Can Microsatellite Status of Colorectal Cancer Be Reliably Assessed after Neoadjuvant Therapy?

A new wave of therapeutic intervention has had an impact on cancer therapy. The use of novel immune regulators has revolutionized cancer medicine and provided therapeutic options that previously did not exist. It is postulated that due to potentially heightened immune infiltration, a higher neoantigenic burden, or other unknown factors, one subset of cancer that appears to obtain benefit from immunotherapy is microsatellite instable colorectal cancers (1, 2). Germline or sporadic genomic alterations in the mismatch repair (MMR) genes MLH1, MSH2/TACSTD1, MSH6, and PMS2 allow for replication errors or instability in repeated DNA sequences, thus leading to a condition termed microsatellite instability (MSI; ref. 3). Germline mutations in these genes cause a hereditary cancer syndrome named Lynch syndrome that predisposes them to development of colorectal cancer and multiple other cancers, including endometrial, ovarian, and urinary tract cancers. Sporadic deficiency occurs secondary to silencing of MLH1 via promoter hypermethylation (4). Testing of MSI is usually accomplished by either PCR to detect instability in mono- or dinucleotide microsatellite repeats or IHC to directly assess the expression of the MMR proteins. Although PCR-based techniques may be slightly more sensitive than IHC, in the clinical setting, the 2 techniques have been determined to be equally reliable methods to detect the presence of MSI, showing high concordance (5, 6). Nevertheless, IHC may be performed more commonly because of its feasibility and lower cost (6).

MSI status has been determined to be a prognostic marker in stage II colorectal cancer (3, 7, 8), and therefore, it has been argued that universal MSI testing should be performed in all surgically resected colorectal cancers (9). Traditionally, MSI testing has been completed on the initial diagnostic patient biopsy sample and may even be used to guide initial treatment decisions and referral to genetic counseling to evaluate for Lynch syndrome. Given this strategy, the majority of tumors tested are treatment naïve. However, endoscopic biopsies may yield insufficient tissue to permit both MSI-PCR and IHC, in which case, additional testing must be performed on the resection specimen. In addition, biopsies may not contain matched normal mucosal specimens for comparison. This scenario is particularly relevant to rectal cancers receiving neoadjuvant chemoradiation, for which the effect of chemotherapy and radiation on MSI status is yet unknown. We sought to determine the effect of neoadjuvant therapy on MSI analysis by PCR and IHC using in vitro, in vivo, and clinical data. Available matched paired tumor samples (pre- and posttreatment) were reviewed to assess for the effect of neoadjuvant therapy on MSI status. Concurrently, in vitro and in vivo studies were conducted using isogenic cell lines conditioned for MMR functioning and with patient-derived xenografts (PDX) to determine the effect of neoadjuvant therapy on MSI tumor status (10). In addition, we tested whether tissue engraftment in PDX models induces changes in MSI status secondary to an increase in genomic instability. In this modern era of treatment with both significant prognostic and predictive information of MSI status, it is imperative to properly define this subset of patients with colorectal cancer.

Between September 2009 and August 2011, 608 patients who underwent surgical resection for colorectal cancer at The University of Texas MD Anderson Cancer Center (MDACC, Houston, TX) were queried. Of these cases, 239 were surgically resected after having received prior neoadjuvant chemoradiation (n = 192) or chemotherapy (n = 47). All tumors were tested by both IHC and PCR-based MSI testing in either the pre- and/or posttreatment specimens. All specimens that had matched paired tumor samples pre- and posttreatment were re-reviewed by an expert gastrointestinal pathologist (M.W. Taggart)

Demographic data, tumor characteristics, treatment types, treatment responses, and survival rates were collected on all 239 patients identified from the electronic medical record. Response evaluation was based on the treating physician's assessment. This study was approved by the MDACC Institutional Review Board (IRB).

A patient blood sample was collected, and DNA was extracted and tested using Sanger sequencing technology and multiplex ligation-dependent probe amplification to evaluate for pathogenic mutations in all exons of MSH6 and for the presence of large deletions and duplications in MSH6 and EPCAM (11).

