Higher Order Chromatin Modulator Cohesin SA1 Is an Early Biomarker for Colon Carcinogenesis: Race-Specific Implications

Colorectal cancer remains the second leading cause of cancer deaths among Americans highlighting the need for more effective strategies (1). The genomic era diagnosis has unprecedented promise of making significant impact on colorectal cancer mortality by employing powerful and increasingly affordable technologies such as next-generation sequencing to identify novel therapeutic targets. A complicating factor is that colorectal cancer is biologically heterogeneous at least 4 consensus molecular subtypes (CMS) and 5 proteomic pattern (2, 3). Moreover, the mutational load spectrum of colorectal cancer is generally quite high with MMR-deficient colorectal cancers (CMS1) having the most mutations of any major cancer (4). Thus, discerning the driver mutations from the numerous passenger mutation background remains a challenging task. Further complicating the biology are the numerous epigenetic events that modify gene expression including methylation, copy number alterations, etc. (5).

Recently, attention has focused on high-order chromatin alterations as the fundamental event in carcinogenesis. Indeed, The Cancer Genome Atlas (TCGA) has revealed that some of the most commonly mutated genes related to higher order chromatin structures are involved in nucleosome positioning and hence in modulation of gene expression (e.g., Arid 1a and other members of the SWI/SNF family; ref. 6). In the human genome, higher order chromatin remodeling can be driven by self-associating topological domains that are enriched in cohesins which play an important role in governing long-range nuclear interactions and gene expression (7). These cohesins are composed of 2 complexes that consist of Smc1, Smc3, Rad21/Scc1 and either of the stromal antigens (SA1 or SA2) which regulate chromatin looping, unmask promoters from the histone shield, and control translation (8). While cohesins have been found to be mutated in a small proportion of colorectal cancers (9), a recent provocative report indicates that most colorectal cancers possess higher frequency of mutations in specific DNA sites which are usually assigned to bind cohesin proteins (10). These studies highlight the important role of alterations in the higher order chromatin in colon carcinogenesis.

Our group has previously developed partial wave spectroscopic microscopy (PWS), a powerful modality to detect changes in the higher order chromatin. By measuring fluctuations in the nanoscale parameters (disorder strength or Ld) with PWS, we have demonstrated that tumors have a large degree of chromatin heterogeneity (11). An even more significant finding from a clinical perspective was that when endoscopically normal rectal mucosa from patients with colonic neoplasia was analyzed, they exhibited markedly altered higher order chromatin (12). These alterations in the higher order chromatin represent a striking manifestation of field carcinogenesis which has previously also been noted through alterations in gene expression, methylation, and proteomics along with ultrastructural changes as detected by transmission electron microscopy (13–15). The potential clinical utility is for improved risk stratification for colon cancer screening through sampling the readily accessible mucosa (i.e., rectal examination). The clinical imperative is that while colonoscopy is recommended for the entire population at age 50 years and older, the yield of advanced adenomas (adenomas ≥1 cm or high-grade dysplasia or ≥25% villous features) is only about 7% to 8%, meaning that a vast majority of colonoscopies do not have any cancer preventive implications (through identification and removal of the biologically significant lesions; ref. 16). Therefore, developing a rectal risk stratification tool would help in identifying at-risk population where colonoscopy is more likely to be high-yield. This in particular is urgently needed in African-Americans who are known to suffer disproportionately from colorectal cancers with regards to both lifetime colorectal cancer incidence risk and mortality than Caucasians (by 23% and 53% in males; 22.5% and 45.5% in females, respectively; refs. 17, 18), with limited progress made in reducing such disparities over the last 2 decades (19). Furthermore, data suggest that African-Americans develop colorectal cancer earlier than Caucasians (20). Therefore, American College of Gastroenterology had recommended earlier screening for African-Americans (21) but other major societies did not endorse that view. Biologically, there are specific racial differences in colorectal cancers with a next-generation sequencing study with mutations in 2 genes (ephrin type A receptor and folliculin) limited only to African-Americans and not Caucasians (22). However, to the best of our knowledge, there are no previous reports on race-specific alteration in colorectal cancer biomarkers, including during early-field carcinogenesis. Thus, finding reliable biomarkers to better risk stratify this supposedly high-risk population (African-Americans) is important.

