If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Ovarian clear cell carcinoma (OCCC) is an understudied poor prognosis subtype of ovarian cancer lacking in effective targeted therapies. Efforts to define molecular drivers of OCCC malignancy may lead to new therapeutic targets and approaches. Among potential targets are secreted proteases, enzymes which in many cancers serve as key drivers of malignant progression. Here, we found that inhibitors of trypsin-like serine proteases suppressed malignant phenotypes of OCCC cell lines. To identify the proteases responsible for malignancy in OCCC, we employed activity-based protein profiling to directly analyze enzyme activity. We developed an activity-based probe featuring an arginine diphenylphosphonate warhead to detect active serine proteases of trypsin-like specificity and a biotin handle to facilitate affinity purification of labeled proteases. Using this probe, we identified active trypsin-like serine proteases within the complex proteomes secreted by OCCC cell lines, including two proteases in common, tissue plasminogen activator and urokinase-type plasminogen activator. Further interrogation of these proteases showed that both were involved in cancer cell invasion and proliferation of OCCC cells and were also detected in in vivo models of OCCC. We conclude the detection of tissue plasminogen activator and urokinase-type plasminogen activator as catalytically active proteases and significant drivers of the malignant phenotype may point to these enzymes as targets for new therapeutic strategies in OCCC. Our activity-based probe and profiling methodology will also serve as a valuable tool for detection of active trypsin-like serine proteases in models of other cancers and other diseases.
Ovarian cancer is the most lethal gynecological malignancy, with an average 5-years survival rate of only ∼50% (https://seer.cancer.gov/statfacts/html/ovary.html). Ovarian clear cell carcinoma (OCCC) is a relatively little-studied histotype of ovarian cancer with particularly poor prognosis when diagnosed in advanced stages (
). To significantly improve ovarian cancer treatment, there is urgent need to better define the underlying biology of different histotypes and their distinct paths of progression and to identify individualized therapeutic approaches based on molecular cancer drivers.
Over 700 proteases are found in the human proteome (MEROPS Peptidase Database (
). Serine proteases of the S1 or chymotrypsin family represent the largest family (Table S1); a majority of these proteases (79 in humans) possess trypsin-like specificity for cleavage after arginine or lysine residues within peptides and proteins (Table S2). Most of these proteases are secreted, with a lesser number tethered to the extracellular cell surface (Tables S1 and S3); in general, these enzymes perform their functions extracellularly. Trypsin-like serine proteases have been identified as potent regulators of numerous pathways in humans and are vital for digestion, blood coagulation, fibrinolysis, reproduction, and immune responses (
Our preliminary studies using models of OCCC implicated one or more trypsin-like serine proteases in malignant phenotypes of this cancer subtype, posing the challenge of how to identify the relevant target(s) from among 79 candidates. The identification of these specific proteases that contribute to OCCC progression could lead to novel targeted therapeutic strategies. However, proteases often show high levels of regulatory complexity and organization into activation cascades and networks, as their activity is regulated by posttranslational proteolytic activation and by interactions with endogenous inhibitors (
). Importantly, transcriptional profiling methods and even most proteomic profiling methods are unable to discriminate among active and inactive forms and thus do not accurately reflect protease activity in complex biological samples, presenting a significant technical challenge.
Activity-based protein profiling (ABPP) is a method of proteomic profiling in which activity-based molecular probes are employed to identify and quantify enzymes on the basis of enzyme activity rather than simple protein abundance. When applied to identification of proteases, an advantage is the ability to distinguish the population of active proteases of a particular family, unconfounded by inactive zymogen forms or inhibited proteases (
). Activity-based probes are generally designed to bind specifically in the active site of an enzyme or class of enzymes and to react irreversibly with catalytic residues in the enzyme active site, via a reactive “warhead”. The warhead is connected via a peptide or chemical linker to an isotope, fluorescent, and/or affinity tag utilized for detection and/or purification of the probe-labeled protein population (
). The target enzyme class dictates the choice of electrophile; for example, global detection of serine hydrolases can be achieved through fluorophosphonate electrophiles, which are highly reactive to capture this diverse pool of enzymes (
). To target proteases specifically, less broadly reactive electrophiles have been developed; warheads that have been employed to selectively target serine proteases over other serine hydrolases include peptidyl diphenylphosphonates (
). Peptidyl groups adjacent to the warhead form interactions with the substrate binding subsites of target proteases, and may further restrict selectivity of the probe to protease subsets or even to individual targets (
Here, we aimed to broadly profile active serine proteases with trypsin-like specificity, to identify those that drive malignancy in OCCC. Whereas previously developed probes of trypsin-like enzymes featured lysine diphenylphosphonate warheads as substrate mimics (
), here we aimed to achieve more comprehensive coverage of the pool of potential targets by employing an arginine diphenylphosphonate warhead, since >80% of human trypsin-like proteases show preference for Arg over Lys at the P1 subsite
Substrate residues surrounding the cleavage site are numbered P1, P2, P3, etc. in the direction of the N-terminus and P1ʹ, P2ʹ, P3ʹ, etc. in the direction of the C-terminus, following the nomenclature of Schechter and Berger (
). The identity of the P1 residue is a primary determinant of substrate specificity for serine proteases.
†Substrate residues surrounding the cleavage site are numbered P1, P2, P3, etc. in the direction of the N-terminus and P1ʹ, P2ʹ, P3ʹ, etc. in the direction of the C-terminus, following the nomenclature of Schechter and Berger (
). The identity of the P1 residue is a primary determinant of substrate specificity for serine proteases.
, whereas only one (1.5%) shows marked preference for Lys over Arg (see Tables S1 and S2). Not finding the desired features among commercially available reagents, we developed a novel probe for our purpose. Additional features of our probe include an extended generic protease recognition sequence spanning P4 – P1 subsites, a biotin handle to facilitate affinity purification of labeled proteases, and an intervening tobacco etch virus (TEV) protease cleavage site for compatibility with tandem orthogonal proteolysis-ABPP approaches (
). The rationale for inclusion of the TEV protease site is to enable selective proteolytic release of probe-labeled proteins or peptides captured on streptavidin beads for proteomic analysis, thus eliminating background from proteins that bind nonspecifically.
