cells were kept at 37°C in a 5% CO 2 environment as described previously (32,35). For expression of ER-retained constructs, COS7 cells at 60 -80% confluence were transfected with pcDNA3.1 expression vectors containing WT PC1/3 truncated at Arg 618 with a C-terminal -KDEL sequence (32) or vectors containing the noted PRO PC1/3 variants using TransIT LT-1 transfection reagent (Invitrogen) according to the manufacturer's instructions. The constructs also contained an HA tag at the N terminus of the propeptide and a FLAG tag at the N terminus of the protease domain (Fig. 1A).
pH-dependent Activation Assays-Twenty-four hours posttransfection with ER-retained PC1/3 constructs, cells were washed two times in PBS and incubated for 15 min in fractionation buffer (270 mM sucrose, 10 mM Tris, pH 7.4) with protease inhibitors (cOmplete EDTA-free, Roche Applied Science). Cells were lysed with three rounds of sonication (10 ϫ 1-s pulses at 30% power) in an ice bath. Unlysed cells and cell debris were pelleted at 800 ϫ g for 15 min at 4°C, and supernatant was transferred to a clean tube. Samples were incubated with 30% (w/v) PEG 8000 for 16 h at 4°C with gentle agitation. Precipitated proteins were pelleted by centrifugation at 17000 ϫ g for 30 min at 4°C. Pellets were resuspended in activity assay buffer (100 mM NaOAc, 0.1% (v/v) Brij-35, 5 mM CaCl 2 ) at a pH of 7.4, 6.4, or 5.4 and incubated at 37°C for 1 h to allow for activation unless otherwise noted. After 1 h, a 15 M concentration of the fluorogenic substrate Abz-RVKRGLA-Tyr(3-NO 2 ) was added, and activity was assayed in triplicate on a SpectraMax-M2 spectrofluorometer equipped with a 96-well plate reader (excitation, 320 nm; emission, 425 nm) for 1 h at room temperature. Data were fitted and analyzed using GraphPad Prism. Replicate samples were also processed for Western blotting; the anti-FLAG M2 antibody (Sigma) was used to recognize the N-terminally FLAG-tagged mature protease, and anti-HA.11 (Covance) used to recognize the N-terminally HA-tagged propeptide.
For experiments where limited trypsinization was used to digest the propeptide, PEG 8000-precipitated proteins were resuspended in activity assay buffer, pH 7.4 as above and incubated in the presence or absence of agarose-immobilized tosylphenylalanyl chloromethyl ketone-treated trypsin (ϳ10 p-toluene-sulfonyl-L-arginine methyl ester (TAME) units per sample; Thermo Scientific) for 1 h 37°C. Samples were then treated with soybean trypsin inhibitor to a final concentration of 1 g/ml for 20 min at 37°C before proceeding with the activity assay as above.
Protein Production and Purification-Codon-optimized sequences encoding either WT-PRO PC1/3 (mouse) or leucine and arginine variants were cloned into the pET11b backbone and expressed in BL21(DE3) Escherichia coli as described previously (32,46). Protein was purified from the soluble fraction by ion exchange after cell lysis via French pressure cell and dialyzed into 6 M guanidinium HCl-containing buffer (50 mM Tris-HCl, 6 M guanidinium HCl, pH 6.5) for long term storage. Before use, proteins were refolded by dialysis against refolding buffer (10 mM Tris, 10 mM cacodylate, 10 mM NaOAc, 150 mM KCl, 5 mM CaCl 2 ) at pH 6.8 unless otherwise noted. Concentration was determined after refolding by absorption at 280 nm.
Enzymatically active mature PC1/3 was expressed via the baculovirus system. The coding sequence of mouse PC1/3 was cloned into the pAcGP67A vector (BD Biosciences) using the primer pair TTTGCGGCGGATCCCGGGAAGAGGCAGTT-TGTTAATGAATG and TTCCATGCGGCCGCTCATCAT-CTCCTGTCATTCTGGACTGTATT into the XmaI and NotI sites, resulting in replacement of the endogenous signal sequence with the gp67 signal sequence and truncation at Arg 618 . The resulting coding sequence was subcloned into a pFastBac transfer vector (Life Technologies) containing a C-terminal GFP and His 8 tag.
The baculovirus genome was generated by transfection into DH10Bac cells according to the manufacturer's protocols (Bacto-Bac, Life Technologies). P1 virus was generated by infection of adherent Sf9 cells and amplified to the P3 as described in the manual. Virus titer was measured by end point dilution and quantification of infected cells by GFP fluorescence.
