|Home | About | Journals | Submit | Contact Us | Français|
Pancreatic cancer is the fourth leading cause of cancer-related death in the US (1-3). At the time of diagnosis more than 80% of patients have locally advanced or metastatic disease. The late diagnosis, disease aggression, and scarcity of treatment options combine to give a median survival time of 6 months, and a 5-year survival rate of less than 5% (2, 4). Cancer of the exocrine duct cells (pancreatic ductal adenocarcinoma, PDA) constitutes 90% of pancreatic cancers (5).
Mutations and/or deletions in oncogenes and tumor suppressor genes are common in PDA. As for many cancers, the genes most affected lie in pathways that control the cell cycle; they include Kras, p16/CDKN2A, p53, SMAD4/DPC4, and BRCA2 (6, 7). Our knowledge of the events that initiate and promote tumor growth in pancreatic cancer has increased, but the precise molecular signature remains incomplete. A better understanding of the earliest carcinogenetic events is required to develop tools for earlier diagnosis, and to identify novel candidate molecules for targeted therapies.
The inhibitor of differentiation (Id) proteins are transcriptional regulators with critical roles in normal cell growth and differentiation (8, 9). The primary function of the four Id proteins (Id1-Id4) is to bind to and inhibit the activity of basic helix-loop-helix (bHLH) transcription factors, e.g. E47 (10). The bHLH proteins activate transcription by forming homodimers or heterodimers that bind to regulatory ‘E-box’ DNA sequences in target genes. Id proteins lack the basic DNA-binding domains, and form non-functional Id/bHLH heterodimers (10) thus, sequestering bHLH proteins.
Because Id proteins antagonize the activity of many bHLH transcription factors involved in cell cycle control (8, 11), fluctuations in Id protein expression correlate with proliferation in many cell types. Overexpression of Id proteins may thus contribute to tumorigenesis (12, 13). Consistent with this possibility, abnormal expression levels of Id mRNA and protein have been reported in numerous human cancers, including those of the colon (14), and liver (15). Furthermore, ectopic Id expression in transgenic mice is associated with development of tumors, including lymphomas (16, 17) and intestinal adenomas (18). Importantly, changes in Id expression have been proposed as prognostic factors for some cancers, providing further impetus to understand the role played by Id proteins in tumorigenesis and tumor growth (19).
In the present study, we have investigated the role of Id3 in cell cycle control of normal and pathogenic ductal cells. We identify high Id3 expression as an early event in pancreatitis and pancreatic intraepithelial neoplasia (PanIN) formation and a prominent characteristic of human PDA. In primary cells we demonstrate that Id3 is sufficient to induce cell cycle entry in quiescent human pancreatic ductal cells. We further demonstrate a functional requirement for Id3 in proliferation of PDA cells. Growth arrest can be induced in PDA cells by altering the balance of bHLH/Id proteins either by knockdown of Id3 or by overexpression of E47. The results support the hypothesis that Id3 expression plays a key role in pancreatic tumorigenesis and suggest that restoring the balance of the Id3/E47 axis is a promising therapeutic approach to PDA.
To determine whether the Id family of proteins plays an early role in the pathological growth of pancreatic exocrine tissue, we examined Id protein expression in a rapid model of pancreatitis. Normal murine pancreatic ducts are quiescent and do not incorporate BrdU (20). Ligation of the main pancreatic duct (PDL) induces pancreatitis distal to the ligature within 7 days (20, 21). The tissue is characterized by acinar cell involution and ductal hyperplasia. We previously reported that, at 7 days, all cells in the hyperplastic ducts exhibit BrdU incorporation (20). Ducts proximal to the ligation site remain normal, thus serving as an internal control.
In normal mouse pancreata or in control tissue proximal to the ligation site, Id3 expression was low and the proliferation marker Ki67 was not detectable. (Fig. 1O, P, and Supp. Fig.1). Remarkably, distal to the ligation site 100% of cells in hyperplastic ducts expressed robust, nuclear Id3 expression and many cells co-expressed Ki67, evidence that they were actively cycling (Fig. 1M, N, Q, and Supp. Fig.1). The data reveal a temporal and spatial link between Id3 and the onset of duct hyperplasia consistent with a role for Id3 in the initiation of pancreatitis.
