miR-181a Targets RGS16 to Promote Chondrosarcoma Growth, Angiogenesis, and Metastasis

Chondrosarcoma is the only primary bone cancer without an effective systemic treatment. Conventional cytotoxic chemotherapy is not effective, and targeted therapeutics have yet to be developed. Chondrosarcoma is highly metastatic and patients typically succumb to pulmonary metastases. The ability to induce sustained angiogenesis is a necessary condition for the two most important traits of cancer: unrestrained growth and development of metastases (1). Angiogenesis is usually a tightly regulated process that becomes dysregulated during neoplasia (2). Angiogenesis and invasion are intimately connected processes and are amplified by hypoxia. In chondrosarcoma, angiogenesis, invasion, and metastasis are a result of CXCR4 signaling, VEGF and MMP1 expression, HIF-1, and potentially microRNA (miR). We have demonstrated that grade 2 and 3 chondrosarcomas have more microvascularity and VEGF expression than benign or grade 1 tumors (3, 4). Vascularity correlates with clinical behavior, because it is primarily grade 2 and 3 chondrosarcomas that metastasize (5). Accordingly, inhibition of tumor angiogenesis could decrease tumor growth and metastasis (6), and therefore we are interested in pathways regulating these processes.

Hypoxia, which develops as tumors outgrow their blood supply, is an important factor that drives angiogenesis and aggressive behavior in cancer (7). Hypoxia-inducing factor one (HIF-1) is the primary transcription factor that mediates changes in gene expression during hypoxia (8). We and others have shown that HIF-1 expression is higher in grade 2 and 3 chondrosarcomas than in low-grade cartilage tumors (3, 9, 10). We have also shown that, in chondrosarcoma cells, HIF-1 both directly increases VEGF expression and indirectly increases VEGF expression by upregulating CXCR4 expression and signaling (11, 12). CXCR4 signaling also increases MMP1 expression (13, 14).

A posttranscriptional mechanism regulating angiogenesis and metastasis in chondrosarcoma could involve miR. miRs are short, endogenous, noncoding RNAs that negatively regulate gene expression by promoting mRNA degradation or by translational repression through complementarity with sequences in the 3′-untranslated region (UTR). In cancer, miRs can function analogous to tumor suppressors or as oncogenes (oncomiRs) when over- or underexpressed, the net effect dependent on the target genes. miRs may be master regulators of the malignant phenotype (15). They have been implicated in angiogenesis (16) and specific miRs are upregulated by hypoxia (17). We have identified miR-181a as a potential oncomiR that is upregulated by hypoxia in chondrosarcoma that also regulates VEGF expression (18), and hypothesize that its overexpression contributes to angiogenesis and metastasis by amplifying CXCR4 signaling.

Human chondrosarcoma (HCS) from 23 patients and 12 normal articular cartilage specimens were obtained at surgery and either preserved in RNAlater Solution (Applied Biosystems) or snap frozen in liquid nitrogen for later use. The 23 chondrosarcoma included 7 grade 1 and 16 grade 2 and 3 tumors. We previously reported overexpression in one of the grade 2 and one of the grade 3 tumors (18).

Human chondrocytes isolated from normal adult articular cartilage and the chondrosarcoma cell line JJ (a gift from Dr. Joel Block, Rush Medical School, Chicago, IL) were cultured in complete medium (40% DMEM, 40% MEM, and 20% F12), and CS-1 (a gift from Dr. Francis Hornicek, Harvard Medical School, Boston, MA) was cultured in Gibco RPMI-1640 Medium (Life Technologies), all with 10% FBS in a humidified incubator (NuAire Inc.) under 5% CO₂ and either normoxia (ambient oxygen) or hypoxia (2% O₂; ref. 11). JJ was derived from a human grade 2 chondrosarcoma (19), CS-1 was derived from human grade 3 chondrosarcoma and metastasizes in a xenograft mouse model (20). In studies involving CXCR4 signaling, 10 to 20 ng/mL SDF1 (R&D Systems) was added to the media. The JJ cell line was authenticated using short tandem repeat (STR) profiling (ATCC) and was performed on the source cell line in 1999, 2007, and repeated in 2012. There is 94% similarity between the different time points, the cells are human, and there are no matches with any cell lines in the ATCC database. The CS-1 cell line was also authenticated using STR profiling (ATCC) and was performed in September 2012 and matched the STS profiling performed by the source laboratory in 2011, and there were no matches in the ATCC database.