Protein was produced by infecting Sf9 suspension cultures in SF900-III medium (Life Technologies) at 4 ϫ 10 6 cells/ml at a multiplicity of infection of 2. After 24 h, the medium was clarified by two subsequent centrifugation steps at 500 and 10,000 ϫ g, respectively, for 20 min at 4°C. Protein was precipitated using 20% (w/v) PEG 8000 and resuspended in a 100-fold smaller volume of 50 mM MOPS at pH 6.5, 300 mM NaCl, 5 mM CaCl 2 , 0.01% (v/v) Brij-35 (Buffer A) mixed with fresh cOmplete EDTA-free protease inhibitor mixture. After clarification by centrifugation at 10,000 ϫ g for 20 min, 10 mM imidazole was added, and the buffer was agitated for 2 h with 1 ml of nickel-nitrilotriacetic acid resin (Thermo Scientific). Afterward the buffer was packed into a column and extensively washed (Ͼ20ϫ column volumes) with Buffer A with 25 mM imidazole. PC1/3 was then eluted using a linear imidazole gradient to 500 mM, and fractions containing GFP fluorescence and PC1/3 activity were pooled, flash frozen, and stored at Ϫ80°C (Fig. 1B). The pooled fractions had an estimated yield of 100 g/liter and a purity of 90%. A representative enzyme progress curve for PC1/3 is depicted in Fig. 1C. Although PC1/3 is known to exhibit complex enzyme kinetics, including a lag phase in activation, we saw minimal evidence of this at the enzyme concentrations used in our studies.
Circular Dichroism Spectroscopy-Circular dichroism spectroscopy was performed on an AVIV model 215 CD spectrometer at 4°C as described previously (32,35). Briefly, after centrifugation at 100,000 ϫ g for 30 min, protein concentration was adjusted to ϳ0.3 mg/ml, and spectra were obtained using a 1-mm quartz cuvette for far-UV spectra. A minimum of three independent scans were acquired, and data were averaged (31,47,48). Structural changes due to substitutions at individual histidine residues in PRO PC1/3 were monitored by plotting the changes in ellipticity at 222 nm at different pH values using the WT PRO PC1/3 as a control. Equilibrium unfolding can be represented as contributions of two states, the native (N) and the unfolded (U), as described previously (32,48) and given by Equation 1.
Alternately, the Henderson-Hasselbalch equation was used to describe the unfolding as originally described by Tanford and De (49).
where n ϭ 1 or 2 for a one-proton or two-proton model.
Activity Assays and Determination of IC 50 -Activity assays were performed to determine the inhibitory capabilities of the various propeptides as described previously (32,35). Briefly, a 15 M concentration of the fluorogenic substrate (Abz-RVKRGLA-Tyr(3-NO 2 )) was incubated with serially diluted amounts of either refolded WT or mutant PRO PC1/3 in concentrations ranging from 0.15 to 3000 nM in activity assay buffer at pH 6.8, and 0.02 enzyme units (1 enzyme unit of enzyme produces 267 relative fluorescence units/min in the assay buffer) of mature PC1/3 was added to initiate the reaction. Activity was assayed as described above. Data were fitted and analyzed using GraphPad Prism to determine IC 50 values as described previously (32,35).

Sequence and Structural Analysis of PRO FUR and PRO PC1/3
Orthologues-Multiple sequence alignment of PRO FUR and PRO PC1/3 provides several insights that serve as a starting point to generate a hypothesis about the mechanism of pH sensing in PC1/3 and how differences in sequence can encode different sensitivities to pH between the two homologues. The alignment of eight orthologues of PRO FUR and PRO PC1/3 indicates a high degree of sequence similarity, including stretches of completely conserved residues distributed throughout the entirety of both propeptides ( Fig. 2A). A structural comparison establishes that both propeptides adopt almost identical folds and comprise two ␣ helices packed against four ␤ sheets with a flexible loop region containing the secondary cleavage site between helices 1 and 2 (Fig. 2, B and C). As we have previously established that the pH sensor in furin is a histidine residue located at position 69 in the sequence (45), we paid particular attention to the distribution of histidines within PRO PC1/3 compared with PRO FUR . Although PRO FUR has histidines distributed throughout, PRO PC1/3 has only four, all of which are clustered within 10 residues of the secondary cleavage site. Of these, His 72 in PRO PC1/3 aligns with His 69 , the pH sensor identified in PRO FUR , and is positioned similarly at the N terminus of the cleavage loop, and thus was of primary interest. Notably, however, His 72 appears more buried within the core of PRO PC1/3 than His 69 in the PRO FUR (Fig. 2B), and His 75 in PRO PC1/3 that does not have a corresponding histidine in PRO FUR is also located within the cleavage loop of PRO PC1/3 . Based on the structure (Protein Data Bank code 1KN6 (50)), His 75 in PRO PC1/3 is surface-exposed and appears to overlie His 72 , leading us to speculate that His 75 may modulate the solvent accessibility of His 72 . Two other histidines, His 67 and His 85 , are also located within close proximity of the cleavage site loop. His 67 , which corresponds to His 66 in furin, is N-terminal to the cleavage loop and is situated at the interface that makes contact with the protease domain. Although His 66 was proposed as a potential pH sensor (50,51), subsequent biochemical and cell-based experiments demonstrated that this residue does not play a role in the pH-dependent activation of furin (45), making the corresponding His 67 in PRO PC1/3 a less likely player in the pH sensing mechanism of this protease. The remaining histidine residue (His 85 ) does not contribute significantly to the packing of the core of the propeptide and is solvent-exposed within helix 2 located at the C-terminal side of the cleavage loop. Because local loop movement is critical for facilitating accessibility of the secondary cleavage site for proteolysis, His 85 appears to be an unlikely candidate for a pH sensor, a supposition that is consistent with our preliminary IC 50 values and circular dichroism data that showed the protonation at these positions does not have a significant impact on the binding, structure, or stability of the isolated PRO PC1/3 (data not shown). Hence, here we focus on His 72 and His 75 as putative pH sensors within PRO PC1/3 .