MSI-PCR and IHC testing was performed on pre- and/or posttreatment resection specimens from the 239 patient tumors. Exclusion and inclusion criteria were applied to exclude nonadenocarcinoma histology, known cases of familial adenomatous polyposis, cases where no neoadjuvant chemoradiation or chemotherapy was given, and cases where there was no IHC or MSI-PCR on both the original biopsy and surgical specimens. Cases were included for analysis when there was IHC performed on paired biopsy and surgery specimen (regardless of MSI testing) or when MSI was done on paired biopsy and surgery specimen (regardless of IHC testing). Sections of paraffin-embedded formalin-fixed tissue from blocks containing the most viable tumor were utilized in the analysis. Representative 5-μm section(s) from block(s) containing tumor (and normal tissue for MSI analysis) were stained with hematoxylin and eosin.

For MSI analysis, DNA was obtained by manual microdissection. If viable (nonnecrotic) tumor represented less than 30% of the designated tissue, microdissection was performed per standard of care. MSI-PCR analysis was performed using an expanded NCI panel of 7 markers (BAT25, BAT26, BAT40, D2S123, D5S346, D17S250, and TGFBR2; refs. 12, 13). MSI-high (MSI-H) was defined as the presence of 2 or more (or >30%) loci showing instability, MSI-low (MSI-L) as the presence of 1 (or <30%), and MSI-stable (MSS) as no loci (14).

IHC was performed to detect the level of expression in Ki67 (MIB-1, 1:100; Dako) and DNA MMR proteins MLH1 (G168-15, 1:25; BD Biosciences Pharmingen), MSH2 (FE11, 1:100; Calbiochem), MSH6 (44, 1:300; BD Biosciences Pharmingen), and PMS2 (Alb-4, 1:125; BD Biosciences Pharmingen). In patients with MLH1 loss, a methylation-specific PCR of the MLH1 promoter was conducted. All of these analyses were performed as per standard of care in Clinical Laboratory Improvement Amendments–approved laboratories (15).

HCT116 cells were purchased from ATCC and HCT116+Ch3 colorectal cancer cells were generously provided by Dr. Alan Clark (NIEHS, Research Triangle Park, NC). HCT116 cells harbor a hemizygous mutation in MLH1 (c.755C>A, p.S255*) and therefore are MMR deficient (16). MMR functioning has been restored in HCT116+Ch3 cells via transfer of chromosome 3 and hence transfer of a functional MLH1 gene (Supplementary Fig. S1A and S1B; ref. 10). Cultured cells were maintained in DMEM (Gibco) supplemented with 10% FBS (Gibco). Cell lines were authenticated using short tandem repeat fingerprinting service provided by the Characterized Cell Line Core Facility of MDACC. In addition, mycoplasma contamination was ruled out using a PCR-based method.

Treatment consisted of 3 cycles of single-agent chemotherapy (5-fluouracil, oxaliplatin, or irinotecan), radiation, or 5-fluouracil in combination with radiation (Supplementary Fig. S2). Drug dosage used was based on tested IC₅₀ concentration (50% inhibitory concentration) in the cell line model (F6627, 5-fluorouracil, 25 μmol/L; O9512, oxaliplatin, 3.5 μmol/L; I1406, irinotecan, 25 μmol/L, Sigma Aldrich). Cells were treated with radiation at 2 Gy for 60 seconds on 2 consecutive days using the RS-2000 Biological System. After each round of treatment, pools of cells were assessed for MSI-PCR and IHC of MSH6.

MSI-PCR analysis in the context of in vitro studies was performed using a panel of 10 microsatellite markers. Microsatellite loci were amplified by 3 multiplex PCRs as following: (i) D10S197, BAT26, β-catenin; (ii) D18S58, BAT40, D2S123; (iii) D17S250, BAT25, TGFBR2, D5S346. All 3 multiplex PCRs were performed under the same conditions (available upon request). The PCR fragments were detected by capillary electrophoresis on ABI370 at MDACC Sequencing Core and were analyzed using the software Peak Scanner v1.0 (Applied BioSystems). The patterns of the microsatellite markers before and after treatment were compared to identify changes as described previously (17).