Examination of the remaining Id family members revealed that Id2 and Id4 expression also increased in hyperplastic ducts following PDL, but the staining pattern suggested a diffuse subcellular distribution, rather than nuclear expression. PDL did not affect the expression of Id1 (Fig. 1A-L).
Because Id3 expression was elevated in the duct ligation model of pancreatitis and because pancreatitis is a known risk factor for PDA (22), we investigated Id3 expression during the progression of ductal hyperplastia to neoplasia. For these experiments the Pdx1-cre;LSL-KrasG12D transgenic mouse model of PDA was employed (23). The mice express a constitutively active form of Kras under the control of the Pdx-1 homeobox gene promoter. Oncogenic Kras activation is commonly found in human PDA (24), and the Pdx-1-cre;LSL-KrasG12D mice faithfully recapitulate the changes in pancreatic morphology and gene expression patterns that are seen in human PDA development (23). By 2 months of age, the mice exhibit premalignant changes in duct morphology termed pancreatic intraepithelial neoplasia (PanIN). As the mice age, PanIN lesions progress in severity and by 12 months of age approximately 10% of mice exhibit frank PDA.
We examined Id3 expression in pancreata of Pdx-1-cre;LSL-KrasG12D mice and control Pdx-1-cre mice, beginning at 8 weeks of age. The earliest detectable PanIN lesions of Pdx-1-cre;LSL-KrasG12D mice exhibited robust Id3 staining which co-localized with the ductal marker muc-1. Id3 expression remained pronounced in advanced PanIN lesions and carcinoma (Fig. 2C-F, and Supp. Fig. 2). In contrast, Id3 was only weakly detectable in the exocrine pancreas of control Pdx-1-cre mice (Fig. 2A, B and Supp. Fig. 2). The data reveal that Id3 is expressed throughout PDA disease progression, suggesting that Id3 plays a role in pancreatic carcinogenesis.
Upregulation of mRNA and protein from Id family members has been detected in various human cancers (25-27). However, Id3 expression in human PDA has been controversial, and the precise contribution made by Id3 to cell proliferation and differentiation in human PDA is unknown (28-29). To address this question, and to extend the relevance of our observations to human disease, we analyzed Id3 expression in human PDA tissues. Id3 protein was weakly expressed or undetectable in normal, non-neoplastic areas of exocrine tissue (Fig. 2G, H, and Supp. Fig. 3). In contrast, prominent Id3 staining was observed in the majority of cytokeratin 19 (CK19) positive ductal cells within neoplastic tissue areas (Fig. 2I-L and Supp. Fig. 3), confirming that Id3 expression in the human exocrine pancreas is associated with PDA.
To determine whether Id3 is functionally involved in cell proliferation in PDA we examined the effects of Id3 depletion in the human pancreatic cancer cell line Panc-1 (30). Transfection with a short interfering RNA (siRNA) sequence specific for Id3 inhibited Id3 mRNA by 57% (Supp. Fig. 4). Id3 knockdown resulted in a marked reduction in BrdU incorporation in Panc-1 cells, demonstrating that Id3 activity is required for DNA synthesis of human PDA cells (Fig. 3A-E). Interestingly, Panc-1 cell growth is also reportedly inhibited by Id2 antisense treatment (25).
Id proteins inhibit cellular differentiation and stimulate proliferation by blocking the activity of DNA-binding bHLH regulatory proteins (8, 13), We and others have shown that bHLH transcription factors are essential mediators of cell fate in pancreatic tissue (31-32). Moreover, we recently demonstrated that the bHLH protein E47 causes growth arrest in a cell line developed from human fetal pancreatic tissue (33).
Our working model predicted that Id3 promotes PDA growth by inhibiting bHLH protein activity. Thus, we reasoned that enforced ectopic expression of the bHLH protein E47 should shift the Id3/E47 balance, and suppress proliferation of human PDA cells. To test this hypothesis, Panc-1 was engineered to stably express an inducible form of the E47. Here, the E47 protein was fused with a modified estrogen receptor (E47MER), which rendered it sensitive to the estrogen analog tamoxifen (33, 34). In the absence of tamoxifen E47 MER is sequestered in the cytoplasm of Panc-1/E47MER cells and is thus functionally inactive. Addition of tamoxifen induced translocation of E47MER to the nucleus (Supp. Fig 5) allowing it to become transcriptionally active. Consistent with our hypothesis, tamoxifen induced E47 activity resulted in cell cycle exit as illustrated by downregulation of Ki67 expression (Fig. 2F-J). Taken together, the Id3 depletion and E47 overexpression studies establish that the Id3/bHLH axis directly controls the proliferative status of PDA cells. The data suggest that it is the balance of activating (e.g. E47) and repressing (e.g. Id3) Helix-Loop-Helix factors, rather than the absolute level of a particular factor, that regulates proliferation.