Analysis of the Role of Histidine Residues in the pH-dependent Activation of the Precursor Protein Pro-PC1/3-To examine the role that the two histidine residues (His 72 and His 75 ) may play in the pH-dependent activation of the precursor pro-PC1/3 in vivo, we first measured the ability of propeptide variants of PC1/3 to undergo activation using ER-localized constructs as described previously (Fig. 1A and Ref. 32). Here pro-PC1/3 (ϳ78 kDa) undergoes efficient autoprocessing to produce MAT PC1/3 (ϳ67 kDa) that remains non-covalently associated with PRO PC1/3 (ϳ10 kDa) to form a stable inhibition complex (32). Because the KDEL motif restricts proteins to the neutral environs of the ER, PRO PC1/3 in the inhibition complex does not undergo the second autoprocessing step that is necessary for degradation and release of the propeptide. Hence, at pH values of 7.4 and 6.4, the ER extracts of MAT PC1/3 failed to degrade the PRO PC1/3 from the inhibition complex (Fig. 3A, inset) and displayed negligible catalytic activity against the fluo- Top, insert cloned into pcDNA3.1 where the coding sequence for WT mouse PC1/3 has been truncated at Arg 618 with the addition of a -KDEL sequence, an HA tag inserted at the N terminus of the propeptide, and a FLAG tag inserted at the N terminus of the protease domain. Bottom, the coding sequence for WT mouse PC1/3 was truncated at Arg 618 , and a C-terminal GFP tag and His 8 tag were added for insertion into the pFastBac vector for baculovirus production. B, representative SDS-polyacrylamide gel showing input loaded (Inp) and fractions collected following elution from a nickel-nitrilotriacetic acid column (6,(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36). Fractions indicated in the red box were pooled, concentrated, and used for assays. C, representative enzyme progress curve showing baculovirus-produced WT PC1/3 processing of the fluorogenic substrate. The curve suggests that PC1/3 has fully matured and displays normal enzyme characteristics. RFU, relative fluorescence units. rogenic peptide substrate (Fig. 3A). However, preincubation of the ER extracts of MAT PC1/3 at pH 5.4 caused a robust increase in the catalytic activity (Fig. 3A) that coincided with degradation of PRO PC1/3 as seen by the Western blot analysis (Fig. 3A, inset). It should be noted that this second autoprocessing event is time-dependent as degradation of the propeptide did not take place immediately upon exposure to low pH and at longer times of incubation continued to be degraded (Fig. 3E).
We then compared the WT pro-PC1/3 with variants in which His 72 or His 75 individually or both His 72 and His 75 together were substituted with either a leucine or arginine to mimic the constitutively deprotonated or protonated states of histidine, respectively, in accordance with our earlier work on the pH sensor in PRO FUR (32,35). Substituting either His 72 or His 75 with leucine allowed for pro-PC1/3 variants to undergo processing at the primary cleavage site to produce MAT PC1/3 (Fig. 3D) but abrogated degradation of the histidine to leucine variants of PRO PC1/3 at pH of 5.4, thereby preventing MAT PC1/3 from cleaving the fluorogenic substrate (Fig. 3, A and inset). Furthermore, no additional decrease in activity was observed when both histidines were mutated to leucine.