Total RNA, including miR, was extracted from HCS tissue, cartilage, chondrocytes, JJ, and CS-1 cells using the miRNeasy Mini Kit (Qiagen). The concentration and quality of total RNA were determined with a NanoDrop 2000C spectrophotometer (Thermo Fisher Scientific). Samples with purity of 1.8 to 2 and integrity over 1.6 were used for analysis of miRNA expression.

Total RNA was reverse transcribed using the miScript Reverse Transcription kit (Qiagen). Quantification of the ubiquitously expressed miRNA, U17a, was used as an internal control that is expressed consistently in normoxia and hypoxia. A reaction mixture (20 μL) containing the SYBR Green Master Mix (Qiagen), 2 ng of cDNA template plus miScript Universal primer and miScript Primer Assay (miR-specific primer for miR-181a) in a 96-well plate was used for real-time PCR using the miScript SYBR Green PCR Kit (Qiagen). The reactions were done in triplicate on the DNA engine Opticon 2 PCR amplification system (Bio-Rad). PCR conditions: an initial step at 95°C for 10 minutes, followed by 40 cycles of amplification at 94°C for 10 seconds, 55°C for 30 seconds, then 70°C for 30 seconds. To determine the expression level of RGS16, total RNA was analyzed using the Reverse Transcription System (Bio-Rad) followed by real-time PCR with SYBR Green Master Mix (Qiagen). 18S and B2M were used as internal controls (21, 22). The primers for RGS16 and 18s have been previously published (23, 24). The primers for B2M were 5′-GTGGAGCATTCAGACTTGTCTT-3′ and 5′-GCGGCATCTTCAAACCTCC-3′, respectively. The data analysis was performed as previously described (23, 25).

Transient miR-181a knockdown or overexpression was achieved with syn-hsa-miR-181a miScript miRNA mimic, control miR, anti–hsa-miR-181a miScript miRNA inhibitor, and miScript inhibitor negative control (Applied Biosystems). Transfections were performed with GenMute transfection reagent (SignaGen Laboratories). pmiRZIP lentivector expressing anti–miR-181a or control sequence (SBI) was used for permanent miR-181a knockdown experiments. Transduction-ready FIV-based pseudoviral particles were generated using pPACK-H1 Lentivector Packaging System together with the 293TN cell line (SBI), at a titer of 1.06 × 10⁹ IFU/mL. Control was Lenti-scramble Hairpin control pseudoviral particles at a titer of 1 × 10⁹ IFU/mL. Cells were cultured in 12-well plates at a density of 1 × 10⁵ per well for 1 day, infected by pseudoviral particles (using a multiplicity of infection of 100 viruses/cell) and cultured for 72 hours, then selected for puromycin (5 μg/mL) resistance for stable cell lines. Stably transduced cells were used for in vitro and in vivo experiments. Cells were transfected with human regulator of G-protein signaling 16 cDNA (RGS16, NM_002928) cloned into a pCMV vector (pReceiver-M02) and negative control Vector (EX-NEG-M02; GeneCopoeia) and selected for neomycin resistance.

Cell membrane proteins were isolated using hypertonic buffer (Active Motif), centrifugation at 40,000 rpm for 40 minutes and proteins from whole-cell lysates were suspended in 2× loading buffer. Equal amounts of proteins in cell lysates were separated with SDS-PAGE and probed with antibodies for anti-RGS16 (Abcam), or VEGF (VEGFA), MMP-1, and actin (Santa Cruz Biotechnology). The fluorescent signals were detected using a fluorescently labeled secondary goat anti-rabbit antibody (Alexa Fluor 680; Molecular Probes) and analyzed on Licor Odyssey Scanner (LI-COR Biosciences).

Lysates from homogenized xenograft tumors and conditioned media (CM) from cultured cells were used. Cells were cultured for one day, then the medium was changed to 1% FBS O/N, and the CM were collected for ELISA to measure VEGF and pro-MMP1 concentration (R&D system; refs. 12, 13). Each sample was measured in duplicate and each experiment was repeated at least three times. VEGF and MMP1 were normalized to the lysate protein concentration as determined by Quick Start Brandford protein assay (Bio-Rad).