Human pancreatic duct cells, a progenitor population for pancreatic ductal adenocarcinoma, are normally quiescent and efforts to induce cell cycle entry in cultured human duct cells have met with little success (35). The finding that Id3 was highly expressed in the earliest hyperplastic ducts in pancreatitis suggested that Id3 could play an initiating role in cell cycle entry of duct cells. To address this question directly, duct cells were isolated from human pancreata and cultured according to our previously reported method (35). The cells were transduced with an adenoviral vector expressing either Id3 (Ad-Id3) (36), or the control protein β-galactosidase (Ad-LacZ). In order to avoid vast overexpression of Id3, conditions were chosen in which many cells remain untransduced (Fig. 4K) or expressed Id3 at levels similar to that observed in pancreatitis and PDA (Fig. 2 and Supp. Fig. 1,2,3). As expected, duct cells expressing LacZ remained quiescent, exhibiting virtually undetectable levels of the proliferation markers Ki67, phospho-histone H3 (pHH3), or phospho-cyclin E (pCyclinE) (Fig. 4 A, D, G). In contrast, Id3 efficiently triggered cell cycle entry as revealed by nuclear Ki67 expression in 25% of Ad-Id3 infected duct cells (Fig. 4 B-C). Following Id3 infection a large proportion of cells also exhibited expression of the G2/M marker, pHH3 (Fig. 4E, F) or the G1/S marker, pCyclinE (Fig. 4H, I, and Supp. Fig. 6). We found that pCyclinE expression was restricted to Id3 positive cells, revealing that Id3 acted in a cell autonomous manner (Supp. Fig. 6). The data establish that the product of a single gene, Id3, is sufficient to induce cell cycle entry in human pancreatic ducts.
Dysregulated expression of Id proteins has been noted in a variety of cancers, and this has prompted interest in Id proteins as both diagnostic markers and as potential therapeutic targets (37). In the present study, we hypothesized that Id3 plays both early and sustained roles in pathogenic pancreatic duct cell growth. Our results are consistent with this hypothesis, and strongly suggest that Id3 activity contributes to the pronounced growth that is a hallmark of pancreatic ductal adenocarcinoma (PDA).
To test the hypothesis that Id3 expression and duct hyperplasia are temporally and spatially linked we examined the pancreatic duct ligation (PDL) model of pancreatitis in which duct hyperplasia reproducibly occurs within 7 days. Our rapid model of pancreatitis allowed us to precisely define the temporal relationship between id3 expression and the onset of ductal hyperplasia. These studies demonstrate that Id3 expression is upregulated in all duct cells at the inception of hyperplasia. The fact that not all cells in hyperplastic ducts were Ki67 positive, although they are BrdU positive (ref Carla) suggests that not all duct cells are cycling at one time.
In contrast to Id3, Id1 was present in pancreatic tissue at low to moderate levels but the expression was not affected by PDL. Id2 and Id4 expression increased after PDL, but the staining patterns were diffuse, in contrast to the nuclear staining of Id3. One study examining Id mRNA and protein expression in pancreatic tissue from chronic pancreatitis (CP) patients (26) noted that Id mRNA levels were similar in CP and normal pancreas but that Id1, Id2, and Id3 proteins were more highly expressed in dysplastic pancreatic ducts than in normal tissue. This latter finding is similar but not identical to our observations in murine pancreatitis.
Activation of the mutant Kras oncogene is one of the earliest transformative events in human PDA, and mutated Kras is present in up to 90% of human PDA cases (24). The Pdx1-cre;LSL-KrasG12D transgenic mouse is an excellent experimental model, as it faithfully recapitulates the progression of PanIN lesion severity observed in the development of human PDA (23). Elevated Id3 expression was observed in PanIN lesions of animals at 8 weeks of age, when most lesions are at the earliest stage (1A/1B) (23). Collectively, the results in the pancreatitis and PanINs place increased Id3 activity at a very early point in the dysregulation of ductal cell growth.