To confirm whether MAT PC1/3 obtained using the His 72 or His 75 PRO PC1/3 variants are catalytically active, the inhibition complexes were treated with trypsin ( Fig. 3C) to cleave the arginine-and lysine-rich propeptides. The results establish that the inhibition complexes, which failed to undergo autoactivation at pH 7.4, can be activated by incubation with trypsin at this pH, indicating that the protease itself is functional and that the His 72 or His 75 PRO PC1/3 variants are incapable of undergoing pH-dependent activation.
In contrast with the leucine substitutions, when His 72 or His 75 were replaced with arginine, the activity of the resulting MAT PC1/3 was more nuanced (Fig. 3B). Although the double mutant H72R/H75R appeared to be constitutively active at all pH values, the maximal activity of the H72R variant was only marginally greater than the control at all pH values, suggesting that protonation at this position, although necessary (as evidenced by the inability of the H72L variant to be activated), must be appropriately timed. The Western blot depicting the processing of the precursor (Fig. 3D) demonstrates that the H72R pro-PC1/3 variant did not undergo appreciable processing and suggests that inappropriate protonation of this residue may impair folding and/or processing similar to results seen in furin (45). An alternate explanation is that this mutation in the propeptide impairs autocatalytic activity, resulting in reduced propeptide excision in the ER similar to the previously reported S307L mutation, which also displayed diminished activity in trans against a peptide substrate (25). Interestingly, the H75R variant showed activity at pH 7.4 that was higher than that of the wild type and appeared to be completely active at pH 6.4, making the resulting MAT PC1/3 more furin-like. Not surprisingly, the Western blot analysis established that the H75R variant of pro-PC1/3 was efficiently autoprocessed and that H75R-PRO PC1/3 was completely destroyed at pH of 7.4 on a time scale that we were not able to capture in our analysis (Fig.  3, B and inset). Thus, our mutational analysis suggests that both histidines at position 72 and 75 within PRO PC1/3 may be playing roles in the pH-dependent activation of pro-PC1/3 as it transits the secretory pathway. A final caveat of these studies is that retaining the proteins in the ER/cis-Golgi prevents terminal glycosylation, which has been shown to play a role in processing and secretion of mature PC1/3. The extent of glycosylation of these constructs has not been determined; however, one report suggests that it is the addition of the precursor glycan (which occurs in the ER), not the terminal trimming, that is indispensable for PC1/3 function (52).
Effect of pH on Structural Changes in PRO PC1/3 -Because the protonation status of the pH sensor alters the structure of PRO FUR (32,35), we next analyzed how the protonation of His 72 and His 75 affects the secondary structure of isolated PRO PC1/3 by circular dichroism spectroscopy using leucine and arginine variants as mimics for the constitutively protonated and deprotonated states, respectively. The CD spectra reveal the presence of significant secondary structure within isolated PRO PC1/3 that was not significantly perturbed when the histidines were substituted by leucines (Fig. 4). This is consistent with our hypothesis that at pH 6.8 the histidine residues are in their deprotonated forms and thus demonstrates that leucine can act as a faithful representation of the constitutively deprotonated state.
When His 72 and/or His 75 was replaced with arginine, an amino acid that mimics a constitutively protonated histidine, the secondary structure of the variant PRO PC1/3 decreased significantly as seen from the changes in ellipticity at 222 and 208 nm (Fig. 4). It is important to note two key points. First, the effects of the H72R substitution alone were greater than the H75R substitution alone, and second, the effects of the substi-  tutions are cumulative wherein there was a greater loss of secondary structure in the H72R/H75R double mutant than in either of the single mutants. For comparison, the spectrum of WT-PRO PC1/3 at pH 4.0 is shown (Fig. 4, black dashed line). We can see that the behavior of H72R/H75R-PRO PC1/3 at pH 6.8 recapitulates the behavior of the WT at pH 4.0 where we would expect both histidines to be protonated and thus the propeptide to be maximally unfolded. With respect to the shifts in the troughs of the spectra around 208 nm, an indicator of ␣ helicity, we can see that the spectra of H75R is slightly left-shifted toward 194 nm with respect to the wild type spectra, indicating a higher percentage of random coil. In contrast, the trough in H72R is right-shifted but significantly decreased in intensity, suggesting that the although the propeptide has more ␣ helical content it is losing overall structure; this perhaps reflects a transition from a more structured to a less structured form similar to what has been observed in subtilisin (31). The cumulative effect of the arginine substitutions can be seen in the spectra of the double mutant where the spectral trough is roughly equivalent to that of WT-PRO PC1/3 at pH 4.0.
The loss in structure is consistent with our earlier observations of the in vitro catalytic activities. Notably, because the double mutant (H72R/H75R) is the only variant that exhibited constitutive activity at pH 7.4 (Fig. 3, A and B), these results strongly suggest that protonation of either His 72 or His 75 causes destabilization of the isolated PRO PC1/3 ; however, there is likely a requisite role for both residues in the pH-dependent activation of PC1/3.