The probe, which showed high stability in aqueous cell culture medium, was used to profile the secreted proteome of two cultured OCCC-derived cell lines. We identified multiple active serine proteases in each OCCC cell line, including two proteases in common, tissue plasminogen activator (tPA) and urokinase-type plasminogen activator (uPA). We interrogated the involvement of these proteases in malignant phenotypes of the OCCC cell lines and found both proteases to be involved in cancer cell invasion and proliferation. Thus, our activity-based probe of trypsin-like serine proteases successfully identified two mediators of tumor progression in OCCC, which may offer targets for new therapeutic strategies. Our probe and approach also offer valuable tools for detection of active trypsin-like serine proteases in models of other cancers and other diseases.
Trypsin inhibitors implicate unknown active trypsin-like serine proteases as drivers of OCCC invasion and proliferation
A sensitivity screen of cancer cell lines toward diminazene aceturate (Fig. 1A), a drug which we and others have previously identified as an inhibitor of trypsin-like serine proteases (
), showed significant inhibition of malignant phenotypes of multiple OCCC cell lines. In JHOC5 and JHOC9 cell lines, representative models of OCCC, we detected significant sensitivity as exemplified by phenotypic assays for both invasion (Fig. 1, B and C) and proliferation (Fig. 1, D and E), suggesting that targets of diminazene are important drivers of OCCC malignancy.
Because diminazene is also an activator of angiotensin-converting enzyme 2 (
) and may potentially exert effects on cells via this or other alternative targets, we sought to corroborate our initial findings using additional broad-spectrum trypsin inhibitors gabexate mesylate (Fig. S1A) (
). Consistent with the effects of diminazene, gabexate mesylate and hydroxystilbamidine both similarly inhibited the invasion and proliferation of OCCC cell lines JHOC5 and JHOC9 (Figs. S1, B–E and S2, B–E). Given that the three compounds possess different chemical scaffolds and are not reported to inhibit any nonprotease targets in common, we interpret our observations as evidence of involvement of trypsin-like serine proteases in the invasion and proliferation of OCCC.
To identify the specific targets that are key to these phenotypic changes, we initially evaluated expression of the three human trypsin isoforms, each of which has been described previously among potential malignant drivers in other types of ovarian cancer (
). However, in the OCCC cell lines, we found trypsins to be either undetected or barely expressed at levels below that of normal ovarian surface epithelium cells (Fig. S3). Thus, we reasoned that trypsins are unlikely to be the relevant target(s) of the serine protease inhibitors that are responsible for promoting invasion and proliferation in OCCC. Rather, we hypothesized that alternative trypsin-like serine proteases targeted by diminazene, gabexate mesylate, and hydroxystilbamidine are expressed in active form in OCCC where they represent drivers of the observed malignant phenotypes.
Design and synthesis of an arginine diphenylphosphonate probe to label active trypsin-like serine proteases
There are 79 candidate trypsin-like serine proteases in the human proteome (Tables S1 and S2). Because they are typically regulated by posttranslational cleavage and by endogenous inhibitors, levels of transcript and protein expression often do not reflect levels of enzyme activity (
). To directly identify the most likely candidate enzymes based on activity, we designed an activity-based probe with an arginine diphenylphosphonate warhead to target serine proteases with trypsin-like specificity as broadly as possible (Fig. 2A). Arg, rather than Lys, was selected for the warhead due to the observation that among 68 trypsin-like proteases with data for Arg versus Lys preference, 56 show marked preference for Arg but only one shows marked preference for Lys (Tables S1 and S2).
Because many serine proteases show greatly enhanced binding and activity toward substrates with compatible residues in extended subsites P2 – P4 (
), we incorporated the Arg diphenylphosphonate warhead into the sequence Ala-Ala-Pro-Arg. Proline was selected for the P2 position because prior studies have shown Pro backbone angles to be compatible with the S2 subsite but incompatible with S1, S3, and S1ʹ subsites of S1 family serine proteases; thus Pro in the P2 position can favor productive binding and prevent nonproductive binding configurations (
). P2 Pro is strongly preferred by thrombin and is tolerated by other trypsin-like proteases for which extended specificities have been broadly profiled, including plasmin, uPA, factor Xa, KLK4, β-tryptase, matriptase, matriptase-2, matriptase-3, hepsin, HAT, and DESC (
These specificity profiling studies further demonstrate diverse but generally permissive specificity for P3 and P4 substrate residues. We selected alanine for these positions with the rationale that its small hydrophobic side chain could be widely accommodated without introducing steric clashes. The recognition sequence was incorporated as the C-terminal portion of the amino acid sequence GTENLYFQGGAAPR-diphenylphosphonate, which also contains a TEV protease cleavage site (ENLYFQ↓G). At the N-terminus of the peptide, a PEG linker serves as a spacer connecting a biotin tag used as an affinity handle (Fig. 2A). Solid phase peptide synthesis was completed using standard automated fluorenylmethyloxycarbonyl (Fmoc)/tert-butyl (tBu) chemistry followed by cleavage from the resin, solution phase coupling to (D/L)Arg(di-Boc)-diphenylphosphonate, deprotection, and purification of the final product.
The probe interaction with a protease is expected to initiate as a substrate-like interaction in which the Arg side chain occupies the primary specificity pocket. Upon nucleophilic attack by the enzyme serine nucleophile on the phosphorus center of the diphenylphosphonate, the complex is expected to proceed through a pentacoordinate transition state leading to a stable covalent adduct with displacement of phenol (
Probe characterization verifies Ser195 specific labeling of a model catalytically active protease
We characterized the anticipated probe-protease behavior using active mesotrypsin, a human trypsin isoform, compared to a catalytically inactive mesotrypsin-S195A mutant lacking the reactive serine nucleophile. Both mesotrypsin and the S195A mutant were incubated with the probe and then samples were subjected to Western blotting to detect the biotin handle with horseradish peroxidase–conjugated streptavidin. The results showed a clear and intense band at the approximate gel mobility of mesotrypsin for the WT protein. In contrast, no signal was detected in the S195A mutant sample, demonstrating the specificity for labeling of the active site serine by the probe (Fig. 3A, loading control Fig. 3B). We further found that the active site modification by the probe was quantitative and complete, by using an activity assay with a colorimetric peptide substrate of mesotrypsin. Whereas untreated WT mesotrypsin demonstrated high turnover of the substrate, a comparable concentration of mesotrypsin incubated with the probe showed almost no residual catalytic activity (Fig. 3C). This assay further verified the specific and complete targeting of the active site serine 195 by the ABPP probe, resulting in elimination of catalytic activity.