Interestingly, only about 10% of Pdx1-cre;LSL-KrasG12D mice progress to invasive adenocarcinoma, suggesting that additional genetic or epigenetic events are needed to reconstitute the advanced human disease. Consistent with this possibility, mice engineered to express both a pancreas-specific KrasG12D mutation and deletion of a conditional p16INK4A allele display accelerated development of PanIN lesions which progress to invasive adenocarcinoma. In contrast, mice carrying the single deletion of p16INK4A display no pancreatic pathology (38). Transforming mutations in several other oncogenes and tumor suppressor genes are common in human PDA, and accumulate as early, intermediate, and late events. Allelic deletions of p16INK4A occur in human PanIN1A/1B lesions as often as do Kras mutations (39), whereas mutated p53, SMAD/DPC4 or BRCA2 genes appear in more advanced PanIN lesions (40). Thus, while Kras is the initiating genetic defect in Pdx-1cre;KrasG12D mice, the observations with p16INK4A /Kras mice support the idea that additional insults or changes are necessary for the invasive phenotype to develop.
Understanding the relationship between Kras activation and additional targets in development of pancreatic cancer is an area of intense study. It seems unlikely that the Id proteins are bone fide oncogenes, as there are no known tumor-associated Id3 mutations (13, 40). However, the increased Id3 expression found in the earliest PanIN lesions strongly suggests that Id3 is a downstream effector of Kras or cooperates with Kras to initiate/increase ductal proliferation Pdx-1cre;KrasG12D mice. In support of this possibility, a direct pathway between Ras activation and Id3 transcription has been shown in thymocytes (41).
In human PDA tissue we detected intense nuclear expression of Id3 in neoplastic ductal cells, confirming the relevance of the murine PanIN studies to human disease. These results extend and clarify previous studies, which have identified varying degrees of Id1, Id2, and Id3 expression in human PDA tissue (42, 25-29). In one study, slightly increased Id3 in PDA tissues was observed compared with normal pancreas. However, they observed moderate and occasionally strong staining of Id3 expression in normal pancreas, which we did not see (26). Several factors may explain the different observations, including variations in analytical techniques (mRNA, protein, immunostaining methods). Our Id3 immunohistological results are in agreement with a recent study of Id3 and p48 in PDA (43).
Id1 and Id3 undoubtedly have overlapping roles in many tissues, and their expression patterns are similar (44). However, many functional studies have simultaneously targeted Id1 and Id3 expression, making it difficult to obtain a clear understanding of their individual roles. Id1/Id3 knockdown in pancreatic cancer cells inhibits their metastatic potential in an ectopic tumor model (45). Similarly, Id1/Id3 knockdown has an inhibitory effect on gastric cancer (46). In a recent study, Id3 silencing reduced the size and weight of ectopic small cell lung cancer tumors (47). However, aberrant Id3 expression is not always associated with a pro-proliferative effect. For example, ovarian adenocarcinomas display decreased Id3 expression (48), and overexpression of Id3 suppresses the invasiveness of human squamous cell carcinoma (49). Thus, it seems likely that the impact of aberrant Id protein expression on cancer growth is cell-type specific, as is the case in development and differentiation.
In functional studies using siRNA-mediated knockdown of Id3 in Panc-1 cells, we established that Id3 activity is required for DNA synthesis. Similarly, overexpression of the bHLH protein E47 also induces cell cycle exit in Panc-1. Therefore, altering the balance between Id3 and E47 is both necessary and sufficient to regulate the cell cycle in PDA cells. Interestingly, Panc-1 cells carry transforming gene mutations in p53, p16, and DPC4 in addition to Kras. Our finding that suppression of Id3 activity alone can inhibit cell cycle progression in Panc-1 cells therefore implicates the E47/Id3 axis as a critical convergence point for oncogenic signals in PDA.
In a previous study we demonstrated that normally quiescent human pancreatic duct cells were not efficiently stimulated to proliferate in response to high serum or growth factors (35). The observation that Id3 expression was elevated in even the earliest pancreatitis and PanIN lesions suggested that Id3 might be an initiating event in duct pathogenesis. Our hypothesis predicted that Id3 overexpression would be sufficient to induce cell cycle entry in primary human pancreatic ductal cells. Remarkably, we found this to be the case. Ectopic expression of Id3 using adenovirus induced primary human duct cells to enter and progress through the cell cycle, as shown by the proliferation markers Ki67, pCyclinE, and pHH3, although we cannot rule out the possibility that lower Id3 expression would be insufficient to induce cell cycle entry. To our knowledge this is the first demonstration that overexpression of a single Id protein is sufficient to induce cell cycle entry in quiescent primary ductal cells. The fact that our previous attempts to stimulate cell cycle entry in primary human duct cells by exposure to growth factors did not induce proliferation speaks to the potency and specificity of the Id/bHLH axis in their growth control (35).