In this study, we explore the role of the higher order chromatin modulator, cohesin SA1, in colon carcinogenesis. We demonstrate that downregulation of SA1 is an early alteration during neoplastic transformation and is a robust marker of field carcinogenesis. Importantly, the basal expression and diagnostic abilities differed by race providing the first demonstration of this concept. Finally, we show that SA1 is biologically important altering phenotype of higher order chromatin as demonstrated by the novel live-cell PWS and the classic micrococcal nuclease (MNase) assay.

All human studies were conducted with approval from the Institutional Review Board (IRB) guidelines of Boston University School of Medicine (Boston, MA). Rectal biopsies were collected from patients undergoing screening or surveillance colonoscopy at Boston Medical Center. There were 40 males (51%) and 38 females (49%) with mean ages 54.5 ± 4.3 and 52.3 ± 1.3 years, respectively. The samples were distributed equally between African-Americans and Caucasians. In this cohort, 34% had adenomas and the remainder was considered controls (negative colonoscopies or left-sided hyperplastic polyps). Given that this was an asymptomatic population, no carcinomas were noted (estimated prevalence ∼0.7%; ref. 23). Exclusion criteria were: (i) incomplete colonoscopy (poor preparation, inability to intubate cecum, or failure to recover polyps for pathological evaluation); (ii) concurrent anticoagulant therapy; (iii) mucosal abnormalities (i.e., inflammatory bowel disease); and (iv) isolated right-sided proximal serrated polyps. Six biopsies were collected from endoscopically normal rectal mucosa. Biopsies were transported in PBS and 2 biopsies snap-frozen in liquid nitrogen and stored at −80°C for RT-PCR and 4 biopsies fixed in 10% buffered formalin for 24 hours and stored in 70% alcohol for immunohistochemistry (IHC).

To determine the expression and localization of the cohesin SA1, IHC was performed on (i) formalin-fixed rectal biopsies (as above) that were embedded in paraffin blocks, sectioned (4 μm thick), and mounted on Superfrost⁺ glass slides and (ii) human colorectal cancer tissue arrays (US Biomax, Inc.) assembled from 72 different grade tumors, 10 tumor-adjacent histologically normal tissues, and 10 control samples (non–colorectal cancer tissue). The slides were deparaffinized by heating at 60°C (∼1 hour) and 2 washes in xylene followed by rehydrating with graded alcohol washes. The tissue sections were subjected to antigen epitope retrieval by pressure microwaving (Nordic-Ware) at high-power setting (2 × 9 minutes) in antigen unmasking solution (Vector Laboratories). After quenching the endogenous peroxidase activity in 3% hydrogen peroxide, the slides were incubated in 5% horse serum for 2 to 3 hours at room temperature to block nonspecific binding. The sections were then incubated with anti-SA1 antibody (Abcam; rabbit polyclonal; 1:200 dilutions) overnight at 4°C. After standard PBS washings, the sections were incubated with the universal biotinylated secondary antibody (1:2,000) for 30 minutes followed by avidin–biotin peroxidase using Vectastatin Elite ABC Reagent Kit (Vector Laboratories). Finally the antigen-specific brown staining was developed by exposing the sections to 3, 3′-diaminobenzidine (DAB) as the chromogen substrate (for 1–3 minutes). For the negative controls, duplicate sections on the same slide were processed in the absence of the primary antibody. The color intensity was scored on a scale of 0 to 3 by an investigator blinded to colonoscopic findings for rectal biopsies.

