When tissues are injured and blood vessels clot, the local environment becomes ischemic, meaning that there is a lack of adequate supply of oxygen and glucose delivered to the surrounding cells. The heat shock protein‐90 (Hsp90) family proteins protect tissues from various environmental insults and participate in the repair of damaged tissue. Here, we report discovery of a new ischemia‐responsive mechanism in which the two Hsp90 isoforms Hsp90α and Hsp90β (also known as HSP90AA1 and HSP90AB1, respectively) work together to promote cell motility in wounded skin and accelerate wound closure. We demonstrate that Hsp90α and Hsp90β have distinct and non‐exchangeable functions during wound healing. Under hypoxia and when there is a lack of serum factors, Hsp90β binds to the cytoplasmic tail of the LDL receptor‐related protein‐1 (LRP‐1) and stabilizes the receptor at the cell surface. Hsp90α, however, is secreted by the cell into extracellular space where it binds and signals through the LRP‐1 receptor to promote cell motility, leading to wound closure. In addition to skin injury, we suggest that this repair mechanism applies broadly to other non‐cutaneous injured tissues.
INTRODUCTION
The microenvironment of wounded tissues is hypoxic and lacks continued nutrient supply owing to vascular disruption and high oxygen consumption by cells at the wound edge (Pai et al., 1972). Acute hypoxia in injured tissues is also a crucial environmental cue that triggers initiation of the wound healing processes (Tandara and Mustoe, 2004). For instance, it has been shown that hypoxia promotes migration of both human keratinocytes, the cell responsible for wound closure through re‐epithelialization (O'Toole et al., 1997), and human dermal fibroblasts, which deposit new extracellular matrices to the wound and support subsequent wound remodeling (Mogford et al., 2002). By contrast, impaired responses to hypoxia are associated with impaired wound healing, as seen in the environment in chronic diabetic wounds (Botusan et al., 2008). In diabetic foot ulcers, in particular, the stability of the hypoxia‐inducible factor‐1α (HIF‐1α) protein is compromised owing to the hyperglycemic environment, although the mechanism remains unclear (Catrina et al., 2004; Fadini et al., 2006; Gao et al., 2005). These in vitro and in vivo studies suggest that acute hypoxia is a natural shock signal to the injured tissues and, more importantly, a call for an immediate jump‐start of wound healing. However, how the cells in the injured tissue cope with the new environment of hypoxia and then respond to it remained little beyond speculations. In this study, we show that the two heat shock protein‐90 family members, Hsp90α and Hsp90β (also known as HSP90AA1 and HSP90AB1, respectively), work together in a unique fashion to repair skin wounds. Hsp90β works inside the cell to stabilize the low‐density lipoprotein (LDL) receptor‐related protein‐1 (LRP‐1) receptor at the cell surface. Hsp90α, however, is secreted into the extracellular space where it binds to and transmits a pro‐motility signal through the stabilized LRP‐1 receptor. Together, these two Hsp90 isoforms promote the early phase of wound healing – wound closure – under hypoxia and nutrient paucity.
RESULTS AND DISCUSSION
Hsp90α, but not Hsp90β, acts extracellularly to mediate hypoxia‐triggered dermal fibroblast migration
To establish the individual role of Hsp90α and Hsp90β in hypoxia‐driven human dermal fibroblast (HDF) migration, a crucial event during wound healing, we wanted to selectively downregulate Hsp90α or Hsp90β in the cells. In order to prove isoform‐specific downregulation, we first verified the specificity of anti‐Hsp90α and anti‐Hsp90β antibodies. As shown in Fig. 1A, known amounts of human recombinant Hsp90α and Hsp90β proteins were resolved in an SDS gel and visualized by Coomassie Blue staining (panel a). Duplicate membranes were immunoblotted with either anti‐Hsp90α or anti‐Hsp90β antibody. The results clearly show that anti‐Hsp90α (panel b, lanes 1–3) and anti‐Hsp90β (panel c, lanes 4–6) antibodies are specific for each of the Hsp90 isoforms with little cross reactivity. Using these antibodies, we were able to verify specific downregulation of Hsp90α or Hsp90β protein. As shown in Fig. 1B, lentiviral infection of a short hairpin RNA (shRNA) against Hsp90α (sh‐Hsp90α) selectively downregulated Hsp90α (panel a, lane 2), but not Hsp90β (panel b, lane 2). Similarly, infection of an shRNA against Hsp90β (sh‐Hsp90β) only downregulated Hsp90β (panel b, lane 3), but not Hsp90α (panel a, lane 3). In these cells, however, simultaneously knocking down both Hsp90α and Hsp90β compromised the viability of the cells.