Using mass spectrometry, we next queried the total mass difference of intact mesotrypsin modified by the probe compared to unmodified mesotrypsin. The spectra revealed a mass shift of 2160.06 Da between the test samples, within experimental tolerance of the calculated mass addition of 2160.53 Da expected for labeling with 1 mol equivalent of the probe to yield the monophenoxyphosphonyl derivative (Figs. 3, D and E and S4). As a control, we also incubated the mesotrypsin-S195A mutant with the probe and compared the resulting spectrum to that obtained from an untreated control sample. We did not detect any labeling of the catalytically inactive mesotrypsin-S195A mutant via mass spectrometry, confirming the high specificity of the probe for the active site Ser-195 (Fig. S5).
Despite the clear evidence of quantitative labeling of active mesotrypsin by the probe, initial efforts to directly detect probe labeling of Ser-195 in MS/MS sequencing of trypsin/TEV protease digests of probe-treated mesotrypsin were unsuccessful, instead revealing the unlabeled form. Further investigation revealed that the covalent probe adduct was unstable to standard mass spectrometric higher-energy collisional dissociation (HCD) fragmentation, which preferentially fragmented the probe from the peptide (Fig. S6). Subsequent modification of the MS/MS protocol to employ the softer electron transfer dissociation (ETD) fragmentation approach ultimately led to successful detection of fragment ions deriving from probe-labeled mesotrypsin, directly confirming the site of probe labeling as the active site Ser-195 (Fig. S7).
Activity-based probe detects secreted active trypsin-like serine proteases in OCCC cell cultures
Following the characterization of the probe, we developed and optimized a standardized workflow for probe labeling of secreted active proteases in cell culture. We focused on the secreted proteome because most trypsin-like serine proteases are secreted enzymes, with the remaining few being extracellularly displayed transmembrane proteases that are shed from the cell surface through proteolysis (
) (Tables S1 and S3). Cells were cultured for 24 h in a minimal volume of serum-free media with exposure to the probe during the final hour, followed by clarification and concentration of the conditioned media. The secreted proteome was recovered by methanol/chloroform precipitation to eliminate excess unreacted probe and then resolubilized. Labeled proteins were then captured via the biotin affinity handle using streptavidin coated beads, resolved by SDS-PAGE, and captured proteins were analyzed by tandem MS/MS protein identification (Fig. 4A). We examined two OCCC cell lines using this workflow, JHOC5 and JHOC9, with the aim to query their complex proteomes for secreted active serine proteases and to identify as candidate mediators of malignancy the active proteases present in both models. Western blot analysis detecting the biotin adduct revealed three distinct bands for analysis in the JHOC5-conditioned media proteome (Fig. 4B). The JHOC9 proteome displayed evidence of labeled proteins across a broader spectrum of molecular weights, from which we chose the four most intense bands for analysis (Fig. 4D). Bands of corresponding gel mobility excised from duplicate silver stained gels (Fig. 4, C and E) were analyzed using mass spectrometry. We identified two trypsin-like serine proteases that were present in both OCCC cell lines: tPA and uPA (Tables 1 and 2). In the JHOC9 cell line, we identified a number of additional proteases: kallikrein 6 (KLK6), 10, and 13, and transmembrane protease serine 11F (Table 2). Taken together, we detected common and unique active proteases in the two OCCC cell lines and next proceeded to interrogate the candidates identified in common for their role in tumor progression.
Table 1Enriched and identified active serine proteases in JHOC5 OCCC cells
Serine proteases uPA and tPA identified by ABPP promote OCCC cell invasion and proliferation
Focusing on the two active proteases that were detected in both JHOC5 and JHOC9 cells, we next conducted experiments to assess the potential contributions of these enzymes to OCCC malignancy and further define their expression and activity in OCCC models. The serine protease uPA, also known as urokinase, contributes to the activation of plasminogen to plasmin, a process involved in clot lysis, wound healing, and tissue remodeling. Previous work has shown that uPA can also be a driver of malignancy, and its expression has been linked to poor prognosis in multiple tumor types (
). To assess the relevance of uPA for tumor cell malignancy in OCCC, both cell lines were subjected to lentiviral shRNA gene silencing targeting the PLAU gene. We found that knockdown of PLAU using two different lentiviral constructs (Fig. 5, A and B) significantly impaired cellular invasion through artificial basement membrane in a standard Boyden chamber assay (Fig. 5, C and D). Additionally, knockdown of PLAU significantly impaired tumor cell proliferation compared to the control in both JHOC5 and JHOC9 cells (Fig. 5, E and F).
We next examined presence and susceptibility to activity-based labeling of different uPA protein species in JHOC9 cultures, since this model had shown most robust detection of uPA in proteomic analyses (Fig. 4D and Table 2). Cultures transfected with a control nontarget lentivirus or with a PLAU-targeted knockdown lentivirus were labeled with the activity-based probe, and then conditioned medium was analyzed directly on Western blots with detection for uPA or for the biotinylated probe. Two uPA species were detected with anti-uPA antibody, a major band representing the single-chain mature zymogen (∼49 kDa) and a minor band representing the beta-chain of proteolytically activated uPA containing the catalytic domain (∼30 kDa) (Fig. 6A). Both bands were completely absent from the PLAU knockdown culture, confirming the efficacy of knockdown at the protein level. Detection with streptavidin revealed labeling of the 30 kDa band coinciding with the catalytically active beta-chain; absence of this band in the knockdown sample further supports its identification as active uPA (Fig. 6A). No streptavidin signal was observed coincident with the 50 kDa zymogen, consistent with specificity of the activity-based probe toward proteolytically activated uPA; the very low basal level of activity that has been reported for the single-chain zymogen (
To evaluate the potential of our activity-based probe for interrogating the more complex proteome of the tumor microenvironment in vivo, we conducted a labeling experiment using ascites fluid collected from tumor-bearing mice. We have previously established the JHOC9 intraperitoneal injection model to recapitulate processes occurring during intraperitoneal metastasis of OCCC (
). Incubation of ascites fluid collected from mice-bearing JHOC9 tumors with our activity-based probe resulted in detection of several labeled protein bands including the 30 kDa band also identified by anti-uPA antibody (Fig. 6B), confirming the relevance of this active protease species in vivo.
