Partial Proteasome Inhibitors Induce Hair Follicle Growth by Stabilizing β-Catenin

Authors


Abstract

The activation of tissue stem cells from their quiescent state represents the initial step in the complex process of organ regeneration and tissue repair. While the identity and location of tissue stem cells are becoming known, how key regulators control the balance of activation and quiescence remains mysterious. The vertebrate hair is an ideal model system where hair cycling between growth and resting phases is precisely regulated by morphogen signaling pathways, but how these events are coordinated to promote orderly signaling in a spatial and temporal manner remains unclear. Here, we show that hair cycle timing depends on regulated stability of signaling substrates by the ubiquitin-proteasome system. Topical application of partial proteasomal inhibitors (PaPIs) inhibits epidermal and dermal proteasome activity throughout the hair cycle. PaPIs prevent the destruction of the key anagen signal β-catenin, resulting in more rapid hair growth and dramatically shortened telogen. We show that PaPIs induce excess β-catenin, act similarly to the GSK3β antagonist LiCl, and antagonize Dickopf-related protein-mediated inhibition of anagen. PaPIs thus represent a novel class of hair growth agents that act through transiently modifying the balance of stem cell activation and quiescence pathways. Stem Cells 2014;32:85–92

Introduction

Despite extensive interest in the identification of stem cells and their location within vertebrate tissues, a key remaining issue is the nature of signals that regulate the activation of stem cells in situ [1, 2]. Identification of such signals would provide therapeutic targets for organ regeneration without the use of ex vivo cell/tissue replacement technologies. Spontaneous reactivation that allows the constant and measured replacement of key portions of the tissue operates in organs such as the hematopoetic or skeletal system [3]. By contrast, similar signals also activate stem cells to provide replacements during injury or to increase tissue size during periods of stress [4]. Regeneration, once thought to be lacking in higher mammals, contributes to tissue repair and renewal in varying degrees in different tissues [5]. The pathways or mechanisms guiding these distinct processes have been the subject of intense investigation.

One of the best-studied stem cell systems is the vertebrate hair, where hair cycling between growth and resting phases is precisely regulated [6]. The first postnatal cycle lasts approximately 6 weeks in mice and results in synchronized growth (anagen), degeneration (catagen), resting (telogen), and hair ejection (exogen) phases. Within a given inbred murine line of similar mass, the timing of each portion of the first two cycles occurs within 8 hours, making the system ideal for studying factors that control stem cell timing. The intrinsic murine hair cycle can be influenced by environmental factors affecting survival and reproduction, including day length, temperature, and hormonal state [7]. While there are notable differences with human hair cycling, the similarities make murine hair studies relevant for human hair growth therapeutics.

Studies of known signaling pathways in the skin have defined a general hierarchy in the control of hair follicle stem cell activation in two distinct competing systems, one for stem cell activation and another for quiescence. Providing signals that activate stem cells into anagen are the Wnt, Sonic hedgehog, and TGF-β pathways. Transgenic animals expressing a stabilized β-catenin or SHH ligand induces neofolliculogenesis, while those expressing Dickopf related protein 1 (DKK) or dominant negative LEF1 demonstrate a paucity of new hair follicle growth [8-10]. By contrast, an opposing and independent set of signaling pathways appears to maintain the stem cells in telogen [11, 12]. Enforcing quiescence involves two pathways, Bone Morphogenetic Protein (BMP) and NFAT. BMP signals emanating from the inner bulge and dermal papilla inhibit anagen [13]. Telogen-phase hair follicle stem cells lacking the ability to respond to BMPs immediately re-enter anagen, reinforcing the central importance of BMP signaling. Similar phenotypes are seen with the removal of Nfatc1, a transcription factor transcriptionally regulated by the BMP pathway [13]. While major pathways controlling stem cell behavior have been identified, how these events are coordinated to promote orderly signaling in a spatial and temporal manner remains unclear.