Once cell lines were treated, MSH6 staining (primary antibody, L990, Cell Signaling Technology) of the cells was performed using VECTASTAIN ABC Elite Kit from Vector Laboratories using standard procedures. MSH6 IHC was performed in vitro to assess for variable loss in posttreatment MSH6 seen in clinical data. IHC specimens were scored on the basis of robustness of staining, with 0 meaning no stain, 1+ weak staining, 2+ moderate staining, and 3+ strong staining by an expert gastrointestinal pathologist (M.W. Taggart)

To evaluate the induction of genomic instability by treatment, 3 additional coding microsatellites in ATR, BLM, and CHK1 were assessed by fragment analysis. These microsatellites are known to be stable in MMR-proficient and deficient cell systems. Analysis was performed using the same methodology used for MSI analysis. Primer sequences and PCR conditions are available upon request.

Primary human tumor xenograft in vivo models were established as described previously (18). Tumors were obtained from specimens of 7 patients with metastatic colorectal cancer at MDACC and collaborating institutions (Supplementary Table S4). All patients provided written informed consent for their tumors to be used for research purposes, including the creation of xenografts. Specimens were obtained with approval of the IRB. A total of 7 PDXs were propagated in NU/J 6-week-old female mice (The Jackson Laboratory). Animal experiments using PDXs were performed according to the protocol approved by the IACUC at MDACC.

After tumors were established, when median tumor volume exceeded 300 mm³, treatment was initiated in 2 PDXs. One mouse received FOLFOX (5-fluouracil 100 mg/kg and oxaliplatin 10 mg/kg via intraperitoneal administration) on day 1 for 1 dose. One mouse received FOLFIRI (5-fluouracil 25 mg weekly and irinotecan 15 mg/kg via intraperitoneal injection) for 4 constitutive weeks. Tumors were then harvested and DNA extracted. MSI-PCR analysis in the context of in vivo studies was performed using the panel of 10 microsatellite markers as described above. The patterns of the microsatellite markers before and after treatment and also before (original tumor) and after implantation (established PDX) were compared to find changes as described previously (17).

All the statistical analyses were performed using Microsoft Excel software. The significance of difference was obtained by performing the χ² test, and the level was set at P < 0.05. For comparison of 2 grading methods, Cohen kappa statistical test was performed. Simple t test and Fisher exact test were used to compare paired and unpaired patient cohorts.

We identified 239 patients with resectable colorectal cancer having received prior neoadjuvant chemotherapy or chemoradiation. Patients were included in the study who had matched paired pre- and posttreatment specimens subjected to IHC only (34 patients), MSI-PCR analysis only (2 patients), or both (1 patient). Of these 37 patients, median age at diagnosis was 55 years (range, 31–85), 57% were male, 89% received chemoradiation prior to resection, with the remaining receiving chemotherapy alone. Patient characteristics are displayed in Supplementary Table S1 with comparison between paired and unpaired patient cohorts. Both cohorts were similar with more patients with poorly differentiated tumors in the unpaired cohort.

Of the 239 patients treated with neoadjuvant chemoradiation or chemotherapy alone, 35 tumors had both pretreatment and posttreatment IHC performed. Thirty of the 35 tumors (86%) showed intact protein for all 4 MMR proteins in both pre- and posttreatment specimens. One patient, a 35-year-old woman with a strong family history, had equivocal MLH1 staining on pretreatment IHC and definitive loss of MLH1 and PMS2 on posttreatment IHC. Posttreatment MSI showed allelic shift in 7 of 7 markers, thus being consistent with maintenance of MSI-H status despite treatment. The remaining 4 patients all showed completely intact MMR proteins in the pretreatment samples but isolated changes in MSH6 expression in the posttreatment samples that were interpreted as loss of staining. One sample showed complete MSH6 loss in posttreatment resection specimens, and 3 showed patchy staining of MSH6. No change in status was seen in the posttreatment specimens of the other 3 MMR proteins tested (Fig. 1). Of note, based on MSI-PCR analysis, also performed on either pre- or posttreatment samples, all tumors were MSS (Supplementary Table S2). Therefore, based on the observed changes in this retrospective review of the clinicopathologic data, we decided to perform further in vitro MSH6 analysis using IHC as described below.

Of the 239 patients treated with neoadjuvant chemoradiation or chemotherapy alone, 3 tumors had both pretreatment and posttreatment MSI-PCR analysis performed. All 3 tumors from pretreatment samples displayed MSS phenotype, with 0/7 markers showing MSI. Two of 3 posttreatment resected specimens again showed 0/7 markers, and 1 tumor displayed instability in 1/7 markers consistent with an MSI-L phenotype (Supplementary Table S3). On the basis of PCR, there was no evidence of change in MSI status.