Like uPA, tPA is also an activator of plasmin, with roles in regulation of fibrinolysis; however, the role of this protease in cancer is much less clear and may vary by tumor type. To investigate the significance of tPA for hallmarks of malignancy in OCCC, we used lentiviral shRNA constructs against tPA to knockdown the PLAT gene in both OCCC cell lines (Fig. 7, A and B), and cells were tested for the ability to invade through artificial basement membrane and to proliferate. Similarly to cells with PLAU knockdown, following PLAT knockdown, we found significant reduction of both invasive capability (Fig. 7, C and D) and proliferation (Fig. 7, E and F) compared to the controls in both cell lines. To further explore the specificity of these phenotypic effects, we also examined the impact of silencing KLK-6, a serine protease that was identified by probe labeling in JHOC9 cells but not in JHOC5 cells. RT/PCR showed that KLK6 transcripts were robustly expressed in JHOC9 cells but not detected in JHOC5 cells, consistent with proteomic results. Unlike our findings with uPA and tPA, however, silencing of KLK-6 in JHOC9 cells did not have any significant effects on invasion or proliferation (Fig. S8), thus highlighting the specific impact of uPA and tPA on these malignant phenotypes.
We next directly examined tPA protein species and susceptibility to probe labeling in JHOC5 cultures, where tPA had been most robustly detected in proteomic results (Fig. 4B and Table 1). Atypically for a serine protease, single-chain tPA possesses most of the enzymatic activity of the cleaved two-chain form; the activating cleavage, catalyzed by its product plasmin in a positive feedback loop, increases the activity of tPA only modestly (
). Conditioned medium from the probe-labeled control culture showed a band detected by anti-tPA at ∼68 kDa, consistent with the expected size of single-chain tPA (Fig. 8A); no two-chain tPA was detected. This band was also labeled by the activity-based probe, demonstrating that the activity of single-chain tPA is sufficient for reactivity with this probe. The band was absent in samples from the PLAT-knockdown culture (Fig. 8A), further confirming the band identity as tPA and demonstrating the efficacy of the knockdown at the protein level.
We also evaluated the presence and probe reactivity of tPA in ascites samples collected from tumor-bearing mice from our previously established JHOC5 intraperitoneal injection model of OCCC (
). The 68 kDa band was detected by anti-tPA in all available ascites samples and was strongly labeled by the activity-based probe as indicated by streptavidin staining (Fig. 8B), confirming the significant presence of this reactive species in the tumor microenvironment. The corroboration of both uPA and tPA protein expression, activity, and probe labeling in ascites samples from tumor-bearing mice (Figs. 6B and 8B) suggests that these mediators of malignancy are indeed relevant to the in vivo setting. These results further highlight the potential utility of our probe for directly interrogating complex tumor microenvironments.
Altogether, our data identify uPA and tPA as mediators of OCCC growth and progression, suggesting possible points of therapeutic intervention and provide an example illustrating how our activity-based probe can successfully identify trypsin-like serine proteases of interest in biological model systems.
The advent of gene expression analysis and fast progress in the field of genomics have led to increased understanding of how genetic mutations and transcriptional changes predispose to disease (
). There remains, however, a knowledge gap between the transcriptome and the functional proteome, as protein activity is shaped by posttranslational regulation through covalent modification, limited proteolysis, and interactions with activity-modifying ligands and binding partners. Differences between transcript abundance, protein abundance, and protein activity can be particularly stark for proteases, given their multiple layers of posttranslational regulation. Proteases are typically produced as zymogen precursors that persist as latent enzymes until activation by limited proteolysis (
), further titrating their actual availability to catalyze proteolytic reactions. Unsurprisingly in light of these mechanisms of regulation, it is often the case that the overwhelming majority of a protease in its biological context is present in various inactive forms. Whereas both transcriptional readouts and traditional proteomics approaches will thus fail to give an accurate picture of the protease landscape in terms of activity, an activity-based proteomic approach can address this need and identify catalytically active proteases in highly complex proteomes. The present study highlights the capability of such an activity-based approach, as we successfully detected proteases that were directly involved as drivers of malignancy in OCCC cell lines.
Some of the earliest applications of ABPP focused on the large group of serine hydrolases (
). ABPP methods were easily recruited to target these enzymes, facilitated by a conserved enzyme mechanism featuring a nucleophilic serine that forms a covalent intermediate by attacking a substrate electrophilic center. Activity-based probes were designed with highly electrophilic and reactive warheads that undergo nucleophilic attack by serine to form stable adducts, thus detecting a plethora of serine hydrolase targets including lipases, esterases, amidases, and serine proteases (
). Warhead design initially exploited relatively promiscuous fluorophosphonates as the preferred electrophile. Such an activity-based probe against serine hydrolases was used to demonstrate that MDA-MB-231 breast cancer cells implanted in vivo show differential hydrolase activity that is influenced by the host stroma, including striking increases in activity of both tPA and uPA (
). Use of a similar probe identified the lipolytic enzyme monoacylglycerol lipase as highly expressed in cancer cells and discovered its function as a liberator of free fatty acids within tumor cells, which in turn are used as an energy source and lead to increased tumor cell survival and proliferation (
) have synthesized a series of probes with alternative substituted phenylphosphonate warheads to fine-tune selectivity toward different subsets of intracellular serine hydrolases, in addition making use of a shorter linker and CuAAC click-chemistry to produce highly cell permeable probes.
A smaller subset of serine hydrolases can be more specifically interrogated by probes with a diphenylphosphonate-containing warhead, which reacts more uniquely with serine proteases. Irreversible serine protease inhibitors featuring this functional group were pioneered by Powers et al. (
) enhanced selectivity toward uPA further by employing a guanidinylated benzyl diphenylphosphonate group instead of the natural lysine or arginine at the P1 position and additionally optimizing the P4 position. Here, we aimed instead to more broadly target the trypsin-like serine proteases, which critically house within the S1 specificity pocket either aspartic acid or glutamic acid, leading to substrate specificity for peptide cleavage after lysine or arginine. We chose to design the probe with a diphenylphosphonate warhead with arginine in the P1 position, as preferred by the great majority of human trypsin-like proteases in the MEROPS database (Tables S1 and S2) (
) and additionally to populate the P2-P4 positions with an extended recognition motif well tolerated by many serine proteases. This probe design was validated by our labeling and identification of a variety of trypsin-like serine proteases secreted by JHOC5 and JHOC9 OCCC cell lines.