The coordination of many complex biological events such as the cell cycle and inflammation is controlled by the ubiquitin-proteasome system (UPS), which transfers ubiquitin to key substrates to alter functions including enzymatic activity, protein-protein interactions, and protein stability [14]. An emerging theme is the central role for protein polyubiquitination in controlling the timing of morphogenic signaling including the cell cycle, immune response, and reproduction [15, 16]. Each of these transitions are characterized by the presence of opposing signaling pathways that maintain the cell in a metastable state of quiescence, and the abrupt change in the stability of one or more of the driving substrates is what serves to progress the cell to the next phase. Because of the similarity with the hair cycle, we hypothesized that the anagen-telogen transition of the hair cycle, like the cell cycle and other biological timing events, may be proteasome-dependent. Here, we show that the UPS differentially regulates key substrates of the hair cycle and controls the maintenance of the resting phase. Partial proteasomal inhibitors (PaPIs) appear to prevent the destruction of β-catenin while maintaining other regulators of the hair cycle, resulting in more rapid hair growth due to lengthened anagen and dramatically shortened telogen. PaPIs thus could represent a novel class of hair growth agents that act through transiently modifying the balance of stem cell activation and quiescence pathways.

Materials and Methods

Animal Strains

TOP-GAL transgenic mice that report the Wnt pathway were provided by Roel Nusse (Stanford, CA). Msx2-LacZ mice that report aspects of the BMP pathway were kindly provided by Cheng-Ming Chuong (Los Angeles, CA). Ubiquitin-G76VGFP mice that report proteasome activity on ubiquitinylated substrates were purchased from Jackson Lab (strain # 008111).

K5-Gli 1 mice were generated as previously described [18]. K5-Gli 1 tumor prone mice (12 in each group) were treated with PAPIs or vehicle every 3 days for 3 months until the spontaneous onset of tumors. Animals were assessed for alopecia or tumor formation daily.

Topical Treatments

Treatments were performed as indicated according to Administrative Panel of Laboratory Animal Care-approved protocols at Stanford University.

PaPI Treatment

To ascertain the effects of blocking the proteasome, we treated shaved, 25–30-g, postnatal Day 28 Female C57BL/6 mice (Jackson Laboratory, Bar Harbor, ME, Jax.org) with either Bortezomib 1% Millennium (Boston, MA, millennium.com), NeoSH101 2% Neosil (Emeryville, CA), or vehicle only (50:50 [vol/vol] of ethanol and propylene glycol) every other day for 2 weeks on the dorsal back. No ulceration or scaling was observed. We used several measures to determine hair follicle staging. These included the change in skin pigmentation and epidermal thickness that were confirmed by histology. To measure the effect on hair growth, we shaved and bleached mice as in [9]. These mice were then treated and the extent of guard hair pigmented hair growth used as an indication of hair growth. Similar effects were seen on the growth of zigzag, awl, and auchenne hairs as well.

LiCl Treatment

LiCl or control NaCl 1% (wt/vol) (50:50 [wt/vol] of ethanol and propylene glycol) was applied as above for PaPIs for 2 weeks. No adverse effects on differentiation were seen.

Adenoviral Injections

Adenovirus-expressing DKK was generated as previously described in [28]. Particle forming units (3.3 × 108) of adenovirus-expressing β-Catenin or control GFP protein (Vector Biolabs) were injected subcutaneously at a concentration of 3.39 particle forming units/ml into clipped dorso-medial skin along the cephalo-caudal axis of the mice indicated. The injections were angled toward the head and produced a 9-mm wheal. Ascertainment of hair follicle stage followed guidelines above. Biopsies were taken as indicated after 1 week of PaPI or vehicle treatment.

Antibodies

The following antibodies were used for immunostainings. β-Catenin (Sigma, sigmaaldrich.com), ABC (Millipore, Billerica, MA, Millipore.com), CD34 (BD Pharmingen, San Jose, CA, BDbioSciences.com), Keratin14 (Covance, Covance.com), Keratin 10 (Covance), AE13 (Abcam), Keratin17 (gift of P. Coulombe), pSmad 1/5/8 (Cell signaling, Boston, MA, cellsignal.com), NfatC1 (Santa Cruz Bio, Dallas, Tx, scbt.com), NFKB p50 (Santa Cruz Bio), and Caspase 3 (Cell Signaling). Sections were stained using standard immunohistochemical staining procedures.

Cell Culture

Mouse keratinocytes were explanted from wild-type or UBG76VGFP mice using standard methods [18]. PaPIs, including MG132, were added at dosages used on animals.