A luciferase reporter construct (0.5 μg) with RGS16 3′UTR wild-type or mutant (GeneCopoeia, Gene Accession: NM_002928.3, UTR length: 1668 bp) was cotransfected with 20 nmol/L miR-181a into JJ cells using GeneMut transfection reagent (SignaGen Laboratories), and cultured for 48 hours. The sequence for the miR-181a–binding site is 1495 ACC AGA CTC TAC CTCTGAATGTG. RGS16 3′UTR-mut was derived by mutating the miR-181a–binding site to 1495 ACC AGA CTC TAC CTCTATCAGTG. Luciferase assays were performed using the Luc-pair miR Luciferase Assay Kit per the manufacture's instruction (GeneCopoeia). Briefly, firefly luciferase activity was measured first, followed by Renilla luciferase activity, and data were recorded in a GLOMAX 20/20 luminometer (Promega). Firefly luciferase activity was normalized to Renilla luciferase activity in the same well (23).

Invasive activity of CS cells was analyzed with Matrigel-coated BD Falcon Cell Culture Inserts (BD Biosciences) as previously described (13). Briefly, 180 μL of BD Matrigel Matrix Growth Factor Reduced (BD Biosciences) diluted 1:3 with serum-free medium was used to coat 12-well inserts and incubated at 37°C for 2 hours. Cells (800 μL; 10⁶/mL) in complete medium containing 1% FBS were added to the upper wells. A 1.5 mL of 5% FBS complete medium containing recombinant SDF-1 (50 ng/mL; R&D Systems) was added to the lower wells. After incubating for 72 hours in hypoxia, cells that invade across the membrane were stained with Cell Stain Solution (Millipore), washed, photographed, then lysed and cell number quantitated by absorbance at 560 nm on a standard microplate reader. The invasion index was calculated by normalizing to the number of cells invading when the lower well had no SDF1 and no FBS.

Cells were cultured for 72 hours and cell proliferation was measured by CellTiter 96AQueous One Solution reagent (Promega). Briefly, cells (5,000 cells/well) were seeded in 96-well culture plates, four replicate wells were used for each condition. At the indicated time points, 20 μL of CellTiter 96AQueous One Solution reagent was added to each well and the plates were incubated for 4 hours. The absorbance at OD490 was measured using SpectraMax 190 Absorbance microplate reader (Molecular Devices).

Xenograft tumors in nude mice were generated as previously described (12). Briefly, 1 × 10⁶ cells in 100 μL culture medium were mixed with 300 μL Matrigel (BD Biosciences) and injected s.c. in the back of nude mice (nu/nu 6–8-week old, female; Charles River Laboratory).

In vivo bioimaging with fluorescence-based tomography (FMT; PerkinElmer) was performed at 3 weeks after injection of tumor cells. Twenty-four hours before imaging, mice were injected via tail vein with 2 nmol/L MMPSense 680 and Angiosense 750 (PerkinElmer). Mice were anesthetized with ketamine (i.p.) during FMT imaging. FMT is acquired with a continuous wave-type scanner capable of acquiring transillumination, reflectance, and absorption data at 680-nm excitation and 700-nm emission or 750-nm excitation and 780-nm emission (PerkinElmer). AngioSense and MMPSense content in xenograft tumors was determined by region of interest analysis as previously described (12).

Tumor size was measured with calipers in three dimensions and volumes were calculated with the formula: height × width × length × 0.52. Weight was determined after 5 weeks of growth at the time of excision. Part of the tumor was fixed in 10% formalin overnight, paraffin embedded, and used for hematoxylin and eosin (H&E) staining and immunostaining with CD34 antibody. Immunostained slides were photographed with a Spot RT camera (Diagnostic Imaging) and Nikon E800 microscope at ×200 to determine microvascularity in the tumor as previously described (4, 12). Some of the tumor was stored in RNAlater or lysis buffer for protein extraction.

Lungs were analyzed with microscopy after fixation in 10% formalin. Transverse sections were made at 100 μm intervals, yielding 12 to 15 sections per lung. H&E-stained slides were scanned with an Aperio Scanscope (Aperio Technologies, Inc.). Each slide was examined for metastases and metastatic burden was quantified as the number of slides with metastases per lung.