A limitation of the above biophysical descriptions is the fact that they are done at static pH, whereas the PCs experience a broad range of pH as they transit the secretory pathway. Therefore, to further investigate the effect of these substitutions on the structural changes in the propeptide in a physiologically relevant way, we monitored unfolding of the isolated PRO PC1/3 via changes in secondary structure between pH 8.0 and pH 4.0 (Fig. 5). The WT-PRO PC1/3 underwent a cooperative sigmoidal transition from a well structured state at pH 8.0 to a less structured state at pH 4.0 (Fig. 5, A and B, black points). Data were fit to various models, including two-state unfolding (black line) and the Henderson-Hasselbalch equation accounting for either a single protonation event (green line) or two protonation events (red line). In each case, the midpoint of the transition between the well structured state and the less structured state occurred at ϳpH 5.3, coincident with the pH of the later, more acidic compartment in which MAT PC1/3 is activated in the secretory pathway.
Identical treatment of PRO PC1/3 variants yielded some interesting insights. At pH values above ϳ6.5, the leucine variants of PRO PC1/3 appeared similarly well folded as the WT PRO PC1/3 ; however, as pH dropped, the behavior of the mutants diverged. The H72L-PRO PC1/3 variant was largely stable, retaining its secondary structure at pH 5.0 and only undergoing insignificant unfolding (ϳ20% when compared with WT-PRO PC1/3 ) as the pH approached 4.0 potentially due to the influence of secondary residues. Under identical conditions, the H75L variant also appeared unaffected by pH until pH 5.0 after which it lost about 60% of its secondary structure when compared with WT PRO PC1/3 at pH 4.0. Thus, our results indicate that the overall pH-dependent stability of the H72L variant is greater than that of H75L as well as WT PRO PC1/3 and suggest that the His 72 residue contributes more toward the pH sensitivity. When the two mutations were combined, we observed that the variant PRO PC1/3 was largely unaffected by the changing proton concentrations, reflecting the replacement of both histidines with nonprotonatable mimics. This suggests that although the effects of protonation and/or salt bridge formation on the structure of PRO PC1/3 cannot be ruled out the histidine protonation status is the primary driver of these structural changes.
In contrast, we noted a striking effect of making the alternate substitutions, replacing histidines with positively charged arginine. Both the single H72R and double H72R/H75R exhibited significantly less secondary structure than WT-PRO PC1/3 even at pH 8.0, and although the H72R variant did undergo some pH-dependent unfolding at pH ϳ6.0, again suggesting a minor role for protonation of secondary residues in the lower pH range, these changes were small. The H75R exhibited the most striking of pH-dependent changes: at pH 8.0, the H75R mutant had significant secondary structure and appeared to undergo a sigmoidal transition to a less folded state with a midpoint of transition at pH ϳ7.0, again making this variant the most furin-like of those examined. Taken as a whole, these data again suggest a role for both His 72 and His 75 ; in replacing both histidines with leucine, the pH-dependent loss of secondary structure could be abrogated, whereas the single sub- stitutions merely attenuated this loss. Similarly, although replacing both residues with arginine rendered PRO PC1/3 likewise impervious to changes in pH, it existed in a baseline less well folded state. The single arginine variants retained a degree of structural responsiveness to pH, undergoing structural transitions at distinct pH values, suggesting underlying differences in their pK a values.
Effect of His Substitution on the Interactions of PRO PC1/3 with MAT PC1/3 -Although studies of the isolated PRO PC1/3 are useful to understand the role of changing proton concentration on the conformational dynamics of this critical domain, ultimately PRO PC1/3 exists as a complex with its cognate protease domain and exerts its regulatory effects via inhibition. Therefore, we next determined the IC 50 values of PRO PC1/3 and its histidine variants for MAT PC1/3 (Fig. 6). The WT PRO PC1/3 inhibited MAT PC1/3 with an IC 50 of 2.3 nM, which is consistent with previously published values (Fig. 6, A and B). Substituting either histidine singly or in combination with leucine did not significantly perturb the IC 50 values of PRO PC1/3 (Fig. 6B) consistent with our observation that there were not significant changes in secondary structure of these at pH 5.5 (Fig. 3A). Likewise, consistent with our earlier observations, both the single H75R and the double H72R/H75R variants were poor inhibitors with IC 50 values ϳ10 -15-fold higher than that of WT PRO PC1/3 . However, inhibition studies with H72R-PRO PC1/3 variant yielded some surprising results: unlike the other arginine variants, the IC 50 of H72R-PRO PC1/3 was similar to that of the WT PRO PC1/3 (Fig. 6B) consistent with our prior observation that in vitro activity of this variant is only ϳ50% of the wild type. As we are yet unable to directly examine how pH affects the PRO PC1/3 -MAT PC1/3 inhibition complex directly because of the high concentrations of mature PC1/3 required to create stoichiometric complexes and difficulty in obtaining a truly catalytically inactive enzyme for obtaining stable complexes, we are left to speculate that perhaps there is interplay of additional residues at the interface of the protease and PRO PC1/3 that allow the propeptide to remain tightly associated despite having less structure and lower stability at acidic pH.