In these studies, we compared the mRNA expression of SA1 in random rectal biopsies collected from African-Americans or Caucasian patients with or without adenomas/polyps detected elsewhere in the colon. For these studies, we used frozen colonic tumors obtained from the Boston University Biospecimen Archive Research Core. RNA was extracted from the samples with TRI Reagent (Sigma Chemicals) following the manufacturer's instructions (Molecular Research Center Inc.). The samples were homogenized in a BeadBug homogenizer (Benchmark Scientific) and total RNA isolated using a RiboPure Kit (Ambion, Life Technologies) following the manufacturer's instructions. After establishing the purity and concentration by spectrophotometry (OD 260/280), RNA was reverse transcribed with human SA1-specific TaqMan probes (Applied Biosystems) following the manufacturer's protocol. Samples were processed using Step-One-Plus RT Thermo Cycler (Life Technologies) using RT kit and Universal PCR Master Mix. All the samples were normalized to β-actin, and the relative expression of SA1 was analyzed using the comparative (2^−ΔΔCt) method as described previously (24). The threshold of fold change significance was set as >1.5 (upregulation) and <0.67 (downregulation).

To further explore the functional consequences of SA1, we studied the effects of SA1 downregulation on cellular growth characteristics of human colon cancer cell line HT29 cells (utilized because of relatively high level of basal SA1 expression). For these studies, we created an SA1 knockdown in HT29 cells by stably transfecting them with SA1 specific shRNA vector (Santa Cruz Biotech) with a puromycin selection vector. We also performed transient transfections in cell lines with lower basal SA1 expression (e.g., SW480) using siRNA (Santa Cruz Biotechnology) per manufacturer's instructions.

Western blotting was done using standard techniques. Briefly, 30 μg of sample protein was subjected to SDS-PAGE, transferred to polyvinylidene difluoride membranes (Amersham Pharmacia), blocked with 5% non-fat dry milk, and probed with specific antibody for proliferating cell nuclear antigen (PCNA; 1:1,000; Santa Cruz Biotech), β-catenin (1:500; Santa Cruz Biotech), and p-β-catenin (1:500; Cell Signaling) as described previously. Xerograms were developed with enhanced chemiluminescence (Santa Cruz Biotechnology) and images acquired via UVP Bioimaging Systems and the data processed using Labworks 4.6 software. The data were normalized to the expression of β-actin as loading control.

HT29 cells were seeded in 6-well plates at a density of 0.5 × 10⁶. After 24 hours, media were replenished with 5-azacytidine (5-Aza-dC; 2.5 mmol/L) and sodium butyrate (1.0 mmol/L). After 48 hours of incubation at 37°C/5% CO₂, the protein lysates were subjected to immunoblotting to measure SA1 expression.

Because nucleosome positioning is critical for gene expression and most DNA-related processes, we performed the micrococcal nuclease digestion to examine SA1 nucleosomal positioning using the HT29-SA1shRNA (stable knockdown) and HT29-Scr (scramble vector) cells. The MNase assay was performed with minor modifications as previously reported by our group (15). Cell suspensions (50 × 10⁶) were washed of media with ice-cold PBS (2,000 rpm at 4°C for 10 minutes) and then resuspended in 5 mL of ice-cold NP-40 lysis buffer [10 mmol/L Tris-HCl (pH 7.4), 10 mmol/L NaCl, 3 mmol/L MgCl₂, 0.5% NP-40, 0.15 mmol/L spermine, and 0.5 mmol/L spermidine] and incubated for 5 minutes on ice and the nuclear fraction isolated and washed with 2.5 mL of MNase digestion buffer (10 mmol/L Tris-HCl, 15 mmol/L NaCl, 60 mmol/L KCl, 0.15 mmol/L spermine, and 0.5 mmol/L spermidine) and resuspended in 1 mL of MNase digestion buffer containing 2 mmol/L CaCl₂. From the suspended solutions, 100-μL aliquots were taken and incubated with increasing amounts of the MNase enzyme (0, 80, and 100 units) for 5 and 10 minutes. The reaction was stopped by adding 80 μL of MNase digestion buffer, 20 μL of MNase stop buffer (100 mmol/L EDTA, 10 mmol/L EGTA), 3 μL of Proteinase K (25 mg/mL), and 10 μL of 20% SDS and incubated overnight at 37°C. The samples were then extracted with 200 μL of phenol–chloroform solution and spun and the aqueous layer collected. Following that, 2 μL of RNAse-A (10 mg/mL) was added to the solution and incubated at 37°C for 2 hours and again phenol–chloroform extraction was performed. DNA was precipitated by adding 100% ethanol, and after centrifugation, the pellet was resuspended in DNA hydration buffer or water and 5 μg of the isolated DNA was resolved in a 1.4% agarose gel and visualized by ethidium bromide staining.