All animal studies were approved by the IACUC at Rhode Island Hospital and were carried out in accordance with the NIH Guide for the Care and Use of Laboratory Animals (eighth edition). Human tissue specimens were collected and processed under approved IRB protocols from the Rhode Island Hospital.

All experiments were repeated at least three times. Statistical analysis was performed with GraphPad Prism, v 5.0 (GraphPad Software). Experiments with two groups (qRT-PCR, invasion assay, and tumor weight) were analyzed with the Student t test. Experiments with three or more groups (qRT-PCR, ELISA, and luciferase activity) were compared with one-way ANOVA, followed by the Student t test with Bonferroni correction for individual comparisons. The number of lung sections with metastases were compared with the Mann–Whitney U test. The null hypothesis of no difference was rejected at a significance level of 5%.

Multiple studies have identified overexpressed miRs in cancer that are related to the malignant phenotype. Some of these miRs are also induced by hypoxia, a condition that enhances aggressive tumor behavior. Because chondrosarcoma lacks effective systemic treatments and because strategies to block the effects of hypoxia are also limited, we analyzed miR expression in chondrosarcoma with the goal of identifying miRs that could potentially lead to a targeted therapy. miR-181a was identified as an overexpressed miR in chondrosarcoma by screening of human tumors by miR array and chondrosarcoma cell lines for hypoxia regulated miRs (18). An additional criterion was that the overexpressed miR increased expression of VEGF and MMP1. To validate miR-181a as an oncomiR in chondrosarcoma, a series of primary tumors was analyzed for miR-181a expression. Overexpression of miR-181a was confirmed with qRT-PCR in a series of 23 primary human grade 1, 2, and 3 chondrosarcomas. Expression of miR-181a correlated with tumor grade: 8-fold higher in grades 2/3 compared with grade 1 tumors and 6-fold higher in grade 1 tumors than in cartilage (Fig. 1A). In chondrosarcoma cell lines JJ and CS-1, miR-181a expression was elevated compared with chondrocytes (Fig. 1B). In the tumor cell lines miR-181a expression was more highly expressed in xenograft tumors than in vitro in hypoxic conditions (Fig. 1C). Transfection with an antagomir to miR-181a inhibited VEGF and MMP-1 in a dose-dependent manner (Fig. 2A and B). Time course analysis showed an increasing effect of the antagomir on VEGF and MMP-1 expression (Fig. 2C and D). Western blot analysis on cell lysates from both cell lines from day 2 confirmed the ELISA results (Fig. 2E). We engineered JJ and CS-1 cells to express anti–miR-181a via transduction with a lentivirus construct and this also decreased secreted VEGF and MMP1 into CM. Further overexpression of miR-181a had the opposite effect, increasing VEGF and MMP1 (Fig. 2F–I).

To evaluate whether anti–miR-181a has similar effects in vivo as it does in vitro, and thereby to assess a therapeutic antagomir approach, the metastatic chondrosarcoma cell line CS-1 was transduced with a lentivirus expression construct for anti–miR-181a or control before implantation in nude mice. Knockdown of miR-181a expression in xenograft tumors was > 80% (data not shown). Bioimaging of mice bearing xenograft tumors was performed at 3 weeks with FMT using AngioSense and MMPSense probes, indicators of angiogenesis, and MMP activity. Angiogenesis in xenograft tumors was inhibited by anti–miR-181a (Fig. 3A and B), and this result was confirmed with immunohistochemistry using CD34 antibody (Fig. 3C). VEGF in xenograft tumors was also decreased (Fig. 3D). miR-181a knockdown inhibited cell proliferation in vitro (Fig. 3E) and xenograft tumor volume and weight (Fig. 3F and G).

MMP activity evaluated with FMT was also decreased in xenograft tumors by miR-181a knockdown (Fig. 4A and B), as was MMP-1 content in xenograft tumors (Fig. 4C). Most importantly, knockdown of miR-181a decreased lung metastases (Fig. 4D and E). Anti–miR-181a also decreased in vitro invasion in both cell lines (Fig. 4F).