We next compared motility of these Hsp90α‐ or Hsp90β‐knockdown cells in response either to stimulation with PDGF‐BB (the major growth factor for HDFs) (a physiological condition) or to acute hypoxia (stress) using the colloidal gold migration assay (Li et al., 2004). In the absence of any stimuli in serum‐free conditions, all cells showed a basal motility (Fig. 1C, panels a, b, c). Both stimulation with PDGF‐BB and hypoxia increased migration of the control sh‐LacZ‐infected HDFs (panel d and g), as expected. Interestingly, neither Hsp90α nor Hsp90β downregulation affected the PDGF‐BB‐stimulated HDF motility (panels e and f versus panel d). Instead, we even detected a modest increase in PDGF‐BB‐stimulated HDF motility in Hsp90α‐downregulated cells (panels e versus panel d). However, either Hsp90α or Hsp90β downregulation alone was sufficient to impair the hypoxia‐induced HDF migration (panels h and i versus panel g). We have previously reported that hypoxia triggers the secretion of Hsp90 proteins from HDFs during wound healing (Li et al., 2007). Thus, we tested whether extracellular supplementation with Hsp90α or Hsp90β protein rescues the motility defect in Hsp90α‐ or Hsp90β‐downregulated HDFs in response to hypoxia. We found that only the addition of recombinant Hsp90α, but not Hsp90β, protein was able to rescue the motility of Hsp90α‐downregulated HDFs (panel j versus panel l). However, neither Hsp90α nor Hsp90β was able to rescue the motility defect of Hsp90β‐downregulated HDFs (panels k and m). Computer‐assisted quantification of the migration data is shown in Fig. 1D. The above results indicate that (1) Hsp90α, but not Hsp90β, acts outside the cells to mediate hypoxia‐induced HDF motility, (2) the extracellular role for Hsp90α cannot be replaced for by Hsp90β, and (3) the extracellular Hsp90α action requires the presence of intracellular Hsp90β.
To verify the finding that secreted Hsp90α mediates hypoxia‐stimulated cell motility, we used an antibody neutralization approach. As shown in Fig. 2A, both PDGF‐BB and hypoxia caused increased migration of serum‐starved HDFs (bars 2 and 3 versus bar 1). The addition of a control IgG did not affect hypoxia‐induced migration (bar 4). However, anti‐Hsp90α antibody blocked hypoxia‐induced cell migration (bars 5 and 6 versus bar 3), whereas anti‐Hsp90β antibody had little effect (bars 7 and 8). Neither anti‐Hsp90α nor anti‐Hsp90β antibody had any detectable effect on the PDGF‐BB‐stimulated HDF migration (bars 9 and 10), indicating that the physiological and stress signals use distinct pathways to promote HDF migration. Moreover, extracellular Hsp90α had a substantial chemotactic effect on HDFs, as also seen with the positive control PDGF‐BB (Fig. 2B, panel b versus panel c), providing additional support for our previous finding that topical application of Hsp90α protein promotes both acute and diabetic skin wound healing (Cheng et al., 2011).