While the Ala-Ala-Pro-Arg recognition motif was designed to be broadly reactive with trypsin-like serine proteases, reactivity with individual proteases may vary according to their individual substrate specificities at the P1–P4 subsites. It is possible that we may have failed to capture and detect some active secreted trypsin-like proteases either due to low abundance or because they were poorly reactive with the extended subsite residues of the probe design. In future work, it may be possible to achieve enhanced sensitivity with this or similar probes by adapting our methodology to a gel-free proteomic workflow. Our probe further incorporated a tandem orthogonal proteolysis design, first described by Speers et al. (
), which allows for interrogation of the labeled protein as well as the site of labeling by sequential digestion with trypsin and TEV protease and offers potential advantages for sensitivity and specificity of detection. It may be feasible to modify our probe for intracellular labeling by substituting the TEV recognition site, linker and biotin for the smaller alkyne functionality, and employing click chemistry to introduce these elements after labeling, thus further extending the reach of the method (
Our findings included the discovery that zymogen and activated uPA are secreted by OCCC cell lines, where uniquely the activated form was detected by our activity-based probe. The physiological role of uPA is the activation of plasminogen to plasmin, a pivotal serine protease that degrades many extracellular matrix components during tissue remodeling and activates zymogen forms of other enzymes including matrix metalloproteinases (
). Both plasmin and matrix metalloproteinases have been implicated in cancer, and uPA as a major activator has drawn interest for its role in cancer development and progression with emphasis on metastatic spread (
), providing a possible means of therapeutic intervention. Furthermore, an intricate interplay between uPA, uPAR, and integrins including αvβ3 has been shown to be involved in the regulation of tumor cell invasion (
). Although less has been known about the potential involvement of these mechanisms in the progression of OCCC specifically, our present findings demonstrate uPA to promote both OCCC cell proliferation and invasion, implicating uPA as a significant driver of tumor growth and progression in OCCC.
We also identified single-chain tPA as an active protease secreted by OCCC cells. As a plasminogen activator, tPA is a primary mediator of fibrinolysis and thrombolysis in the vascular system through generation of active plasmin; it has also been widely used in recombinant form as a therapy to lyse clots in acute ischemic stroke (
). In our experiments using OCCC cell culture medium, tPA basal activity was sufficient for activity-based probe labeling (Fig. 8A). Notably, however, we did observe markedly more intense probe labeling of tPA in complex ascites samples (Fig. 8B), which might be explained from prior observations of human ovarian cancer and mouse models showing influx of fibrinogen and accumulation of fibrin degradation products in ovarian cancer ascitic fluid (
). Thus, allosteric stimulation of tPA activity in ascites by incompletely digested fibrin fragments may enhance reactivity with our activity-based probe; this possibility will require further investigation.
The role of tPA has been little studied in most forms of cancer, but substantive data support a malignancy promoting role for tPA in pancreatic cancer and glioma. In pancreatic cancer models, tumor cell expression of tPA was shown to promote cell proliferation and tumor growth through mechanisms involving epidermal growth factor signaling (
). In glioma, tPA has been implicated as a key factor in malignancy, since its knockdown in glioma-initiating cells, or its downregulation by miRNA-340, inhibits both proliferation and invasion in vitro and tumor growth in vivo (
). While the potential significance of tPA in ovarian cancer previously has not been a major focus of study, the present study implicates tPA along with uPA as a driver of invasion and proliferation in OCCC. Our work suggests that both of these plasminogen activators, along with downstream pathways activated in common by these enzymes (
), may represent potential targets in OCCC. Additional work is needed to identify the most effective therapeutic approach to reduce the protumorigenic effect of both uPA and tPA.
In summary, we show that our arginine diphenylphosphonate activity–based probe can successfully profile serine protease activity in a complex secreted proteome. Interrogation of OCCC cells resulted in the detection of multiple target proteases for which we demonstrate active involvement in malignant phenotypes that contribute to cancer growth and progression. Together, our findings highlight the potential uses for ABPP in the rapidly expanding field of proteomics to aid in the identification of therapeutic targets and guide the future development of targeted therapy.
Activity-based probe design and synthesis
The activity-based probe was custom synthesized in collaboration with Peptides International (Cat# PCS-31822-PI, Lot# 003066C, formula C103H152N23O30SP).
Solid phase synthesis
Biotin-[NH-(CH2)5-COO]-dPEG4-Gly-Thr(tBu)-Glu-Asn(Trt)-Leu-Tyr(tBu)-Phe-Gln(Trt)-Gly-Gly-Ala-Ala-Pro-OH was synthesized (0.4 mmol scale) by standard automated (Prelude, Protein Technology Inc) Fmoc/tBu chemistry using a solid support H-Pro-2-ClTrt resin (substitution: 0.58 mmol/g). The coupling of the amino acids in dimethylformamide (DMF) was carried out with diisopropyl carbodiimide and oxima in DMF activation chemistry for 3 h. The Fmoc group was removed by treating the peptide resin with 20% piperidine/DMF (2×, 2 min and 20 min each). The resin was thoroughly washed with DMF, dichloromethane, and methanol and dried to obtain a constant weight. The dried peptide resin was then cleaved with 1% TFA in dichloromethane (3×, 30 min each). The resin was filtered off and the filtrate was evaporated and the residue was triturated with isopropyl ether. The protected peptide was centrifuged, washed with isopropyl ether (2×), and dried under high vacuum.
Solution phase coupling
The protected peptide and 1-Hydroxy-7-azabensotriazole were dissolved in DMF, and diisopropyl carbodiimide, (D/L)Arg(di-Boc)-diphenylphosphonate (Peptides International), and N-methylmorpholine were added. After 18 h, the reaction was completed and the peptide was precipitated by adding H2O and washed two times with isopropyl ether, after which the solid material was dried under vacuum.
Final cleavage, purification, and validation
Complete cleavage of the protecting groups was carried out with a cleavage cocktail containing TFA (87.5%), triisopropylsilane (2.5%), H2O (5.0%), and anisole (5%). After 3 h at ambient temperature, the product was precipitated with isopropyl ether, washed with isopropyl ether (3×), and dried under vacuum. The peptide was then purified by using a C-18 reversed phase column and characterized by LC-MS. The average molecular weight was determined to be 2255.53, and the probe showed 92.6% purity of the racemic mix of D + L isomers as determined by HPLC.