Discussion
Organellar pH is a critical regulator of a wide range of intracellular events, including zymogen activation, signaling cascades, ion channel activity, and receptor-ligand interactions. Each of these events must be tightly regulated both temporally and spatially to maintain organismal homeostasis, and many eukaryotic proteins have evolved to exploit environmental pH as a cue to regulate their activation via alterations in the protonation status of titratable amino acid side chains. Not surprisingly then, protonation with increasing organellar acidification is the major biochemical cue that regulates the final activation step of PCs. We are only just beginning to understand how the differing pH sensitivity of individual PCs is encoded and how the addition of a single proton can drive the biochemical and biophysical changes required to initiate protease activation. We have earlier demonstrated that protonation of His 69 can disrupt a hydrophobic pocket within PRO FUR to drive local unfolding of the secondary cleavage site loop to facilitate PRO FUR proteolysis that results in furin activation (35,45). Although His 69 is conserved within all PCs, individual PCs nonetheless undergo activation at different pH values with actual proton concentrations that vary over 10-fold, suggesting that additional complexities remain to be determined.
Our results demonstrate that although histidine residues are mostly conserved within orthologues of PC1/3 and furin individually they show significant differences between the two families with only the primary pH sensor residue (His 69 in PRO FUR and His 72 in PRO PC1/3 ) absolutely conserved throughout the alignment (Fig. 2). The remaining histidines, His 67 , His 75 , and His 85 , in PRO PC1/3 are located within proximity of the cleavage site loop. His 67 (which corresponds to His 66 in furin) is situated at the interface with the protease domain and has been demonstrated to have no influence on pH sensing in furin (45). Furthermore, preliminary studies using leucine and arginine substitutions at positions 67 and 85 in PC1/3 suggest that these variants do not affect the secondary structure of the isolated propeptide domain or their IC 50 values (data not shown). It is noteworthy that (i) the counterpart to His 69 , which is solventaccessible in furin, appears more buried within PC1/3; (ii) the imidazole side chain of His 75 is solvent-exposed and does not have an analogous counterpart in furin; and (iii) His 75 precedes a proline residue in PC1/3. Experimental studies have established that when a histidine side chain precedes a proline residue the proline isomerization rates increase up to 10-fold when the imidazole side chain is protonated relative to the deprotonated state (53,54). In this study, we therefore analyzed the role of the conserved histidine in PC1/3 (His 72 ) as well as that of a second histidine (His 75 ), which we propose acts as a "gatekeeper" that regulates solvent accessibility and thus protonation of the primary pH sensing histidine.
To decipher the role of the two histidine residues in regulating the pH-dependent activation of PC1/3, we used constructs with a C-terminal KDEL sequence that retains the noncovalently bound inhibition complex within the ER; in this case, pro-PC1/3 has undergone the pH-independent primary processing subsequent to folding, but the propeptide remains associated, acting as an inhibitor of protease activity. Our results indicated that mutation of either His 72 or His 75 to a nonprotonatable mimic was sufficient to block the second processing event and thus activation of PC1/3 at all pH values assayed, indicating that the protonation of both histidine residues is necessary and that neither alone are sufficient to drive pH-dependent activation of PC1/3. It is noteworthy that when replaced with an arginine, a surrogate that mimics a constitutively protonated state, the H72R-PRO PC1/3 did not promote activation, whereas the H75R-PRO PC1/3 underwent marginal activation at pH 7.4. Although activation at pH 6.4 and 5.4 increased for both H72R-PRO PC1/3 and H75R-PRO PC1/3 , only the double variant (H72R/H75R-PRO PC1/3 ) mediated robust activation at all pH values. On this basis, we propose that both His 72 and His 75 appear to work cooperatively to mediate the pH-dependent activation of PC1/3 (Fig. 3).