To confirm the different mechanisms of action by Hsp90α and Hsp90β in vivo, we tested topical application of the proteins on wound healing using our newly established model in pigs (O'Brien et al., 2014). As shown in Fig. 2C, with the control carboxymethylcellulose (CMC) vehicle treatment, 1.5×1.5 cm full thickness wounds closed slightly more than 50% on day 7 (panel d versus panel a). Treatment with recombinant Hsp90α protein accelerated wound closure to ∼75% (panel e versus panel b). Wounds treated with recombinant Hsp90β protein also showed a lesser degree of acceleration of wound closure (panel f versus panel c). However, when we carried out histological analysis of wounds from three treatment groups, we noticed a dramatic difference in quality between Hsp90α‐ and Hsp90β‐repaired wounds. As shown in Fig. 2D, the Hsp90α‐treated wounds achieved a similar epidermal thickness to that of unwounded skin (panel c′ versus panel a′). CMC treatment led to significant reduction in the epidermal thickness (panel b′). However, the worst outcome with regard to the thickness of the re‐epithelialized epidermis came from the Hsp90β‐treated wounds, which closed on the same day to the naked eye as other treatments but had extremely thin epidermis (panel d′). We conclude that extracellular Hsp90α is superior to Hsp90β in promoting wound healing quantitatively and, moreover, qualitatively. However, the reason that Hsp90β promoted wound closure remained unclear and the possibilities include its effect on wound contraction, its lower pro‐motility activity on skin cells, as we previously reported (Cheng et al., 2008), or both.
The role for Hsp90β is to bind and stabilize LRP‐1 receptor from inside the cell
What is the role for Hsp90β in promoting HDF motility in response to hypoxia? We focused on the stability of LRP‐1, the cell surface receptor that mediates hypoxia‐induced cell migration (Cheng et al., 2008; Woodley et al., 2009). We first examined which Hsp90 isoform, Hsp90α or Hsp90β, binds to the short (100 amino acids) cytoplasmic tail of LRP‐1 as a chaperone. To focus on the cytoplasmic tail only, we pre‐incubated the cells with receptor‐associated protein (RAP) prior to lysis of the cells. Given that RAP is known to bind to and block the entire extracellular‐ligand‐binding domains of LRP‐1 (Bu and Marzolo, 2000), this ‘pre‐occupying’ approach prevents endogenous Hsp90 proteins from binding to the extracellular part of LRP‐1 after cell lysis. Anti‐LRP‐1 immunoprecipitates of the cell lysates were subjected to western blot analysis with anti‐Hsp90α or anti‐Hsp90β antibodies to identify the LRP‐1‐associated Hsp90 isoform(s). As shown in Fig. 3A, increasing concentrations of RAP protein in the pre‐incubation did not affect the amount of LRP‐1‐associated Hsp90β (panel a, lanes 1–3). The slight increase in Hsp90β binding to LRP‐1 was likely due to less internalization of LRP‐1 following RAP binding from outside the cells (Bu and Marzolo, 2000). Surprisingly, however, we were unable to detect any Hsp90α from anti‐LRP‐1 immunoprecipitates (panel b). Equal amounts of LRP‐1 pulldown from the cells were verified by blotting the immunoprecipitates (10%) with an anti‐LRP‐1 antibody (panel c) (Note that the 150‐kDa, instead of 55‐kDa, IgG seen in the SDS‐PAGE gel is due to the required non‐reducing working conditions for the anti‐LRP‐1 antibody). These data indicate that Hsp90β, but not Hsp90α, binds to the cytoplasmic tail of LRP‐1 receptor. Therefore, Hsp90β acted as an intracellular chaperone to stabilize LRP‐1 at the cell surface and allow secreted Hsp90α to bind LRP‐1 and initiate cell motility signaling. Indeed, as shown in Fig. 3B, we found that Hsp90β downregulation (panel a, lane 3), but not Hsp90α downregulation (lane 2), caused almost complete disappearance of LRP‐1 in the cells. Under the same conditions, the PDGF receptor‐β and EGFR levels either remained unchanged (panel b) or were slightly reduced (panel c). The Hsp90β down regulation caused LRP‐1 to be down regulated, likely at the post‐translational level because LRP‐1 mRNA was still present (</emph>Fig. 3C, panel a, lane 3 versus lane 1).