Expression of recombinant mesotrypsin and catalytically inactive mesotrypsin-S195A
Recombinant human mesotrypsin and catalytically inactive mutant mesotrypsin-S195A were expressed as zymogen precursors in Escherichia coli, isolated from inclusion bodies, refolded, purified by ecotin affinity chromatography, proteolytically activated using bovine enteropeptidase, and purified by benzamidine affinity chromatography as described previously (
). Enzyme concentrations were routinely measured by Nanodrop (NanoDrop One Thermo Scientific) using a calculated extinction coefficient at 280 nm of 41,535 M−1 cm−1.
Western blot to detect probe labeling of mesotrypsin
Lyophilized aliquots of the activity-based probe were reconstituted in 25 μl dimethyl sulfoxide (DMSO) to make a stock concentration of 17 mM. Catalytically active mesotrypsin or the catalytically inactive mutant mesotrypsin were incubated at a concentration of 4 μM with 4-fold molar excess of the probe for 1 h at 37 °C, in a final volume of 100 μl in serum-free, phenol red–free Dulbecco’s modified Eagle’s medium (DMEM)/F12 media (Cat 21041025, Gibco). Following the incubation, samples were run under denaturing conditions on a 4 to 20% gradient SDS-PAGE (BioRad). Protein was transferred (20% methanol transfer buffer) onto nitrocellulose membrane. Following the transfer, the membrane was blocked for 1 h in 5% bovine serum albumin (BSA) TBST. Streptavidin-HRP (CST #3999, dilution 1:4000) was added to the membrane in 5% BSA TBST for 1 h at room temperature. Following three wash steps, Clarity Western ECL (BioRad) was added for 5 min prior to exposure of the membrane to film.
Mesotrypsin activity assay
Proteolytic activity of mesotrypsin and mesotrypsin labeled with the ABPP probe were determined as previously described (
). Briefly, enzyme at a concentration of 4 μM was incubated with 2.5 fold excess of probe or DMSO vehicle control (final DMSO concentration 0.11%), for 1 h at 37 °C, in DMEM/F12 phenol red–free medium. The colorimetric peptide reporter substrate Z-GPR-pNA (Sigma) was prepared as a 40× stock solution in 100% DMSO. Using the Varian Cary-100 spectrophotometer, the assays were carried out at 37 °C. Assay buffer (760 μl; 100 mM Tris–HCl pH 8.0, 1 mM CaCl2) and substrate (20 μl) were mixed and equilibrated in cuvettes prior to reaction initiation. Enzyme + probe (20 μl) or enzyme + DMSO (20 μl) were added to start the reaction. The final reaction concentrations were as follows: enzyme at 0.5 nM, probe at 1.25 nM, and substrate at 100 μM. Reactions were followed for 5 min, and rates were determined from the absorbance increase upon the release of p-nitroaniline (ε410 = 8480 M−1 cm−1).
Intact mass analysis with LC-MS using high resolution mass spectrometry
Mesotrypsin and catalytic mutant mesotrypsin-S195A at a concentration of 4 μM were incubated with 17 μM probe or DMSO control (final DMSO concentration 0.11%) for 1 h at 37 °C, in DMEM/F12 phenol red–free media at a final volume of 100 μl. LC-MS using a Thermo Scientific Orbitrap Elite Hybrid Mass Spectrometer (Thermo Fisher Scientific) coupled to a Thermo Ultimate 3000 RSLC HPLC system running a ProSwift RP-4H capillary monolithic column were used for capturing of data. The Elite was set to scan 500 to 3000 m/z at a resolution setting of 120,000 at 400 m/z. The spectra were deisotoped to give the intact MH+ mass using the Xtract function in Xcalibur.
Identification of the probe-labeled mesotrypsin peptide by nano-LC/MS using multiple fragmentation methods
Mesotrypsin (concentration 4 μM) was incubated with the probe (10 μM) for 1 h at 37 °C, and then samples were resolved by SDS-PAGE under denaturing conditions. The gel bands were destained and the gel pieces were dehydrated in acetonitrile, followed by reduction with rehydration in 50 mM tris(2-carboxyethyl)phosphine for 45 min at 60 °C. The reduced samples were then dehydrated in acetronitrile followed by rehydration in 25 mM iodoacetamide for alkylation at room temp for 45 min, followed by dehydration with acetonitrile. The proteins were digested using 0.25 μg trypsin (Promega) in 50 μl 25 mM Tris pH 8.2 and 0.0002% Zwitt 3-16 and incubated for 20 h at 37 °C. The digest solution was then subjected to TEV protease cleavage by adding 50 μl of TEV protease solution (0.0666 U/μl in 1× TEV buffer, Thermo) for 3 h at 37 °C. The extracted peptides were analyzed by nano-flow liquid chromatography electrospray tandem mass spectrometry using a Thermo Scientific Orbitrap Elite Hybrid Mass Spectrometer (Thermo Fisher Scientific) coupled to a Thermo Ultimate 3000 RSLCnano HPLC system with a 100 μm × 25 cm PicoFrit column packed with Agilent Poroshell EC C18 solid phase. The Elite mass spectrometer experiment was set to perform a FT full scan from 600 to 1600 m/z with resolution set at 120,000 (at 400 m/z), followed by orbitrap HCD MS/MS or ETD MS/MS scans on the top five ions set up for product ion–triggered data dependent MS3 on the neutral loss of the probe mass of 640.2966. The FTMS AGC target was set to 3e6 and the orbitrap MSn target was set to 4e5 with a max ion inject times of 100 ms for both. Dynamic exclusion was set to one and selected ions were placed on an exclusion list for 30 s. The raw files were analyzed using PEAKS software (Release Version 10) allowing for probe mass addition of 621.28 at serines, oxidation of methionine, and carbamidomethyl cysteine as variable modifications. A manually assembled small database of 118 human serine proteases which included human mesotrypsin and mesotrypsin-S195A mutant was searched to find probe-modified peptides, which were then manually validated. Search parameters used semi-specific trypsin specificity allowing for two missed cleavages with the precursor mass tolerance of 10 ppm and fragment tolerance of 0.02 Da. The protein identification criteria required a two peptide minimum and a false discovery rate (FDR) less than 1%. PEAKS estimates FDR using a decoy-fusion approach where a corresponding decoy sequence is appended to each sequence in the database. The matched probe-labeled mesotrypsin peptides gave peptide confidence scores (-10lgp) that were higher than the confidence scores removed when 1% FDR filter was applied.