Finally, we combined the pre- and posttreatment results of the IHC and MSI comparisons and made a formal statistical analysis. Our observed Cohen kappa was 0.3, thus indicating a fair consistency between pre- and postratings of MSI status. Then, we assessed statistical significance by resampling MSI and MSS calls, pre- and post-, given their marginal probabilities from the observed data. More than 97% of the time, the resulting kappa statistic was less than our observed value of 0.3, which rendered a P value of 0.03.

Of the 4 patients with intact pretreatment MSH6 expression, but variable posttreatment MSH6 IHC staining, 2 had received 8 cycles of neoadjuvant FOLFOX and bevacizumab chemotherapy and 2 patients received neoadjuvant chemoradiation with either 5-FU or capecitabine. Two samples were evaluable for Ki67 expression. One sample, pretreated with FOLFOX plus bevacizumab, showed 5% Ki67 positivity, and 1 sample pretreated with neoadjuvant 5-fluorouracil chemoradiation showed no Ki67 positivity in tumor cells, which is consistent with a decreased expression of MSH6 secondary to a reduction in cellular division rate (Fig. 1E).

All 3 tumor samples tested for MSI showed posttreatment tumor viability between 10% and 20%. Of the 4 tumors tested via IHC for MMR protein loss, with variable MSH6 staining on postresection specimens, posttreatment tumor viability ranged from 0% to 20% (Supplementary Table S2).

One of 4 patients who showed loss of MSH6 expression by IHC after treatment had undergone germline sequencing of DNA extracted from a peripheral blood sample. The germline blood sample was negative for mutations or large deletion/duplications of MSH6, thus ruling out Lynch syndrome. The rest of the patients did not return or declined further analysis.

MMR-deficient (HCT116) and proficient (HCT116+Ch3) isogenic cell lines were chosen due to their clinical relevance and ability to recapitulate MMR-deficient tumor biology (19). Cell lines were exposed to treatment conditions described above. HCT116 cells displayed MSI in all 10 markers tested. Conversely, all HCT116+Ch3 showed stability in all 10 markers tested. This remained consistent in both cell line models through all 3 cycles of treatment, and no change in MSI status was detected in either cell line (Table 1).

Fragment analysis in alternative microsatellite markers for the assessment of the maintenance of genomic instability was performed on all cells lines tested in vitro. There was no increase in genomic instability posttreatment in cell lines in all 3 microsatellites tested (Table 1).

HCT116 and HCT116+Ch3 cells were exposed to 3 consecutive cycles of monotherapy chemotherapy, radiation, or combined radiation with 5-flurouracil. The MMR-deficient cell line HCT116 showed 2+ moderate staining of MSH6 in all conditions after the first cycle of treatment. Following the second and third cycle of treatment with radiation, MSH6 staining remained 1+ but was never completely lost. MSH6 staining remained 2+ throughout all cycles of treatment with combined chemotherapy and radiation. After treatment with the second and third cycle of chemotherapy monotherapy, MSH6 staining was variable, however was never completely lost (Fig. 2A). Similarly, the MMR-proficient HCT116+Ch3 cells showed 2+ moderate staining of MSH6 in all conditions after the first cycle of treatment except chemoradiation with MSH6 stained 1+. Following the second cycle of treatment in all conditions, staining increased in intensity to 3+, although staining remained 2+ in the irinotecan-treated cells. By cycle 3, MSH6 staining again decreased in all conditions but was never completely lost (Fig. 2B). Overall, MSH6 expression was never lost neither in MMR-deficient nor in proficient cell lines.

We decided to test whether genomic instability generated from tumor implantation and engrafting could influence MSI status in PDXs. We completed our in vitro assessment using 5 PDX mouse models derived from patients who had not been exposed to prior neoadjuvant chemotherapy or chemoradiation. These 5 PDXs were derived from patients with a median age at diagnosis of 55 years (range, 38–74); 60% were male, and 40% were MSS. Of the 3 MSI-H PDXs, 1 was derived from a patient with MLH1 methylation, 1 with MLH1 loss, and 1 from MSH6 loss. Patient characteristics are displayed in Supplementary Table S4. After xenograft creation, MSI status did not change postimplantation in all 5 models (Fig. 3).