The biophysical analysis of the isolated propeptides further supported the idea that both histidine residues were important in the mechanism of activation for this protease. Here, the pHdependent structural transitions seen in the various mutant propeptides are particularly insightful. The H72L-PRO PC1/3 , H75L-PRO PC1/3 , and H72L/H75L-PRO PC1/3 variants behaved similarly to WT PRO PC1/3 between pH 6.0 and 8.0. However, at pH less than 6.0, the WT PRO PC1/3 began to unfold with a calculated pK a of ϳ5.2 when the data were fit to a two-proton model (Fig. 5A). Under these conditions, all histidine to leucine PRO PC1/3 variants were significantly more stable than the WT PRO PC1/3 , and although the H75L-PRO PC1/3 lost about 50% of its ellipticity at 222 nm when compared with WT PRO PC1/3 , H72L-PRO PC1/3 and H72L/H75L-PRO PC1/3 variants underwent only marginal unfolding (ϳ20%) under identical conditions (Fig. 5B). Although these results indicate that a single substitution of histidine with nonprotonatable leucine residues at position 72 or 75 blunts the ability of PRO PC1/3 variant to respond to more acidic pH, the subtle differences in the behavior of the individual variants allows one to begin teasing apart the differing roles of these residues.
First, given that the single H72L variant was sufficient to attenuate the pH-dependent structural changes within the propeptide (Fig. 5B) and block pH-dependent activation (Fig. 3A), we propose that the protonation status of His 72 is a major driver of the requisite unfolding of the cleavage loop of PRO PC1/3 . Second, the observation that the H75L underwent partial unfolding, albeit to a lesser degree than the WT PRO PC1/3 , indicates that although protonation of His 72 is necessary it is not sufficient, and thus the combinatorial effect of protonation of both residues appears to be necessary for activation of PC1/3. As a final note, the small change in secondary structural content in the double H72R/H75R-PRO PC1/3 variant was likely due to the influence of other titratable groups, which at low pH values may enhance the local changes required for propeptide processing and release, and thus further studies are needed to more fully investigate this possibility.
By introducing an arginine in place of the histidines of interest, we were able to assess the effect of a positive charge at these positions within the propeptides. Again, the differences between the variants shed light on the differing role that each of these histidines play and lends further support to the notion that there is a synergistic effect between His 72 and His 75 that allows for the precise spatiotemporal regulation of the loosening of secondary structure required for PRO PC1/3 processing. Even at pH 8, H72R-PRO PC1/3 displayed markedly less secondary structure than the WT-PRO PC1/3 , reflecting its role as a major driver of destabilization. There was an additional loss of ␣ helicity as pH dropped below ϳ6 either due to the aggregate effect of protonation of His 75 or the influence of other side chains. The behavior of H75R-PRO PC1/3 is again of particular interest: although somewhat destabilized relative to WT-PRO PC1/3 , it was nonetheless well structured at pH 8 when compared with H72R-PRO PC1/3 and H72R/H75R-PRO PC1/3 . As pH was lowered, H75R-PRO PC1/3 immediately began to lose secondary structure but plateaued at pH 6. This again highlights our assertion that the positive charge at either of these positions singly is not sufficient to account for the activation behavior of PC1/3 but rather that cooperative action between early protonation of His 75 followed by later protonation of His 72 is required.
A final piece of insight is offered by the ability of the propeptides to inhibit the protease domain (Fig. 6). As would be expected, IC 50 values (at pH 6.8) for the leucine variant propeptides roughly approximated those of the wild type, whereas those for the H75R and H72R/H75R variant were ϳ15-fold higher, indicating that they are notably weaker inhibitors. At first, it may seem surprising that the H72R variant has an IC 50 that is comparable with that of the wild type; however, we propose that this reflects a critical role of the histidine at position 75: in addition to modulating the solvent accessibility of the hydrophobic pocket it overlays, protonation of His 75 may be involved in a destabilization of the propeptide-protease complex. As we are yet unable to produce the mature protease in sufficient quantities, we are unable to more directly determine k on /k off and more thoroughly characterize the interaction between the protease and its cognate propeptide.