Results

Previous studies of the signaling pathways regulating the hair follicle cycle have shown the dependence on regulated protein stability by the UPS as a major control point. These observations suggested that the UPS system may regulate the timing of the hair cycle. To examine the role of the proteasome in regulating cycle progression, we treated wild-type mice with topical versions of two partial proteasome inhibitors (PaPIs), Bortezomib (PaPI1) and NEOSH1 (PaPI2), both of which inhibit the chymotrypsin protease activity of the proteasome. We determined the efficacy of these PaPIs, using a mouse line expressing ubiquitin-green fluorescent protein (GFP) (UbG76VGFP, Fig. 1). Constitutive activity of the proteasome in this line allows the positive readout of proteasome inhibition through the induced presence of GFP fluorescence [17]. Application of PaPIs to mouse skin during telogen revealed marked accumulation of fluorescence throughout the follicular epithelium and, to a smaller degree, in the dermal cells within 6 hours after application (Fig. 1A, 1B). In order to quantify the maximal effects, levels of GFP were assessed by Western blot after drug application. At 1% concentration, topical PaPI1 (Bortezomib) induced a fivefold increase in Ub-GFP protein levels, similar in magnitude to the maximal induction of MG132, a potent proteasome inhibitor, in cultured Ub-GFP keratinocytes (Fig. 1C, 1D). No stage-dependent differences in permeability or activity were seen, as anagen and telogen skin both had robust GFP immunoreactivity (not shown). We conclude that PaPIs effectively inhibit the proteasome throughout the skin in a stage-independent manner.

Figure 1.

Topically applied PaPIs partially inhibit the proteasome. (A, B): Fluorescent micrographs of UbG76VGFP-expressing mice where GFP fluorescence indicates proteasome inhibition. (A): Control vehicle-treated skin sections (10 µM sections) and (B) PaPI-treated sections showing relatively uniform distribution of proteasome inhibition. (C, D): Quantitation of inhibition by Western blot for GFP in two independent samples of vehicle (V) or PaPI (P)-treated tissue or in explanted keratinocytes (in vitro). WT mice lack the UbG76VGFP allele. (D): Quantitation of PaPI effect by densitometry. Abbreviations: GFP, green fluorescent protein; PaPI, partial proteasomal inhibitor.

To determine the effects of PaPIs on skin, we treated murine skin topically with PaPIs starting in the middle of the first anagen cycle every other day for 2 weeks. PaPI-treated skin demonstrated remarkable enhancement of hair growth (Fig. 2A, 2B) through the shortening of the resting (telogen) phase and increased hair growth during anagen. We measured the stage of the hair cycle using skin pigmentation, and correlated that with hair counts. During the anagen growth phase, the length of the guard hair was increased by approximately 20% relative to vehicle control for both PaPIs we tested (Supporting Information Fig. S1, p < .01). This was true at multiple sites along the dorsal back, ruling out a local penetration effect. Interestingly, the length of anagen was relatively unchanged (Fig. 2C) indicating that the growth rate of the hair was increased. Moreover, PaPI treatment also affected the enforcement of the resting or telogen stage. In untreated animals, quiescent hair follicle stem cells in the second telogen typically remain dormant for 21 days until they asynchronously enter anagen. Remarkably, Bortezomib-treated animals re-entered anagen prematurely, with an average return to anagen after 3–5 days. PaPI treatment results in longer hair due to more rapid anagen growth and inhibition of the quiescent telogen phase of the hair cycle.

Figure 2.

PaPIs induce murine hair growth. PaPI-treated murine skin induces a shortened telogen. (A, B): H+E sections and gross images of vehicle and PaPI-treated skin showing marked hair growth. Hair cycle staging was determined in part by changes in skin pigmentation. Treatment started during the first anagen in postnatal day 28 female B6 mice. (C): Quantitation of the number of mice in anagen and telogen of PaPI treatment using two different topical PaPI treatments. Note marked inhibition of telogen and mild extension of anagen in PaPI-treated skin. (D, E): Normal differentiation in PaPI-treated skin. Differentiation markers keratin 10 (D, green) and keratin 14 (E, green) reveal no abnormalities at Day 47. (F): Cleaved caspase 3 staining shows no difference in PaPI-treated and vehicle-treated skin samples at Day 52. (G): PaPI-treated skin reveals a lot more proliferation compared to vehicle-treated skin at Day 52. Abbreviation: PaPI, partial proteasomal inhibitor.