A full description of the backscattering interference spectroscopic (BaSIS) microscopy technique has been reported previously (25). Briefly, the BaSIS instrument was built into a commercial inverted microscope (Leica DMIRB) equipped with a high NA oil immersion objective with broadband illumination provided by a xenon lamp. Refractive index fluctuations are measured by sampling backscattered light each at wavelength of 500 to 700 nm using a combination of a liquid crystal tunable filter (LCTF) and a CMOS camera. Nanoscale differences in chromatin structure are measured as Σ, which quantifies the spatial fluctuations in the z-direction (25). HT29-SA1shRNA and HT29-Scr cells were maintained at 37°C with CO₂ and imaged in their normal media seeded onto petridishes with coverslip bottoms (MatTek). To calculate the average Σ, nuclear regions of interest (ROI) were selected and averaged over all of the cells and biological repeats (n ∼ 80 nuclei, per group). Standardized Student t tests generated 2-tailed P value (assuming unequal variances) using Microsoft Excel.

HT29-SA1shRNA and HT29-Scr cells were seeded in triplicates at a density of 2.5 × 10³ cells/10-cm dishes in McCoy's medium containing 1.0% FBS. After 2 weeks of growth, the cells were fixed and stained with crystal violet stain (0.1%, w/v) in 20 nmol/L 4-morpholinepropanesulfonic acid (Sigma Chemicals). Colonies were counted and quantified as average of 3 independent experiments.

HT29-Scr and HT29-SA1shRNA cells were seeded in multiples of 8 in a 96-well plate (10,000 cells per well) in complete medium. After overnight incubation, the medium was replaced with 1% FBS and penicillin/streptomycin (100 μL/mL) containing media. This was followed by the addition of 10 μL of the Cell Proliferation Reagent, WST-1 (Roche), to each well and absorbance measured at 450 nm after 4 hours of incubation (26). From these measurements, the values recorded at the reference wavelength (600 nm) were subtracted, following the manufacturer's guidelines. Data analysis was performed for 3 independent experiments. **, P = 0.0002.

Appropriate Excel 2010 statistical tools were used for determining statistical significance. AUROC curves were plotted using STATA 8 software. The 2-tailed Student t test was utilized as appropriate.

IHC expression analysis for SA1 cohesin was performed on human tissue arrays with 92 samples as discussed in Materials and Methods. Ten 40× fields from each specimen were scored (0–3; with 0 being no intensity and 3 very strong). SA1 immunoreactivity was found to be expressed mostly in the nucleus and partially in the cytoplasm. As shown in Fig. 1A, of the cohesins studied, SA1 was markedly reduced in adenomatous polyps (by 62%, P = 0.001) with a progressive loss detected in adenocarcinomas (75%; P = 0.0001; data not shown). To investigate SA1 as a marker of field carcinogenesis, IHC was performed in rectal biopsies obtained from patients with (n = 33) or without (n = 36) any concurrent neoplasia. As demonstrated in Fig. 1B, a marked downregulation was also observed in rectal mucosa from patients harboring colonic adenomatous polyps compared with patients with no polyps (P = 0.052). These studies demonstrate that SA1 is progressively lost during colon carcinogenesis.