Taking the above as a whole, we can then propose a mechanism by which PC1/3 is activated (Fig. 7). We believe that the two histidine residues within the cleavage loop are in proximity of one another and require sequential protonation to allow for activation. His 75 , which sits at the top of the hydrophobic pocket, is likely protonated first and begins the local unfolding of the cleavage loop. It is tempting to speculate that protonation of His 75 facilitates local unfolding by enhancing the kinetics of cis-trans isomerization of Pro 76 as demonstrated by systematic studies in a pentapeptide series, Ac-Ala-Xaa-Pro-Ala-Lys-NH 2 (53). These studies, which examined the influence of each of the 20 amino acids at position Xaa on the energetics of proline isomerization using NMR spectroscopy, demonstrated that the rates of proline isomerization are enhanced severalfold only when the side chains of tyrosine and histidine residues are protonated. The occurrence of histidine residues preceding a proline is higher than the overall frequency of the individual amino acids would predict (55). Thus we propose that the protonation of His 75 enhances local unfolding of the loop, which allows His 72 to become more solvent-accessible and simultaneously disrupt the interface between the propeptide and protease. Once solvent-accessible, His 72 can then be protonated, thus delivering the second and final blow that drives the processing and dissociation of the propeptide, thus releasing inhibition from the protease. In this model, His 75 acts as a gatekeeper residue, blocking access to the hydrophobic pocket in its deprotonated state, a restriction that can be removed via protonation. The model for two protonation events is also supported by CD data on the WT-PRO PC1/3 . Notably, when we fit the CD data from pH-dependent unfolding experiments to a modified form of the Henderson-Hasselbalch equation (49), we found that a model of two titratable groups better fit the conformational changes in PRO PC1/3 in response to changing pH than one that only allowed for a single protonation (Fig. 5A). We also speculate that the close proximity of His 72 and His 75 may also be critical to understanding the mechanism of activation and reason that PC1/3 is only activated in acidic environs. When both residues are unprotonated, hydrophobic packing of the core of the propeptide is the major stabilizing force similar to the furin propeptide. One must ask then is it simply a greater degree of stability in the structure of PRO PC1/3 that requires the stronger disrupting force of two positive charges to drive its unfolding, or does the protonation of one residue influence the protonation of the second? Concomitant work in our laboratory has been focused on defining pK a values of the histidines within PRO FUR and PRO PC1/3 . Although the pK a values of His 67 , His 75 , and His 85 are all ϳ6.0, the pK a of His 72 is acid-shifted with a value of ϳ5.5 (56), which similar to the pH of activation of PC1/3. Therefore, we believe that the spatial arrangement and interaction of His 72 and His 75 can explain why PC1/3 is activated by a 10-fold higher proton concentration than its paralogue furin.
This work offers significant insight into the question of how each member of the PC family encodes unique organelle-specific information about activation and processing. We believe that although the primary pH sensor (His 69 in furin and His 72 in PC1/3) has been evolutionarily conserved throughout PCs, concomitant with histidine enrichment has been a tuning of the pH sensitivity of these proteases. As seen in the multiple sequence alignment, only the position of the primary pH sensor has been absolutely conserved; thus domain-specific conservation as opposed to position-specific conservation seems to be crucial to encoding pH sensitivity in the propeptides of proteases in this family. Looking more broadly at secreted proteases across eukaryotes and prokaryotes, we had reported previously that there is an enrichment of histidine residues in the propeptides of secreted eukaryotic proteases that is not seen in their prokaryotic paralogues. Again, we see that although there is a bias for more histidines within this particular domain there is no appreciable bias for specific positions (33). These observations highlight the importance of spatial juxtapositioning of titratable groups in regions of proteins to tune their pH sensitivity, an evolutionary adaptation that was likely concomitant with the emergence of compartmentalization and specialization allowed by the secretory pathway within eukaryotic cells.
The issue of "mistuning" via mutation is also interesting to consider. The most obvious case to consider would be the loss of the conserved pH sensor, yielding either an inactivatable or constitutively active protease depending on the type of mutation. More interesting, however, would be the mutation of one of the surrounding residues and its effect on the local environs and protonatability of not only the primary pH sensor but also the gatekeeper or tuning residues. Evidence for the plausibility of this notion is lent by several single nucleotide polymorphisms in PRO FUR and PRO PC1/3 . Preliminary characterization of one of these polymorphisms, ⌬R80Q (57), indicates that changes in the propeptide can affect the biosynthesis, secretion, and catalytic activity similar to other previously described SNPs in the catalytic domain, and thus it is hypothesized that it may contribute to cancer (including head and neck carcinoma and lung adenocarcinoma (58)) or metabolic disease. Further characterization of these polymorphisms would be insightful not only in understanding the basis on which they are able to cause disease but also in the broader sense of understanding what role propeptides have in dictating proteolytic activity and processing. Similarly, environmental homeostasis is required for optimal proteolytic function; perturbation of cellular pH through disease (59) may cause activation of a PC prematurely or prevent activation all together. The role of the pH gradient in the secretory pathway in orchestrating proteolytic processing of substrates cannot be understated. Thus as we continue to understand the interplay among the proprotein convertases, proton concentration, and pH sensors, we hope to gain better insight into how things go wrong in the disease state and how we can better approach solutions to restore homeostasis.
Author Contributions-D. M. W. made substantial contributions to conception, design, and acquisition of data. J. E. helped with the analysis. U. S. helped draft and revise the article and helped with the analysis of data. All authors reviewed the results and approved the final version of the manuscript. . Right, both histidines are protonated, causing cleavage loop to unfold completely and exposing cleavage site to allow processing.