To identify potential target genes of miR-181a, we performed a bioinformatics search using three different algorithms, TargetScan (http://www.targetscan.org), miRBase (http://www.mirbase.org), and miRanda (http://www.microrna.org). The latter predicted RGS16 as a target of miR-181a. RGS16 is a negative regulator of CXCR4 activity. Because we have previously reported increased CXCR4 signaling in chondrosarcoma, we pursued miR-181a and RGS16 as another mechanism that might explain increased CXCR4 signaling in this tumor (12).

The predicted seeding site for miR-181a is in the RGS16-3′UTR, which is evolutionarily conserved across different species. miR-181a transfection and RGS16 knockdown have similar effectiveness at increasing invasion, and VEGF and MMP-1 expression (Fig. 4G–I). To confirm the prediction that miR-181a targets RGS16, we created a luciferase reporter construct containing luciferase and the WT or mutated RGS16 3′UTR. The results show that miR-181a transfection inhibits WT RGS16 3′UTR, whereas knockdown with anti–miR-181a had the opposite effect (Fig. 5A). There was no effect of miR-181a with the mutated RGS16 3′UTR (Fig. 5B). We then analyzed the effect of miR-181a transfection and knockdown on RGS16 expression in both cell lines. qRT-PCR showed the expected effects if RGS16 was the target: miR-181a overexpression decreased RGS16 mRNA and knockdown of miR-181a increased RGS16 mRNA (Fig. 5C and E). Western blotting was consistent with the PCR results (Fig. 5D and F).

We analyzed RGS16 expression in human and xenograft tumors, cell lines, and during hypoxic conditions, and found that the expression pattern is opposite to that of miR-181a. That is, RGS16 expression was decreased in human tumors compared with cartilage (Fig. 6A); chondrosarcoma cell lines compared with chondrocytes (Fig. 6B), and further decreased by hypoxia and in xenograft tumors (Fig. 6C).

To establish a functional relationship between miR-181a and RGS16, CS-1 cells were transfected with RGS16 cDNA construct, miR-181a mimic, or both. The effect of RGS16 transfection on RGS16 protein levels was confirmed with Western blotting (Fig. 6F), and RGS16 transfection decreased VEGF and MMP1, and cotransfection of RGS16 and miR-181a reversed the increase in VEGF and MMP1 in CM after miR-181a transfection alone (Fig. 6D and E).

In this study, we show that in vivo growth in a xenograft chondrosarcoma model increases expression of miR-181a even more than hypoxic cell culture conditions and that expression of miR-181a correlates with tumor grade. We have previously reported that miR-181a increases expression of VEGF in vitro. Here, we report that miR-181a also increases MMP1 and at least part of the mechanism by which miR-181a overexpression increases expression of VEGF and MMP1 is by inhibiting RGS16, a negative regulator of CXCR4. miR-181a directly targets RGS16 through binding to the 3′UTR, resulting in decreased RGS16 mRNA and protein. We have previously reported that HIF-1 directly increases CXCR4 expression in chondrosarcoma, resulting in an indirect mechanism of increased VEGF and MMP1 expression (12, 13). Hypoxia-regulated transcription of miR is yet another mechanism driving the expression of these factors in chondrosarcoma by inhibiting an inhibitor of CXCR4 signaling. miR-181a knockdown restored RGS16 expression and decreased VEGF and MMP1 expression, angiogenesis, and, most importantly, lung metastases in a xenograft model (Fig. 6G).