To determine the performance of SA1 as a risk stratification biomarker, mRNA expression of SA1 was quantified by RT-PCR in the rectal biopsies as described in Materials and Methods. As shown in Fig. 2A, baseline expression of rectal mucosal SA1 was observed to be significantly higher in Caucasians than un African-Americans (73%; P = 0.003). In the entire cohort, SA1 expression was significantly downregulated in subjects harboring colonic adenomas (by 47.3%; P = 0.035) consonant with role as a biomarker for field carcinogenesis (Fig. 2B). When stratified by race, African-Americans with adenoma manifested strikingly greater and statistically significant reduction in SA1 expression (by 60.6%; P = 0.012) compared with African-Americans without adenomas. Caucasians on the other hand demonstrated lesser and statistically insignificant decrease in SA1 expression in subjects with adenomas compared with those without adenomas (44.7%; P = 0.113), suggesting a race-specific alteration. Likewise, as shown (Fig. 2C), SA1 mRNA expression was also found to be significantly lower in colon adenocarcinomas from African-Americans than from Caucasians. Figure 2D depicts the ROC curve demonstrating diagnostic performance of rectal mucosal SA1 as a biomarker to risk-stratify subjects for presence of adenomas. Evidently, the biomarker sensitivity of SA1 was low for all subjects (AUROC = 0.606) or Caucasians (AUROC = 0.585), but the test performed significantly better when applied only in African-Americans (AUROC = 0.724).

Furthermore, to elucidate the mechanism of SA1 gene silencing during colon carcinogenesis, we investigated potential epigenetic modifications such as DNA methylation and/or histone deacetylation as catalyzed by DNA methyl transferases (DNMT) and histone deacetylases (HDAC), respectively (Fig. 3A). For these studies, we exposed HT29 cells to either a demethylating agent (5-Aza-dC) or an HDAC inhibitor (sodium butyrate; NaB). After 48 hours, SA1 was found to be overexpressed in both 5-Aza-dC- and NaB-treated HT29, suggesting important relevance of epigenetic modulations in SA1 repression. Furthermore, as methylation of histones can modify chromatin structure by increasing or decreasing transcription depending on which amino acids are methylated and how many methyl groups are attached, we analyzed the effect of SA1 ablation on such methylation patterns. We observed that whereas SA1 knockdown in HT29 cells specifically reduced the trimethylation at histone marker H3 lysine 4 (H3K4), it increased trimethylation at H3 lysine 27 (H3K27) residues without altering monomethylation marks at either site (Fig. 3B). Histone marker H3K27, when trimethylated, is known to shut down transcription by tightly associating with inactive gene promoters, whereas trimethylation of H3K4 produces an opposite effect. Also, trimethylations of another marker H3 lysine 9 (H3K9) showed no modifications by SA1 knockdown.

Nucleosomes, which encapsulate 147-bp segment of DNA wrapped around the histone octamer 1.65 times (27), are not randomly distributed but carefully positioned in certain genomic regions to impact transcriptionally activity. Nucleosome positioning is determined by the combination of DNA sequence, nucleosome remodeling enzymes, and transcription factors. To verify that SA1 has a direct bearing on the higher order chromatin structure as indicated by chromatin accessibility to nuclease digestion, we examined the susceptibility of HT29 cells (control) and SA1 shRNA–transfected HT29 cells to MNase digestion (Fig. 4A). Comparison of the control and SA1 shRNA–transfected HT29 cells showed different MNase digestion patterns. The variation in the pattern of the digested DNA ladder is reflective of changes in the nucleosomal occupancy. The DNA ladder bands correspond to either the mononucleosome (147 bp) or polynucleosomes, in successively increasing size. The ladder shows an increase in intensity with increasing concentration of the MNase enzyme, indicating a higher degree of digestion as the ratio of enzyme to DNA increases. The densitometric analysis shows the difference in intensity of the low-molecular-weight bands corresponding to mono-, di-, and tri-nucleosomes, marked as N1, N2, and N3, respectively, between control and SA1 shRNA–transfected HT29 cells. The results indicate that the internucleosomal DNA linker may be less accessible for MNase digestion in the SA1 shRNA–transfected HT29 cells compared with controls.