The pathways regulating hair cycling can potentially disrupt normal hair differentiation if ectopically active. To determine whether there were any abnormalities associated with PaPI treatment, we analyzed histology of skin in similar stages of the hair cycle with various differentiation markers. Skin differentiation and overall architecture were unchanged by proteasome treatments as indicated by proper maintenance of K14 basal and K10 suprabasal compartments (Fig. 2D, 2E). Similarly, hair follicle markers K17 and AE13 retained their proper expression pattern (data not shown). Furthermore, PaPI treatment did not cause apoptosis, as determined by cleaved caspase 3 staining (Fig. 2F and Supporting Information S2A) but induces an increase in proliferation as observed by Ki67 staining (Fig. 2G and Supporting Information S2B). Finally, we asked whether PaPI treatment altered the susceptibility to skin cancers such as basal cell carcinomas (BCC). BCCs derive from inappropriate activity of the transcription factor Gli, previously shown by our group to be regulated by the proteasome in vitro [18]. We reasoned that if PaPI treatment affected Gli activity, we would see increased tumor size or earlier onset. We treated a cohort of mice carrying the Gli1 transcription factor under control of the keratin 5 promoter (K5) with either vehicle or drug every 3 days for 3 months. We found that treated and untreated tumor susceptible mice develop skin tumors at the same rate and with approximately the same size (Supporting Information Fig. S4), suggesting that PaPIs were not sufficient to enhance Gli1-dependent tumor induction. We conclude that PaPIs act to increase hair by controlling hair cycling and proliferation without altering hair differentiation or tumor susceptibility.

The topical effects of the PaPIs coupled with our in vitro data suggest the tight balance between pathways controlling the onset of anagen and those enforcing telogen hair stem cell quiescence underlie the PaPI effects. To determine which pathways PaPIs effect, we treated reporter-lacZ mice, which provide a readout for multiple hair cycle hair cycle signaling pathways Shh, Wnt, and Bmp. Interestingly, the Msx2lacZ reporter mice for BMP signaling [19] maintained Xgal staining during telogen and turned off during anagen, similar to what is observed in untreated controls (data not shown). Furthermore, the distribution of Psmad1/5/8, the transcription factor mediating the effects of BMP, appeared nuclear in PaPI-treated bulge cells but cytoplasmic in untreated bulge cells (Supporting Information Fig. S3). Psmad 1/5/8 is degraded by the proteasome [20] and because of PaPI treatment, it is still nuclear although the skin is already in anagen. Similarly, the localization of NFATc1 remained distinctly nuclear in the bulge stem cells (Supporting Information Fig. S3) of treated skin. These data indicate that the pathways enforcing quiescence are upregulated or unchanged in PaPI-treated skin despite the induction of the anagen growth phase.

By contrast, PaPI-treated skin showed marked increases of Wnt signaling in the skin. PaPI-treated Top-Gal (Wnt reporter) mice demonstrated early accumulation of β-gal activity in the hair germ, suggesting induction of the Wnt pathway (Supporting Information Fig. S9). Also, nuclear β-catenin levels appeared significantly elevated in the sebaceous gland and bulge cells, especially in the hair germ just before the onset of anagen (Fig. 3A, 3B). We followed increases in β-catenin in the hair germ relative to the onset of anagen as defined by proliferation of the hair germ. At the catagen-telogen transition, no difference was noted in PaPI-treated versus nontreated skin. Within a few days, elevated β-catenin levels were noted prior to anagen onset in PaPI-treated skin. Within a day, the hair follicles went into the anagen phase of the hair cycle as the levels of β-catenin increased in the hair germ. PaPI addition markedly elevated β-catenin levels throughout the bulge and hair germ. Activated β-catenin (ABC) immunoreactivity also increased in the hair germs of PaPI-treated skin compared to vehicle-treated skin (Supporting Information Fig. S5). We conclude that PaPI treatment promotes hair follicle nuclear β-catenin accumulation, Wnt signaling, and anagen initiation.

Figure 3.

PaPIs inappropriately accumulate β-catenin in the hair germ/bulge. (A): Immunoperoxidase staining of β-catenin protein (brown). Little difference in levels of β-catenin between vehicle and PaPI-treated skin during catagen. (B): In contrast, during telogen, there are increased levels of β-catenin in PaPI-treated skin as compared to vehicle treatment. This difference is magnified in late telogen with marked accumulation of β-catenin in the hair germ (arrow, and in magnified view) as the hair enters anagen. β-Catenin accumulation occurs in the bulge and hair germ of treated skin. Abbreviation: PaPI, partial proteasomal inhibitor.