To confirm the specific chaperone function for Hsp90β, not Hsp90α, in stabilizing the LRP‐1 receptor and to rule out possible off‐target effects of the shRNA‐Hsp90β used, we carried out gene rescue experiments. First, we overexpressed Hsp90α in Hsp90β‐downregulated cells to test whether a simple physical replacement of the downregulated Hsp90β with Hsp90α would restore LRP‐1 receptor. As shown in Fig. 3D, Hsp90β protein downregulation was evident (panel a, lane 2 versus lane 1), and overexpression of Hsp90α (panel b, lane 3) did not change the Hsp90β protein level (panel a, lane 3). A slight increase in the Hsp90α protein level in Hsp90β‐downregulated cells was always detected (panel b, lane 2) in our experiments, and has also been seen by others (Chatterjee et al., 2007). However, the exogenous Hsp90α overexpression was unable to recover the downregulated LRP‐1 (panel c, lane 3 versus lane 2), indicating that the chaperone function of Hsp90β could not be replaced for by Hsp90α inside the cells. In contrast, when we re‐introduced Hsp90β into the Hsp90β‐downregulated cells to restore its amount back to its endogenous level (Fig. 3E, panel a, lane 3 versus lane 1), the physiological level of LRP‐1 was completely recovered (panel c, lane 3 versus lane 2). The Hsp90α protein levels remained unchanged in these cells (panel b, lane 3 versus lane 1). Therefore, Hsp90β is the chaperone that specifically stabilizes the LRP‐1 receptor.
Exogenous expression of Hsp90β or LRP‐1 rescues the defect in HDF motility caused by Hsp90β downregulation in response to hypoxia and extracellular Hsp90α stimulation
We expected that the HDFs with shRNA‐Hsp90β plus overexpression of exogenous Hsp90β would behave like their wild‐type counterparts and regain enhanced motility in response to hypoxia and extracellular Hsp90α stimulation. As shown in Fig. 4A, Hsp90β downregulation inhibited hypoxia‐stimulated (bar 6 versus bar 3) and Hsp90α‐stimulated (bar 5 versus bar 2) cell migration. Exogenous re‐expression of Hsp90β, which restored the downregulated LRP‐1 receptor, corrected the defects (bars 8 and 9). In contrast, overexpression of Hsp90α was unable to do the same (bars 11 and 12). More convincingly, directly re‐introducing a LRP‐1 gene into Hsp90β‐downregulated HDFs also restored the cell motility. As shown in Fig. 4B, a mini‐LRP‐1 receptor, LRP‐1‐II, which mediates extracellular Hsp90α signaling (Tsen et al., 2013), was successfully introduced to the cells (panel a, lane 3 versus lane 2) and restored the ability of the cells to migrate in response to extracellular Hsp90α stimulation (Fig. 4C, bar 9 versus bar 6). Taken together, the above findings have revealed for the first time the working relationship between extracellular Hsp90α and intracellular Hsp90β in response to a major tissue injury signal, hypoxia and nutrient paucity. A schematic representation of this novel repair mechanism is shown in Fig. 4D. We believe that this mechanism applies not only to skin wound healing but broadly to other tissue repairs, as well as tumor progression, where secreted Hsp90α plays a crucial role.