Two additional MSS PDX lines were treated with chemotherapy using FOLFOX (Fig. 4) and FOLFIRI. MSI-PCR status was assessed pre- and posttreatment, and MSI status remained stable pre- and posttreatment in both models (Fig. 4).

Immunotherapy has revolutionized cancer care in the past 5 years with promising impact of checkpoint inhibitors in MSI-H colorectal cancer. In addition, in recent years, it has been argued that universal MSI testing should be performed on all patients with colorectal cancer due to its relatively common incidence, prognostic effect, and predictive implications for treatment (9, 20). The majority of this testing occurs in the preresection biopsy sample. Once treated in the neoadjuvant setting with chemoradiation, MSI status is not necessarily retested. Despite the lower incidence of MSI-H rectal tumors (21), there is currently no clear understanding of the effect of chemotherapy and radiation on genomic instability and the impact of treatment on MSI status. It is also unclear whether a resistant MSS subpopulation may be selected over the more chemosensitive MSI-H population within the tumor.

In our study, we looked at in vitro and in vivo modeling systems to assess change in MSI status pre- and posttreatment. In the experimental setting, we focused on MSH6 staining, as this was the only MMR protein seen to change in the clinical dataset. On the basis of these data, it appears that neither chemotherapy, radiation, nor the combination of the 2 led to significant changes in MSI status based on microsatellite testing. In patients, testing showed variable posttreatment expression of MSH6 by IHC; however, MSI status by PCR did not change. In vitro testing of proficient and deficient cell line models also shows variability of expression of MSH6 by IHC after serial rounds of treatment; however, complete loss of expression was not observed in any condition.

On the basis of other studies and consistent with our own results, IHC for MSH6 may render conflicting results. In 2010, Bao and colleagues published a study in which 51 patients who received neoadjuvant therapy and resection underwent posttreatment IHC testing. Any posttreatment loss of IHC markers was confirmed by MSI-PCR as well as MMR IHC protein expression, in the pretreatment tumor samples. All of the 51 posttreatment tumor samples showed preserved MLH1, MSH2, and PMS2; however, 10 posttreatment samples (20%) showed decreased MSH6 staining (22). In our cohort of 34 patients with pre- and posttreatment IHC, we saw a similar loss of 11.7% despite remaining MSI stable by PCR. It has therefore been argued that before moving forward with genetic testing for Lynch syndrome, patients whose tumors show diminished MSH6 staining in treated tumors should prompt IHC testing of pretreatment biopsy samples (22). In at least 1 of 4 patients whose specimens showed patchy loss of MSH6, no germline mutations were detected in MSH6.

One could postulate that the loss in MSH6 protein postneoadjuvant treatment may be secondary to subclonal mutational changes in the MSH6 gene or mutations in intronic regions (e.g., the gene promoter) that may be responsible for altered gene expression. IHC results of other MMR proteins did not show changes in the clinical dataset. As described by Kondo and colleagues, hypoxia and low tissue pH caused by chemoradiation therapy may also select for cells that are MMR deficient and cause focal or patchy loss of MSH6 (23). In addition, in MSS patients, loss or decreased MSH6 expression may be related to decreased cellular division rates and induction of a resting state in response to chemoradiation therapy. We believe that this is the most likely explanation, as we have also observed concurrent posttreatment decreases in Ki67 and MMR protein expression as recently reported by Kuan and colleagues (24).

IHC is frequently performed on small pretreatment biopsy specimens when there is little available tissue. IHC may therefore be more difficult to interpret after treatment. As seen in Supplementary Table S2, patient 1 had patchy loss of MSH6 in the context of a pT0 tumor; this may have been due to MSH6 staining of stroma or lymphocytes, but negative in fibrotic tumor. Improvements in future treatment of colorectal cancer may also mean less viable and more necrotic posttreatment tumor specimens. Factors such as tumor viability and performance of MSH6 antibodies may account for the observed loss of expression in posttreatment specimens.