Nanocytology (PWS) can be measured on live cells using the next-generation iteration BaSIS microscopy (13). BaSIS quantifies the spatial fluctuations in the z-direction, measured as Σ, with sensitivity to structures between 20 and 200 nm that can quantify the dynamics of the nanomolecular organization in live cells without using exogenous labels. Control and SA1 shRNA–transfected HT29 cells were imaged in glass-bottom petridishes (Fig. 4B1 and B2). The Σ was calculated for each nucleus and averaged over 3 experiments. We found that the nuclear Σ decreases in SA1 shRNA–transfected HT29 cells SA1-KD cells compared with control cells (−11%, n ∼ 80 cells per group, P < 0.01). These data suggest that depletion of SA1 alters the nanoscale higher order chromatin structure, leading to more homogeneity of nuclear structure.

For these studies, we first performed a colony formation assay by separately seeding 2,500 HT29 cells and SA1 shRNA–transfected HT29 cells in 10-cm dishes that were cultured for 2 weeks. Cells were fixed and stained with 0.1% crystal violet, and representative photographs were taken for HT29 and SA1 shRNA–transfected HT29 cells. As shown in Fig. 5A, SA1 shRNA–transfected HT29 cells showed significant increase in colony formation as compared with control (HT29 cells; *, P = 0.001). Next, we performed WST-1 proliferation assay. HT29 and SA1 shRNA–transfected HT29 cells were seeded in multiples of 8 in 96-well plates and their growth was monitored by measuring absorbance after incubation with WST-1 reagent. Data are represented as the mean absorbance (OD_{450_nm} − OD_{(Ref = 600_nm)}) of replicates for 3 independent experiments. The SA1 shRNA–transfected HT29 cells showed significant increase in proliferation as compared with control HT29 cells (*, P = 0.0002; Fig. 5B). Finally, we performed Western blot analysis to measure the expression of proliferation marker PCNA and β-catenin. Immunoblotting revealed significant upregulations in protein levels of PCNA and both phosphorylated (pS552) β-catenin and total-β-catenin with siRNA-mediated SA1 downregulations in HT29 cells (Fig. 5C). We replicated the studies in another human colorectal cancer cell line SW480 and found similar results (Fig. 5D).

We demonstrate herein, for the first time, that cohesin SA1 is lost early during colon carcinogenesis. Our data indicate that SA1 expression is suppressed in field carcinogenesis potentially through epigenetic events. Intriguingly, the SA1 expression appeared to have a racial predilection with lower expression in endoscopically normal rectal mucosa and colorectal cancers in African-Americans compared with Caucasians. For risk stratification, rectal SA1 performed somewhat better in African-Americans than in Caucasians underscoring the need to evaluate biomarkers in context of race. From a biologic perspective, the cell culture studies suggested that SA1 loss may modulate key pathways in colon carcinogenesis including β-catenin signaling. Finally, using the novel BaSIS (live cell PWS) technology, we show that SA1 altered higher order chromatin, and the transcriptional implications are further supported by histone mark analysis.

Higher order chromatin has been increasingly realized to be a major determinant of transcriptional activation and abnormalities in its organization are linked to cancer (28). In a mammalian cell, for the chromatin to fold into a relatively smaller nuclear confines (∼10 μm), DNA (∼2 m long) undergoes several levels of compactions. Nucleosomes form the fundamental units of compaction in which DNA is tightly wound around (1.7 turns) the histone core (11 nm) resembling beads on a string that densely fold in the nucleus. However, despite the enormous level of compaction and organization, chromatin selectively makes DNA accessible to interact with specific protein regulators to regulate transcription. The structure is complex with several groups, suggesting that chromatin architecture may represent a compact polymer state under many levels of organization defined as a fractal globule (29, 30). For instance, SWI-SNF family of proteins (such as Arid1a which is well established in colon carcinogenesis) is thought to expose promoter regions through moving nucleosomes (31). The cohesins, on the other hand, serve as modulators of chromatin looping which allows the placement of enhancer elements upstream of promoters to enhance gene transcription. To initiate appropriate gene functions, cohesin proteins, such as SA1, complexed with CTCF (a zinc finger DNA-binding protein) at the promoter region (32). Indeed, in many cancers including colorectal cancers, higher mutation frequency has been attributed to the CTCF/cohesin DNA-binding sites rather than in cohesin proteins alone (10). While there have been relatively few studies demonstrating changes in the protein expression, some reports suggest that cohesin levels may have prognostic value (33). To our knowledge, this study is the first to link cohesin levels to initiation of carcinogenesis.