Together the gain and loss of function studies presented here implicate dysregulation of the Id/bHLH axis as an early effector of duct cell pathogenesis with continued involvement throughout the spectrum of PDA development. Further investigation is necessary to determine whether restoring balance to the Id/bHLH axis will prove an effective therapeutic approach to pancreatitis and PDA.
Male C57BL/6J or ICR mice were purchased from Harlan Sprague Dawley, Inc. (Indianapolis, Indiana). The animals were obtained at 5 weeks of age and housed four per cage on a 12:12 hour light:dark schedule. Pancreatic duct ligation was performed by ligation of the main pancreatic duct as previously described (27, 28). This study was approved by IACUC at the Sanford-Burnham Institute for Medical Research. Pancreata were also obtained from 2-8 month old Pdx-1-cre;LSL-KrasG12D mice and control Pdx-1-cre mice (30) as approved by IACUC at UCSD.
The human pancreatic cancer cell line, Panc-1 (ATCC) was cultured with DMEM/10% FBS/1% Pen/Strep (Invitrogen). Panc-1/E47MER cells were generated by transduction of Panc-1 with a retroviral vector encoding E47MER and CD25. E47MER positive cells were selected by FACS sorting for CD25 (26, 35). To induce E47 activity 4uM tamoxifen (Sigma) or vehicle (ethanol) was added to cultures of Panc-1/E47MER cells for 48hours. For siRNA studies Id3 siRNA or scrambled siRNA (Silencer® Pre-Designed & Validated siRNAs, Applied Biosystems) was transfected with Lipofectamine RNAiMAX (Invitrogen) for 96 hr in Panc-1 cells. Cells were treated with BrdU (1:1000, GE Healthcare) for 1 hour prior to harvest.
Primary adult human pancreatic exocrine tissue was obtained from the NIH Islet Cell Resources-Administrative and Bioinformatics Coordinating Center (ICR-ABCC), JDRF, and University of Alberta Islet Isolation Center, Canada. Primary cells were cultured in 5.5mM glucose RPMI /10% FBS/1% Pen/Strep (Invitrogen) on HTB9 matrix plates or collagen plates (Becton Dickinson) as previously described (35). Human pancreatic cancer samples were obtained under IRB approval from UCSD.
Primary adult human pancreatic exocrine cells were infected with Adeno-Id3 (36) and Adeno-LacZ viruses at MOI= 50 to 100 resulting in approximately 50% infected cells. After 16 hours, cells were washed with media and at 48 hours, cells were fixed for analysis.
For analysis of pancreatic duct ligation (PDL), tissue proximal and distal to the ligature was harvested 7 days following surgery and fixed in 4% paraformaldehyde (USB, OH) for 16-18hr at 4°C followed by freezing in OCT embedding media (Sakura Finetek, CA). Pancreata from Pdx-1-cre;LSL-KrasG12D mice and control Pdx-1-cre mice were harvested and paraffin embedded using standard procedures as previously described (23). All tissue samples were sectioned to a mean thickness of 5μm. Cultured Panc-1 and Panc-1/E47mer cells and primary human exocrine cells were fixed in 4% paraformaldehyde (USB, OH) for 15min at 4°C.
All samples were permeabilized with 0.3% Triton X-100 in PBS for 15min and blocked for 1hr at room temperature. Cells were incubated with the following primary antibodies overnight at 4°C: Id1, 2, 3 and 4 (Santa Cruz, Abcam, US Bio), CK19 and panCK (DAKO), E47 (Santa Cruz), mucin-1 (Muc-1) (Neomarkers), Ki67 (Abcam, DAKO), BrdU (GE Healthcare), phospho-cyclin E, phospho-Histone H3 (Cell Signaling). Id3 antibody specificity was confirmed using competing Id3 peptide. For fluorescent imaging, samples were incubated with Alexa 488 (Molecular Probes, Oregon), Rhodamine, or Alexa 596 (Jackson Immuno Research) fluor-labeled anti-mouse/rabbit/rat/hamster and nuclear counterstained with DAPI (Molecular Probes). Digital images of fluorescently labeled sections stained sections were captured using a fluorescence microscope with a digital camera (Nikon, Tokyo, Japan) or with a confocal microscope (Bio-Rad Laboratories, Inc., CA) equipped with krypton/argon laser. H&E was analyzed by brightfield with a conventional inverted microscope (Olympus, PlanFl 40x/0.60). Image J (NIH) was employed for intensity measurements.