HCD fragmentation led to the discovery of 11 b-type ions showing the probe attached to the peptide (Fig. S3), however, the most intense fragment detected was at 640.30D, which represents the probe by itself as it is removed from the peptide due to the fragmentation method. Using the more gentle ETD fragmentation approach, we identified a total of 19 c-type ions and no high intensity fragment at the size of the probe as detected before (Fig. S4).
Cell culture and reagents
JHOC5 and JHOC9 cells were purchased from RIKEN BioResource Research Center and cultured following supplier protocols in DMEM/F12 media with 1% NEAA (Gibco, Invitrogen) and JHOC5 cells with 10% fetal bovine serum and JHOC9 cells with 20% fetal bovine serum (Gemini Bio-Products), at 37 °C and humidified atmosphere with 5% CO2.
Diminazene aceturate was purchased from Cayman Chemical Company (cat. #18678), gabexate mesylate from Sigma (SML2964-5MG), and hydroxystilbamidine CID 16212515, from NCI-DTP.
ABPP in-situ labeling and purification
Cells were plated in the morning in complete media to allow attachment during the day; seeding was optimized to reach near confluency within 24 h. To collect sufficient secreted protein, cells were seeded into three 10 cm plates per condition. In the evening of the day of seeding, cells were washed 3× with 4 ml PBS followed by addition of 10 ml serum-free, phenol red–free media. The following morning, the PBS washes were repeated and finally 3 ml of serum-free, phenol red–free media was added for 24 h, to generate the conditioned media and accumulate secreted proteins.
The probe was resuspended in DMSO (stock solution 17 mM). All experimental procedures involving the probe were exclusively handled in LoBind protein tubes (Eppendorf). The probe was added directly to cell cultures in conditioned media to achieve final probe concentration of 5 μM and mixed by pipetting. DMSO was used as a vehicle control. The cells were returned to the incubator for 1 h at 37 °C. Following incubation, the conditioned medium was collected and spun 5 min at 1000 rpm at 4 °C to separate cell debris. The supernatant was removed and pooled from the three 10 cm plates; final volume of conditioned medium was 9 ml per test condition. Protease inhibitor (9 μl of 1000× stock Sigma p8340) and EDTA (final concentration 1 mM) were added to the supernatant. The conditioned medium was then concentrated using a 3K MWCO concentrator (Amicon, Millipore) to a final volume of 100 μl.
The concentrated conditioned media were mixed with four parts methanol, one part chloroform, and three parts water to precipitate proteins. Samples were briefly centrifuged and the upper aqueous layer was carefully removed and discarded, leaving the protein at the phase interface. An additional four parts of methanol was added and mixed, followed by high speed centrifugation for 5 min. Supernatant was carefully removed and the remaining protein pellet was briefly air dried.
Resolubilization and binding to beads
The protein pellet was solubilized in 100 μl 2.5% SDS PBS and heated for 5 min at 60 °C, followed by centrifugation at 6500g at room temperature to pellet residual insoluble protein. If a sizeable pellet remained, solubilization was repeated with an additional 50 μl 2.5% SDS PBS and incubation at 90 °C. Supernatants were combined and PBS was added to reach a final SDS concentration of 0.2%. Magnetic streptavidin beads (100 μl; Thermo Fisher, Pierce, #88816) were prepared as per manufacturer’s instruction. Storage supernatant was removed from the beads and the protein solution in 0.2% SDS PBS was added to the beads and incubated under rotation for 3 h at room temperature to allow streptavidin–biotin interaction. Following a brief spin to remove beads trapped in the lid, the tubes were placed on magnetic stands and washed 1× with 0.2% SDS PBS, 5× with 1 ml PBS, and 5× with 1 ml of H2O. Samples were then stored at 4 °C until further processing. Samples were used for mass spectrometry directly or, alternatively, were resolved by SDS-PAGE and then individual bands excised for MS protein identification (
Cellular RNA was isolated following the manufacturer’s protocol, using TRIzol reagent (Invitrogen). RNA concentration was determined by absorbance at 260/280 using the Nanodrop One spectrophotometer. Reverse transcribed cDNA was synthesized using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Reverse transcribed cDNA samples were analyzed using quantitative real-time PCR according to the manufacturer’s protocol on the ABI QuantStudio 7 Flex Real-Time PCR System. Taqman assays for PLAT (Hs00263492), PLAU (Hs00170182), PRSS1 (Hs00605631_g1), PRSS2 (Hs00828418_gH), PRSS3 (Hs00605637), KLK6 (Hs00160519_m1), ACTB (Hs01060665_g1), and GAPDH (Hs99999905_m1) were obtained from Applied Biosystems and run over 40 cycles. Data were analyzed using QuantStudio Real-Time PCR Software (Applied Biosystems).
Lentiviral shRNA against PLAT and PLAU were obtained from the MISSION TRC1 and 1.5 libraries (Sigma). Knockdown efficiency of each gene was tested using five different constructs, and the two with the highest efficiency were selected for further experiments. Selected constructs targeting PLAT included NM_000930.2-1864c1c1 (KD1) and NM_000930.2-946s1c1 (KD2) and for PLAU included TRC1 NM_002658.1-1235s1c1 (KD1) and NM_002658.1-125s1c1 (KD2). For KLK6, KD1: NM_002774.2-297s1c1 and KD2: NM_002774.2-590s1c1. All experiments included control cultures transduced with nontarget shRNA control constructs not recognizing any human genes. Virus for transduction was produced using HEK 293FT cells following supplier protocols. For PLAU, PLAT, and KLK6 KD using viral transduction, 8 × 105 cells of JHOC5 or JHOC9 were plated in 10 cm culture dishes to adhere overnight. Culture media were replaced with a mixture of 2.4 ml lentivirus-containing conditioned media, 3.5 μg/ml polybrene (EMD MILLIPORE Merck KGaA) and 3.6 ml complete culture media. After a 24 h incubation period, the supernatant was removed and replaced with complete media and 2 μg/ml (JHOC9) or 1 μg/ml (JHOC5) puromycin (Corning) for selection.