It is striking that these changes in SA1 occur in the histologically normal mucosa consonant with a role in field carcinogenesis. Field carcinogenesis leads to a permissive milieu for tumorigenesis that reflects the molecular interplay between both genetic susceptibility and impact of exogenous risk factors (such as diet; smoking that introduces toxins in the fecal stream). Indeed, despite being histologically normal, there are profound molecular alterations (genomic, methylation, microRNA, proteomics, etc.) in patients harboring neoplasia. A parsimonious hypothesis assumes a central role for alterations in the higher order chromatin as evidenced by PWS, transmission electron microscopy (TEM), karyometric measurements, etc. (12, 13, 34). The biologic determinants of these changes are varied, but proteins that modulate higher order chromatin, including cohesins, may have an important role. Indeed, in keeping with this, our group has previously noted that PWS alterations from the endoscopically normal rectal mucosa had excellent diagnostic potential (35), supporting our results from the diagnostic studies with SA1. While others have shown that cohesin loss occurs at the adenoma stage (36), ours is the first study to show modulation at the earliest stages of carcinogenesis. A complexity in unraveling the biologic determinants is the pleotropic effects of cohesins including sister chromatid segregation. The relative importance of cohesion-induced alterations in gene expression versus aneuploidy in carcinogenesis is unclear. For instance, some reports in bladder cancer linked SA1 to aneuploidy (37), whereas others found that this tumor suppressor gene promoted bladder malignancies regardless of ploidy (38). SA1 may have other functions such as being downstream of important drivers in colon carcinogenesis such as Wnt signaling, etc. (39). The functional importance is underscored by the multiple developmental abnormalities in germline mutations such as noted in Cornelia de Lange Syndrome (40).

Detection and identification of markers of field carcinogenesis may have strong clinical ramifications. For instance, field carcinogenesis is the biologic underpinning for the clinical guidelines on post-polypectomy surveillance (41). Several groups, including ours, have focused on exploiting the field carcinogenesis for risk stratification to assist personalized screening given its remarkable inefficiency. Utilizing novel array of optical modalities, our group has provided ample evidence that minimally invasive interrogation of the rectal epithelium could risk-stratify patients for colorectal cancer (42, 43). Furthermore, field carcinogenesis detected in the uninvolved rectal mucosa was reported to be indicative of nanoscale changes in higher order chromatin and early tumorigenesis (12, 13).

The mechanism of SA1 loss in colon carcinogenesis is uncertain. Previous reports suggest that mutational inactivation in cohesin proteins as such may be infrequent, alternatively making epigenetic modulations of this tumor suppressor gene a cogent focus a logical candidate. Our data indicate that SA1 may be impacted upon by both histone acetylation and methylation, 2 common themes in cancer biology. Methylation is particularly attractive given its well-established role in field carcinogenesis and several reports suggest race-based modulation, although it needs to be emphasized that our data only relate to cell culture (44, 45). However, the precise modalities of gene regulation that are involved in SA1 loss in early colonic neoplastic transformation remains to be elucidated and other transcriptional or posttranslational processes may be involved.

To understand the functional biology of reduced SA1 levels, we found that SA1 knockdown resulted in increased cellular proliferation. The knockdown model was in a way utilized to recapitulate the observed lower basal levels of SA1 in African-Americans. The increase in cell proliferation in SA1-knockdown HT29 cells (increased WST and PCNA) was corroborated by activation of key signaling pathways including β-catenin. Future studies will focus on further understanding the biologic implications of SA1 loss.

One of the important facets of this work stems from the fact that there has been limited previous appreciation of the demographic considerations in biomarker discovery. On these lines, we have previously demonstrated that specific microRNAs in the field carcinogenesis have gender-specific implications (46). Thus, while the performance is modest, to our knowledge, this is the first demonstration of racial differences in cancer biomarkers for screening. Furthermore, as there is dramatically less basal expression of SA1 in African-Americans, it may suggest a potential mechanism for the disproportionate toll of colorectal cancer in African-Americans.