Levels of free β-catenin are tightly controlled by the proteasome. Previous studies using expression of a stabilized form of β-catenin demonstrated that the presence of elevated β-catenin protein is sufficient to induce anagen and hair neogenesis [8, 10]. We reasoned that if PaPIs affected β-catenin degradation, then overexpression of wild-type β-catenin would phenocopy stable β-catenin only in the presence of PaPIs. Indeed, injection of wild-type β-catenin or GFP-expressing (not shown) control virus into skin resulted in no difference in physical appearance or alterations in the hair cycle. This demonstrated the strong post-transcriptional inhibition of β-catenin normally operative in telogen skin to prevent anagen onset. With PaPI treatment in addition to adenoviral β-catenin injection (Fig. 4A left and middle), the induction of anagen was dramatically accelerated compared to either PaPI alone (Fig. 2B) or β-catenin alone (Fig. 4A, far right mouse). The specificity of the result is seen in that only the injected portion of the skin that had been pretreated with PaPI demonstrated accelerated anagen induction. As expected, we found that bulge and hair germ cells accumulated β-catenin with PaPI treatment (Fig. 4B). ABC immunoreactivity indicated that ABC accumulated in bulge and hair germ cells (Supporting Information Fig. S6).

Figure 4.

Wild-type β-catenin with PaPIs accelerates anagen. (A): Injection of adenovirus-expressing wild-type β-catenin subcutaneously rapidly accelerates the onset of anagen with the application of PaPIs (left and middle) but not vehicle control (right most). Note the onset of anagen only occurs at the injection site. (B): β-Catenin accumulation occurs in the hair follicle bulge. Despite β-catenin adenovirus, little accumulation occurs in mice that were injected with the adenovirus only. However, with PaPI treatment, marked β-catenin accumulation occurs and the hair enters anagen. Abbreviation: PaPI, partial proteasomal inhibitor.

We next reasoned that if the proteasome degradation of β-catenin controlled anagen initiation, blocking β-catenin priming kinases should phenocopy PaPI treatment. Priming kinases such as GSK3β modify β-catenin into a form recognized by the proteasome and can be inhibited by the GSK3β inhibitor lithium chloride [21]. We treated several cohorts of mice in late telogen and ascertained the effect of GSK3B inhibitor lithium chloride compared to the sodium chloride control (Fig. 5A, 5C). As with PaPI treatment, lithium chloride treatment significantly shortens the length of telogen, inducing a new anagen in an average of 5 days. Consistent with anagen induction, LiCl-treated sections show increased staining for ABC compared to sodium chloride-treated sections (Supporting Information Fig. S7). The 70% reduction in telogen length seen during LiCl treatment mirrors that seen with PaPI treatment and provides further evidence of the role for the proteasome and Wnt signaling in hair cycle control (Fig. 5A).

Figure 5.

PaPIs act downstream of DKK. To identify where in the Wnt pathway PaPIs act, Wnt modulators were tested. (A): The number of animals in anagen were measured after topical treatment with LiCl, an inhibitor of GSK3β and control NaCl. LiCl dramatically shortens telogen. (B, C): Histology of animals showing anagen stage hair follicles in LiCl-treated mice. (D): Animals injected with adenovirus-expressing DKK cease anagen and enter catagen-telogen (histology below). (E): Animals treated with PaPIs and DKK remained in late anagen, indicating PaPIs act downstream of DKK (n = 3). Abbreviations: DKK, Dickkopf-related protein 1; PaPI, partial proteasomal inhibitor.

To further localize the action of the UPS in Wnt signaling, we determined whether PaPIs could overcome the action of the Wnt inhibitor DKK. DKK acts at the level of the membrane receptor to prevent axin destruction and promote β-catenin degradation [22]. If PaPIs act to stabilize Wnts or activate pathways upstream of β-catenin, then they should fail to block DKK-mediated hair cycle inhibition. By contrast, if PaPIs act downstream of axin to stabilize β-catenin, then PaPIs would counteract the effect of DKK activity. We injected an adenovirus-expressing DKK into anagen skin and observed the expected induction of the skin to the catagen phase, a well-described result of Wnt signaling blockade (Fig. 5D) [23]. However, anagen skin treated with PaPIs completely inhibited adeno-DKK mediated catagen and continued in the anagen phase (Fig. 5E). Although immunohistochemistry of β-catenin showed no difference in levels (Supporting Information Fig. S2C), throughout the skin that was treated with DKK plus PaPI, the hairs remained in late anagen rather than in catagen/telogen transition. This supports the idea that PaPIs act downstream of DKK to stabilize β-catenin and activate stem cells into anagen.