JHOC5 and JHOC9 cells were transduced with PLAT, PLAU, or KLK6 lentiviral shRNA constructs prior to the invasion assay. Cells were split 1:1.5 the day before the assay. On the day of the assay, BD BioCoat Matrigel Invasion Chambers (cat # 354480) were coated with 0.5 μg per well of Corning Matrigel Basement Membrane Matrix (cat # 354234) in serum-free DMEM/F12 media (cat # 10565-018 (Gibco)) and placed to polymerize for 4 to 5 h at 37 °C and 5% CO2. Cells were trypsinized and pelleted before resuspension in 1 ml of 0.1% BSA/serum free DMEM/F12 media and counted using a Countess automated cell counter (Invitrogen). NIH/3T3 cell–conditioned serum-free medium (750 μl; DMEM supplemented with 50 μg/ml ascorbic acid) was used as chemoattractant. Cells were seeded in four replicates on top of the insert at a density of 5 × 104 (JHOC5) or 1.5 × 105 (JHOC9) cells (1 × 105 JHOC9 cell were used for the KLK6 invasion assay) in 500 μl final volume. For inhibitor treatment studies, 1.5 × 105 WT JHOC9 or 1 × 105 JHOC5 cells were seeded in the presence of 0, 1 μM, 10 μM, or 100 μM diminazene, 100 μM or 500 μM gabexate mesylate, or 10 μM or 100 μM hydroxystilbamidine (dissolved in DMSO; final DMSO concentration 1%) in 500 μl final volume. Cells were allowed to invade for 18 h (24 h for the diminazene and KLK6 experiments) at 37 °C in 5% CO2. At the end of the assay, noninvading cells were removed from the insert by scrubbing with a cotton swab, and cells on the lower surface of the filter were fixed in 100% methanol for 30 min at −20 °C, followed by 1 h room temperature staining with 0.1% crystal violet in 20 mM MES, pH 6.0. Stained filters were photographed at 2× magnification and cells were counted using Image-Pro 7.0 or Infinity Analyze software.
Plate reader–based proliferation assays were performed on cells previously transduced with PLAT, PLAU, or KLK6 KD virus. The cells were under selection for 5 days after virus addition prior to setting up the experiment. The day before the assay, 5000 cells/well in 100 μl complete media were plated into 96 well plates and incubated for 18 h. For diminazene, gabexate mesylate, and hydroxystilbamidine treatment experiments, cells were seeded at 5000 cells/well and incubated with vehicle only (DMSO at final concentration of 1%) or with each drug at the indicated final concentrations for 18 h in complete media before commencement of the assay. The assay was carried out following the manufacture’s protocol (Promega CellTiter Glo 2.0 viability assay). Cells were lysed and incubated with Promega reagent and luminescent signal was acquired following the provided standard Promega protocol (Veritas, Turner BioSystems). Data were analyzed using GraphPad Prism 8.0.
Western blot analysis for tPA and uPA
JHOC9 cells transduced with PLAU-targeted or nontarget control lentiviral shRNA, or JHOC5 cells transduced with PLAT-targeted or nontarget control lentiviral shRNA, were seeded and cultured in 10 cm plates as detailed above for ABPP in-situ labeling. After 24 h in 3 ml of serum-free phenol red–free media at near confluency, cultures were labeled with 5 μM activity-based probe at 37 °C for 1 h as described. Supernatant was clarified and cell debris removed. Samples were then evaluated by Western blot for uPA (GeneTex GTX10046, dilution 1:1000; donkey anti-rabbit secondary GE Healthcare NA934V, dilution 1:3000) or tPA (Abcam #Ab227069, dilution 1:1000; donkey anti-rabbit secondary GE Healthcare NA934V, dilution 1:1000). Probe labeling was assessed using HRP-conjugated streptavidin (Cell Signaling #3999, incubation 1:7000, 1 h at RT). Membranes were washed three times and developed with Clarity Western ECL (BioRad) for 5 min before exposure to film.
Animal model, ascites collections, and Western blotting
Animal studies were conducted with approval by the Mayo Clinic Institutional Animal Care and Use Committee (IACUC) (protocol A00002844-17) according to protocols that we have described in detail previously (
). Female Nod/SCID mice (6–8 weeks old) received an intraperitoneal injection with either JHOC9 (2 × 106) or JHOC5 (5 × 105) tumor cells injected into the lower right quadrant of the abdomen. JHOC9 cohorts were euthanized by CO2 asphyxiation at 15 weeks and JHOC5 cohorts at 7 weeks due to tumor growth kinetics and moribund endpoints. Ascitic fluid was retrieved from the abdominal cavity and centrifuged to pellet the cellular component. The supernatant was collected and stored at −20 °C until use. Clarified supernatant (100 μl) was incubated with 16 μM activity-based probe for 1 h at 37 °C to achieve protein labeling. Samples were then evaluated by Western blot for uPA (Invitrogen MON-U-16-02, dilution 1:1000; goat anti-mouse secondary Thermo Scientific #31432, dilution 1:1000) or tPA (Abcam #Ab227069, dilution 1:1000; donkey anti rabbit secondary GE Healthcare NA934V, dilution 1:1000). Evidence of probe labeling was assessed using HRP-conjugated streptavidin (Cell Signaling #3999, incubation 1:7000, 1 h at RT). Membranes were washed three times and developed with Clarity Western ECL (BioRad) for 5 min before exposure to film.
The authors declare that they have no conflicts of interest with the contents of this article.
We would like to thank the Mayo Clinic Medical Genome Facility Proteomics Core for support and guidance in the development of the protocol. The Mayo Clinic Medical Genome Facility Proteomics Core is a shared resource of the Mayo Clinic Cancer Center (NCI P30 CA15083).
C. M., D. C. R., and E. S. R. conceptualization; C. M., B. M., R. E., and E. S. R. methodology; C. M., A. H., M. C., and B. M. investigation; C. M., A. H., M. C., and B. M. formal analysis; C. M. and B. M. data curation; C. M., A. H., M. C., and E. S. R. validation; C. M. and E. S. R. project administration; C. M., M. C., and E. S. R. visualization; C. M. and E. S. R. writing–original draft; C. M., A. H., M. C., B. M., R. E., D. C. R., and E. S. R. writing–review and editing; B. M., R. E., and E. S. R. resources; E. S. R. supervision; E. S. R. funding acquisition.
Funding and additional information
This work was supported by National Institutes of Health grants R21 CA226302, R01 CA154387, and R01 GM144393 to E. S. R. C. M. acknowledges support from the Mayo Clinic Graduate School of Biomedical Sciences. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.