With regards to potential long-term clinical implications, these center on the paramount nature of risk stratification is paramount for colorectal cancer screening especially in disadvantaged populations. Finding less invasive techniques and tissue acquisition potentially at the point of care (rectal swab) to help identify relatively high-risk individuals may improve compliance of the minority patients with more invasive testing such as colonoscopy. This report may herald translation of recent knowledge regarding biologic differences in the colon carcinogenesis between races to clinical applications such as risk analysis.

This study has many strengths including the novelty of our findings with SA1 in specific and cohesins in general. The observation that SA1 was lost at the earliest stages (field carcinogenesis) was striking and its profound nature strongly implicates high-order chromatin regulators as drivers of colon carcinogenesis. The combination of protein and message in a prospectively collected cohort undergoing colonoscopy is a powerful resource, and the diverse cohort enables insights into racial disparities. Moreover, the demonstration of the impact of SA1 on high-order chromatin is compelling, as we utilized both conventional (MNase) and novel (BaSIS) techniques.

Our study has several limitations that should be acknowledged. First, the number of patients is not large and thus we were unable to perform subgroup analysis on different adenoma types (serrated, advanced, etc.). Second, the expression data were not coupled with mutational data, although this is mitigated by the relative rarity of cohesin mutations in colorectal cancer (∼2%; ref. 47). Third, SA1 was taken as a candidate approach although, as noted here and in a previously reported pilot study, and although the expression of other cohesins including SA2, SMC3, and CTCF was also measured, it was not an exhaustive evaluation in patient populations. Fourth, racial data were self-reported and recent studies have impugned reliability (48). However, while imperfect this is currently the state-of-the-art and should not bias our results. Finally, from a clinical perspective, SA1 performance as a biomarker was modest but the vast majority of the neoplastic lesions were small adenomas. Our previous work has indicated that field carcinogenesis is more pronounced with more significant neoplasia (e.g., advanced adenomas) with diminutive adenomas engendering limited alteration (12). Furthermore, as the major novelty of the work is the demonstration of race-specific biomarkers, the actual performance of SA1 is secondary.

In conclusion, we demonstrate for the first time that cohesin SA1 is an important tumor suppressor gene in early colon carcinogenesis. SA1 modulates higher order chromatin alterations that are the hallmark of field carcinogenesis. Furthermore, lower expression of cohesin SA1 in African-Americans versus Caucasians may provide a potential biologic explanation for the increased incidence and mortality of colorectal cancer in African-Americans. Finally, this provides the proof-of-concept of the need to factor race into biomarker development. Future studies are needed to understand the ramifications of cohesin dysregulation for cancer screening, prevention, and therapeutics.

A. Radosevich is a Research Scientist at Nanocytomics. H.K. Roy has Ownership Interest (including patents) in Nanocytomics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Backman, H.K. Roy

Development of methodology: N. Momi, H.K. Roy, R.K. Wali

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Momi, M.D. Cruz, A.H. Calderwood, Y. Stypula-Cyrus, L.M. Almassalha, A. Chhaparia, B. Latif, H.K. Roy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.H. Calderwood, Y. Stypula-Cyrus, C.R. Weber, A. Radosevich, B. Latif, H.K. Roy, R.K. Wali

Writing, review, and/or revision of the manuscript: N. Momi, M.D. Cruz, A.H. Calderwood, Y. Stypula-Cyrus, A.K. Tiwari, V. Backman, H.K. Roy, R.K. Wali

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.D. Cruz, A. Chhaparia, A.K. Tiwari, B. Latif, H.K. Roy

Study supervision: V. Backman, H.K. Roy, R.K. Wali

The authors thank Ms. Beth Parker for excellent support in manuscript preparation.

H.K. Roy and V. Backman were awarded R01 CA165309, R01 CA156186, R01 CA200064, and R01 CA183101. H.K. Roy and R.K. Wali were awarded R03 CA195143.

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.