Discussion

Stem cell activation forms the basis for tissue regeneration and occurs both spontaneously and during injury or stress. The UPS regulates the stability of key substrates and in turn controls the timing of key developmental events such as cell cycle progression and fertilization. We have shown that the UPS plays a key role in the control of hair follicle stem cell reactivation and murine hair follicle growth through the regulated stability of β-catenin.

A PaPI effect on hair cycle has been reported before [24] but was thought to be through BMP2 activation. While PaPIs appear to affect multiple signaling pathways, we find that they preferentially stabilize the Wnt pathway component β-catenin, downstream of DKK. Studies of known signaling pathways in the skin have defined both stem cell activation and quiescence pathways as targets of the UPS, including the role in quiescence for BMP2. However, our study of topically applied PaPIs reveals little detectable alteration of quiescence regulators and argue against derepression of quiescence as the mechanism for PaPI action. Rather, our study identifies β-catenin as the most sensitive of the potential substrates to UPS inhibition. Moreover, because we found that increasing doses of β-catenin in conjunction with PaPIs can accelerate the onset of anagen, we postulate a threshold of β-catenin protein above which anagen is induced. Such a threshold would likely be set by the activity of BMP and NfatC1. Consistent with this model, GSK3β inhibitors dramatically shorten the length of telogen. UPS regulation of the hair cycle is analogous to that of the cell cycle, where the anaphase promoting complex regulates the destruction of G1 substrates to control the cell cycle checkpoints [14]. For the hair cycle, ongoing destruction of β-catenin at the beginning of telogen is critical to maintain the resting phenotype.

We have uncovered a novel mechanism for the PaPIs that have marked effects on normal hair growth cycling. Previous studies conflict about the mechanism of action of PaPIs and have been ascribed to NFkB signaling and other pathways [25, 26]. Although we find that NFkB levels are mildly elevated in PaPI-treated skin (Supporting Information Fig. S8), NFkB regulates epidermal differentiation that we show remains unaffected [27]. This suggests that PaPIs can affect different tissues in distinct ways by preferentially stabilizing different pathways.

Our data in hair follicle regulation clearly show that PaPIs function through the stabilization of β-catenin signaling leading to a shortened telogen phase. This is consistent with the previously published role for Wnt signaling in mediating anagen and possessing the ability to ectopically activate anagen. While our data support a central role for PaPIs/GSK3β in acting on the Wnt pathway, we cannot rule out smaller contributions from other pathways. Previous studies on the role of Wnt signaling in hair cycling have clearly demonstrated a role for the Wnt ligand in hair cycling, suggesting that Wnt agonists may have similar effects as PaPIs. However, because Wnts have sustained activity, they are associated with the induction of cancer. Transient stabilization and Wnt pathway activation by reversibly stabilizing β-catenin with PaPIs portends a preferable and potentially safer alternative as a hair growth therapy.

Acknowledgments

We wish to thank the generosity of Millenium Pharmaceuticals and Neosil (now Leo Pharmaceuticals) for providing proteasome inhibitors, the Oro lab for helpful comments on the manuscript, NIH Grants R015ARO54780 and ARO52785 and F32 (G.Y), NIH Intestinal Stem Cell Consortium, 1U01DK085527 and 1R01DK085720 (CJK) Stanford Medical Scientist Training program and Howard Hughes Institute (JVA) for funding.

Author Contributions

G.Y., J.V.A., and P.C.M.: conception, collection of data, data analysis, and manuscript writing; E.H.: conception, collection of data, and data analysis; B.A. and M.F.L.: collection and assembly of data; J.Y. and C.K.: provision of study materials; A.E.O.: conception, data analysis, manuscript writing, and financial support. G.Y., J.V.A., and P.C.M. share equal first authorship.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Ancillary