Previous studies have also documented variable MLH1 IHC staining patterns similar to the one in the patient in our series who had decreased MLH1 staining in the pretreatment specimen and loss of MLH1 staining in the posttreatment specimen. Over one-third of all MLH1 mutations are missense mutations resulting in nonfunctional proteins that are antigenically intact (25). These mutant proteins may therefore result in a mildly positive or weak staining pattern. Such individuals could show different MLH1 IHC staining patterns as in our patient (26).

The main limitation inherent to our study includes small numbers of patients with paired pre- and posttreatment specimen MSI and IHC analysis, despite the fact that many samples in the original cohort were tested in either setting. Although our observed correlation between the status of the samples pre- and posttreatment was indicative of a fair consistency, we decided to assess the statistical significance by resampling MSI and MSS calls, pre- and post-, given their marginal probabilities of our observed results. More than 97% of the time, the resulting kappa statistic was less than our observed value of 0.3, which could be interpreted as a P value of 0.03. Of course, the effect of inconsistencies in MSI determinations are not equal. For example, calling a tumor MSI when in fact it is MSS can involve a change of therapy that would be ineffective in the patient and with high toxicity. The reverse inconsistency will result on withholding a potentially very active therapy. In this statistical analysis, we made no attempt to weigh these contrasts but rather scored them equally. Another limitation is that limited tumor viability (10%–20%) in posttreatment specimens also made interpretation of IHC results difficult, and unfortunately, no other tissue was available for staining. In addition, neoadjuvant treatment was relatively inconsistent. Whether a patient received chemotherapy alone versus chemoradiation was practitioner dependent upon tumor location and stage at diagnosis. In regards to in vitro and in vivo experiments, we are limited by experiments on 2 cell lines, 1 with an MSI-H, and 1 with MSS phenotype as well as 2 PDX models used to assess pre- and posttreatment status. Although these are the most commonly used pair of isogenic cell lines. In the future, other colorectal cell lines may be tested to assess for variability.

The effect of neoadjuvant therapy on MSI status in rectal cancer cases has been a subject of debate in the colorectal cancer research community. In this study, we used both in vivo and in vitro modeling systems to assess this question. Our findings indicate that in the posttreatment setting, IHC testing is relatively unreliable and may lead to unnecessary work-up for suspicion of Lynch syndrome. In contrast, it appears on the basis of our study that MSI-PCR status remains unchanged after chemotherapy or chemoradiation. In light of these new data, it appears that MSI-PCR analysis is currently the most reliable test for evaluation of MMR in colorectal tumors and should be performed upon initial evaluation or to confirm a negative IHC result. In the future, as demonstrated by Le and colleagues, with the use of checkpoint inhibitors in MMR-deficient patients with colorectal cancer, mutational load may replace IHC and MSI-PCR analysis as a surrogate to identify treatment candidates and follow treatment response (1). We believe this information is imperative to define the MMR-deficient patient subset in the modern era of cancer immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: J.B. Goldstein, M. Rodriguez-Bigas, Y.N. You, E. Vilar

Development of methodology: J.B. Goldstein, E. Vilar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.B. Goldstein, W. Wu, G. Masand, A. Cuddy, M.E. Mork, S.A. Bannon, P.M. Lynch, M.W. Taggart, J. Wu, S. Kopetz, Y.N. You, E. Vilar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.B. Goldstein, E. Borras, P. Scheet, S. Kopetz, Y.N. You

Writing, review, and/or revision of the manuscript: J.B. Goldstein, E. Borras, A. Cuddy, M.E. Mork, S.A. Bannon, M. Rodriguez-Bigas, M.W. Taggart, S. Kopetz, Y.N. You, E. Vilar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.B. Goldstein, A. Cuddy, M.W. Taggart, E. Vilar

Study supervision: E. Vilar

We thank the Sequencing and Microarray Facility at MD Anderson Cancer Center for their service.

This work was supported in part by a Young Investigator Award of the Conquer Cancer Foundation of the American Society of Clinical Oncology to E. Vilar; the Janice Davis Gordon Memorial Postdoctoral Fellowship in Colorectal Cancer Prevention (Division of Cancer Prevention/The University of Texas MD Anderson Cancer Center) to E. Borras; The University of Texas MD Anderson Cancer Center G.S. Hogan Award in Gastrointestinal Research to Y.N. You; and NCI/NIH Cancer Center support grant P30CA016672, which supports the Sequencing and Microarray Facility Core.

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.