In cancer, chemokines, their receptors, and innate regulators of receptor activity are important determinants of invasion, angiogenesis, and metastasis. There are four groups of chemokine receptors (26). Of those, CXCR4 is the one most commonly expressed in tumors, and the one whose expression is most related to development of metastases (27). The ligand for CXCR4 is the chemokine stromal cell–derived factor one (SDF1; refs. 28, 29). CXCR4/SDF1 promotes metastasis by mediating proliferation and migration of tumor cells and enhancing tumor-associated angiogenesis (30–32). CXCR4 is a seven-transmembrane G-protein–coupled receptor, whose activation also triggers intracellular signaling cascades, whose downstream targets include MMP1 and VEGF (32, 33). The increased expression of chemokine receptors has been most widely investigated in carcinomas (34–36). We and others have found that CXCR4 is overexpressed in chondrosarcoma cells and primary chondrosarcoma (13, 37, 38). We have also shown that CXCR4 expression is further increased when chondrosarcoma cells are cultured in hypoxia and increased even more so when grown in mice as xenograft tumors (12). CXCR4 signaling through p-ERK increases expression of VEGF and MMP1 as well as angiogenesis and invasion (13), thus linking the CXCR4 pathway to the angiogenic switch (2, 39) and two of the hallmarks of cancer (1, 40). Extracellular blockade of CXCR4 signaling with the drug AMD3100 partially inhibits secretion of MMP1 and VEGF (12, 13). Intracellular regulation of CXCR signaling involves RGS proteins, a family of proteins that either enhance or inhibit GPC signaling, the latter by accelerating the deactivation of the activated G_alpha subunit through GTPase activity (41). RGS proteins are critical modulators of signal transduction pathways in normal physiology and in cancer. Individual RGS proteins can either be over- or underexpressed in specific cancers. RGS16 is an inhibitor of CXCR4 signaling in normal megakaryocytes and T lymphocytes (24, 42). In breast cancer, loss of RGS16 enhances growth factor–related PI3K signaling, proliferation, and resistance to tyrosine kinase inhibition (43). In pancreatic cancer, underexpression of RGS16 is a predictor for lymph node metastasis (44). Thus, diminished RGS16 leads to cancer progression. Our data are the first to show loss of RGS16 expression in chondrosarcoma and first to show that the mechanism of RGS16 loss involves miR-181a. Another potentially effective method to inhibit CXCR4 signaling, or a method to augment extracellular blockade, would be an intracellular strategy involving anti–miR-181a.

It has been postulated that overexpressed miRs may be master regulators of angiogenesis, invasion, and metastasis through regulation of multiple target genes related to these phenotypes. Our data show that loss of RGS16 expression results from overexpression of miR-181a. MiR-181a is highly conserved across species. Its overexpression is associated with hypoxia (45), progression of myelodysplastic syndromes to acute myelogenous leukemia (46), and has recently been reported in osteosarcoma (47), and breast cancer (48). Our data as well as a xenograft study using human myeloma cell lines treated with miR-181a antagonists resulted in significant suppression of tumor growth (46). Thus, miR-181a is a candidate therapeutic target. We have shown that AMD3100 decreases tumor growth, angiogenesis, and metastasis in a xenograft model (Sun and colleagues; ref. 12). This study shows that miR-181a knockdown has similar effects as AMD3100 (Sun and colleagues; ref. 12). These data are important in demonstrating that CXCR4 inhibition with either AMD3100 or anti–miR-181a are effective in vivo strategies. Anti–miR-181a may inhibit additional pathways than does AMD3100. The advantage of the anti-miR approach is that multiple, partially redundant signaling pathways related to tumor progression may be targeted; one reason anti-miR strategies are under clinical development (49). Whether antagomir therapy would be more or less durable and effective needs to be evaluated and a method of systemic delivery developed.

There are some limitations to our study. We did not attempt to comprehensively analyze additional mechanisms of miR-181a regulation. Hypoxia and HIF-1 increase miR-181a expression (18), but there may be other mechanisms. We also did not attempt to identify all possible targets of miR-181a. In addition to RGS16, there may be other targets of miR-181a and RGS16 may regulate other pathways besides CXCR4.

In conclusion, miR-181a is an oncomir whose overexpression results in gain of function by inhibiting an inhibitor of CXCR4 signaling. Antagomir-based therapy for chondrosarcoma may prove efficacious and has a potential role in therapeutic strategies.

No potential conflicts of interest were disclosed.

Conception and design: X. Sun, R.M. Terek

Development of methodology: X. Sun, R.M. Terek

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Sun, C. Charbonneau

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Sun, C. Charbonneau, L. Wei, Q. Chen, R.M. Terek

Writing, review, and/or revision of the manuscript: L. Wei, R.M. Terek

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Sun, L. Wei, Q. Chen

Study supervision: R.M. Terek

The authors thank Jason T. Machan, PhD, for statistical support; Jack Wands, MD, for scientific mentoring; and Drs. Joel Block and Francis Hornicek for chondrosarcoma cell lines.

One or more of the authors have received funding from the NIH and the National Center for Research Resources, a component of NIH: grant R01AR059142 (to L. Wei), grants 1P20RR024484-01 and 9P20GM104937-06 (to Q. Chen), and grants 1P20RR024484-01 and 1R01CA166089 (to R.M. Terek).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.