Hsp90β gene knockout is embryonic lethal in mice (Voss et al., 2000). In contrast, mice develop normally without functional Hsp90α (Grad et al., 2010; Imai et al., 2011). These studies support the notion that Hsp90β, but not Hsp90α, represents the long recognized ‘chaperone’ protein that maintains the intracellular signaling networks essential for life. In contrast, Hsp90α plays less of a chaperone role inside the cells than Hsp90β. We argue that Hsp90α is mainly an extracellular tissue repair molecule. Two recent studies show that conditional deletion of Hsp90α in adult mice causes defects in PIWI‐interacting RNA (piRNA) biogenesis and spermatogenesis (Ichiyanagi et al., 2014; Kajiwara et al., 2012). There are two possible explanations for those findings. First, that Hsp90α acts as a chaperone in these cellular events. The other is that those defects might also be due to lack of the extracellular functions of Hsp90α, in particular for spermatogenesis where the local environment is hypoxic and has a lower temperature than rest of the body. A key future experiment is to test whether extracellular supplementation of the cells isolated from the mice with purified Hsp90α protein rescues the defects.
MATERIALS AND METHODS
Primary HDFs were purchased from Clonetics (San Diego, CA) and were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. The third or fourth passages were used in cell migration assays. Recombinant human (rh)PDGF‐BB was purchased from R&D Systems. Antibodies against PDGFR‐β (number 3169) and EGFR (number 4267) were from Cell Signaling Technologies (Dancers, MA). Anti‐LRP1/CD91 antibody (number 37‐7600) was purchased from Life Technologies (Grand Island, NY). Mouse monoclonal antibodies against Hsp90α (number CA1023) and Hsp90β (number SMC 107) were from Calbiochem (Billerica, MA) and Stressmarq Biosciences (Victoria, BC, Canada), respectively. Anti‐GAPDH antibody (number GTX28245) antibody was from Genetex (Irvine, CA). Recombinant RAP protein was as previously described (Cheng et al., 2008). Production of recombinant Hsp90α and Hsp90β proteins and their purification was as previously described (Cheng et al., 2011). The preparation of carboxymethylcellulose gel with Hsp90, the wound healing assay, and H&E and immunohistochemical staining and statistical calculations are as previously described (O'Brien et al., 2014).
Cell migration assay
The colloidal gold migration assay followed the procedure as described previously by us (Li et al., 2004). Data from independent experiments (n>3) were averaged and the migration index was calculated as a percentage.
Chemotaxis assay
The transwell motility assay was carried out according to the manufacturer's instruction (catalog no. 3422, Corning Life Sciences, Tewksbury, MA).
Wound healing in pigs
See O'Brien et al., 2014. All animal experiments were performed according to approved guidelines.
The lentiviral systems
Utilization of lentiviral systems for downregulation and overexpression of genes are as previously described (Li et al., 2007; Cheng et al., 2008; Sahu et al., 2012). The shRNA sequences for human Hsp90α was 5′‐GGAAAGAGCTGCATATTAA‐3′ (sense) and for human Hsp90β was 5′‐GCATCTATCGCATGATCAA‐3′ (sense).
Reverse‐transcriptase PCR
The primers for reverse‐transcriptase PCR (RT‐PCR) of LRP‐1 are: 5′‐CTCCCACCGCTATGTGATCC‐3′ (forward) and 5′‐ACTCATCTTGTGCTCGGCAA‐3′ (reverse) and for GAPDH are 5′‐CCATCACCATCTTCCAGGAG‐3′ (forward) and 5′‐CCTGCTTCACCACCTTCTTG‐3′ (reverse) (Primer‐Blast software). The products were visualized on an 1.5% agarose gel using ethidium bromide staining.
Author contributions
P. J. and H. D. participated in the design and carried out most of the molecular biology experiments. M. Z. helped the analyses of antibody specificity. A. B. purified the recombinant proteins and participated in the pig wound healing study with K. O. M. C. and D. T. W. helped on primary cell culture and lentiviral vectors. W. L. designed experiments, supervised the entire project and wrote the paper with help of all above.
Funding
This work is supported by the National Institutes of Health (NIH) [grant numbers GM066193 and GM067100 to W. L., AR46538 to D. T. W., AR33625 to M. C. and D. T. W.]; and a VA Merit Award (to D.T.W.). Deposited in PMC for after 12 months.
References
Competing interests
The authors declare no competing or financial interests.