Total RNA was extracted after 4 days with RNeasy Mini kit (Qiagen). 1ug of total RNA prepared for cDNA and 2ul of cDNA used as template for qPCR. Quantitative RT-PCR used the Opticon Real-Time System (BioRad), with SYBR Green (BioPioneer) for human Id3 (Forward: actcagcttagccaggtgga, Reverse: aagctccttttgtcgttgg) and GAPDH (33), used for normalization.
Data are presented as means±SEM. The statistical significance of the differences between groups was analyzed by Student’s t test.
In the pancreatic duct ligation (PDL) model of pancreatitis (n=11), duct cell hyperplasia occurs distal (Ligated region, B, D), but not proximal (Unligated region, A, C) to the ligation after 7 days. A normal ductal structure in the unligated region of pancreas is outlined in dotted white lines (A, C). Representative immunostaining for both Id3 (green in A, B), and Ki67 (red in C, D). In contrast to Fig. 1, Id3 and Ki67 immunostains are shown without mucin for added clarity. Hyperplastic ducts in ligated region (B, D) are easily identified morphologically. Replicating cells, identified by Ki67, were found in large numbers in ducts in ligated (D), but not unligated (C) region of pancreas. Blue nuclear counterstain is DAPI. Scale bars, 50μm.
Immunostaining for Id3 (green) and mucin (red) in pancreata from: A,B, normal Pdx-1-cre mice. Note low level of Id3 expression in normal exocrine tissue, compared with C-H, Pdx-1-cre;LSL-KrasG12D mice (n=5) which exhibit extensive id3 expression in all phases of PanIN and carcinoma development. Blue nuclear counterstain is DAPI. Scale bar, 100μm.
Normal and neoplastic regions of pancreases from patients with PDA (n=7) were analyzed for Id3 (green) expression in CK19 (red) positive duct cells. In normal regions of pancreas (A, B) few ducts exist and these are Id3 negative (green). In contrast, Id3 was upregulated in duct cells of PDA regions (C-H). Blue nuclear counterstain is DAPI. Scale bar, 100μm.
Panc-1 cells were transfected with scrambled siRNA (black) or Id3 specific siRNA (white). Ninety-six hours later, Id3 mRNA levels were measured by quantitative PCR (n=3). * p<0.005.
Immunostaining for E47 expression (green) was performed on Panc-1/E47MER cells without tamoxifen (-TAM) or following treatment with 4uM tamoxifen (+TAM) for 48 hours. Blue nuclear counterstain is DAPI. Scale bar, 50μm
Monolayer cultures of human pancreatic duct cells were infected with an adenoviral vector expressing Id3. Immunofluorescence revealed that only Id3 positive cells (green) express the S-phase marker pCycineE (red). Blue nuclear counterstain is DAPI. Scale bars, 25μm.
We thank Dr. Colleen McNamara for Adeno-Id3 and Adeno-lacZ viruses. We thank Tatsuya Kin at the University of Alberta, Betsy Holbrook at Emory, ICR Centers, particularly City of Hope, for their invaluable gift of human pancreatic cells. We thank Dr. Rati Fotedar for helpful discussion. We thank Li Huang and Kaitlyn Kirk for technical assistance. This study was funded by an Academic Senate Award from UCSD (PRIA), DERC Award from UCSD/UCLA (PRIA), CIRM (SHL) and the JDRF Regeneration Program (PRIA and FL).
Author ContributionsSHL performed experiments and analysis. EH performed animal surgeries. AK performed experiments and analysis. JS provided valuable reagents. AL and DS provided valuable reagents, data analysis, and discussion. SHL, FL, and PI-A wrote the manuscript. PI-A and FL designed experiments and PI-A provided overall direction for project planning and execution.