Association of Adiponectin and Vitamin D With Tumor Infiltrating Lymphocytes and Survival in Stage III Colon Cancer

Abstract Background Adipocyte-derived adiponectin may play a role in the host inflammatory response to cancer. We examined the association of plasma adiponectin with the density of tumor-infiltrating lymphocytes (TILs) in colon cancers and with vitamin D, clinicopathological features, and patient survival. Methods Plasma adiponectin and 25-hydroxyvitamin D [25(OH)D] were analyzed by radioimmunoassay in 600 patients with stage III colon cancer who received FOLFOX-based adjuvant chemotherapy (NCCTG N0147 [Alliance]). TIL densities were determined in histopathological sections. Associations with disease-free survival (DFS), time to recurrence, and overall survival were evaluated by multivariable Cox regression adjusting for potential confounders (ie, body mass index, race, TILs, and N stage). All statistical tests were 2-sided. Results We found a statistically significant reduction in adiponectin, but not 25(OH)D, levels in tumors with high vs low TIL densities (median = 6845 vs 8984 ng/mL; P = .04). A statistically significant reduction in adiponectin was also observed in obese (body mass index >30 kg/m2) vs nonobese patients (median = 6608 vs 12 351 ng/mL; P < .001), in men vs women (median = 8185 vs 11 567 ng/mL; P < .001), in Blacks vs Whites or Asians (median = 6412 vs 8847 vs 7858 ng/mL; P < .03), and in those with fewer lymph node metastases (N1 vs N2: median = 7768 vs 9253 ng/mL; P = .01). Insufficiency of 25(OH)D (<30 ng/mL) was detected in 291 (48.5%) patients. In multivariable analyses, neither adiponectin nor 25(OH)D were associated with a statistically significant difference in DFS, overall survival , or time to recurrence in models adjusted for potential confounders. We found a statistically significant association of TILs with prognosis, yet no such interaction was observed for the association of adiponectin with TILs for DFS. Conclusions Lower circulating adiponectin levels were associated with a statistically significant increase in TIL densities in colon cancers, indicating an enhanced antitumor immune response. In contrast to TILs, neither adiponectin nor 25(OH)D was independently prognostic.

Adiponectin, the most abundant hormone secreted by adipose tissue, senses metabolic stress and modulates metabolic adaption by targeting the innate immune system under physiological and pathological conditions (5). Epidemiological studies suggest an inverse association between plasma adiponectin levels and risk of developing CRC (6,7). This inverse relationship was observed among men, but not women, in the Nurses' Health Study and Health Professionals Follow-up Study (6). In a meta-analysis of patients with established CRC, a statistically significant decrease in adiponectin levels was observed compared with non CRC controls (8). Limited data suggest an association between prediagnostic plasma adiponectin and risk of CRC-specific and overall mortality (9). However, the only prospective study of adiponectin measured at the time of diagnosis of CRC was not prognostic (10), and postdiagnosis data are lacking.
Plasma adiponectin was found to be associated with vitamin D levels in non-CRC patients (11). Vitamin D insufficiency (<30 ng/mL) is relatively common in healthy individuals and in CRC populations (12). Serum 25-hydroxyvitamin D [25(OH)D] was inversely associated with CRC risk (13), and some studies suggest that 25(OH)D insufficiency in patients with established CRC may be associated with worse clinical outcome (14)(15)(16). Individuals deficient in vitamin D had statistically significantly higher levels of the serum inflammatory biomarkers interleukin-6 (IL-6) and Creactive protein (17). Interestingly, a prior study found that a high 25(OH)D level was associated with a lower risk of developing CRC with an intense immune reaction (18), suggesting that vitamin D may influence the tumor-host interaction. In a study of the 25(OH)D score measured post-CRC diagnosis, its association with CRC-specific mortality differed by the extent of peritumoral lymphocytic reaction (19).
To date, adiponectin has not been analyzed in relationship to the tumor immune microenvironment. We sought to test the hypothesis that adiponectin may play a role in the host inflammatory response to colon cancer, indicated by tumor-infiltrating lymphocytes (TILs). In patients with CRC, studies indicate that TILs can independently predict recurrence and survival (20)(21)(22). However, only limited data exist for adiponectin or vitamin D in patients with established CRC, and their impact on survival is largely unknown. We determined the association of postsurgical plasma adiponectin levels with TIL densities, 25(OH)D, clinicopathological features, and clinical outcome in patients with stage III colon cancers from a phase III adjuvant trial of FOLFOXbased chemotherapy (NCCTG N0147 [Alliance]) (23).

Study Population
Among 2686 patients with resected stage III colon cancer who had been randomly assigned to adjuvant FOLFOX alone or combined with cetuximab, we randomly selected 600 patients (300 per arm) for analysis of adiponectin and 25(OH)D (NCCTG N0147; ClinicalTrials.gov Identifier: NCT00079274). Right-sided tumors were defined as located proximal to the splenic flexure. N1 tumors had 1-3 metastatic regional lymph nodes; N2 tumors had at least 4 nodes. All patients provided written informed consent at the time of clinical trial enrollment. All data analyses shown here were approved by Mayo Clinic Institution Review Board.

Adiponectin and Vitamin D Assays
Plasma adiponectin and 25(OH)D were analyzed in processed and stored blood obtained at adjuvant study registration. Biomarker concentrations were measured by radioimmunoassay (laboratory of M. Pollak, McGill University, Montreal, QC, Canada) as previously described (24). Circulating levels of total and high molecular weight adiponectin were measured in duplicate using standard enzyme-linked immunosorbent assay methods. Because total and high molecular weight adiponectin levels were almost perfectly correlated (Spearman P ¼ .99), we report total adiponectin. 25(OH)D was extracted from serum or plasma with acetonitrile, and samples were then assayed using an equilibrium radioimmunoassay procedure using an antibody specific for 25(OH)D (24). Blinded quality control (QC) samples were included with test samples, and pooled QC specimens were added to each of the batches. For both assays, masked QC samples were interspersed among case samples. All laboratory personnel were blinded to patient outcomes. The mean coefficient of variation of the assay was 8%.

Histologic Examination of TILs
A representative hematoxylin and eosin-stained tumor section from each patient was scanned at low power to identify areas with the most intraepithelial TILs. Once identified, 5 consecutive 40X fields were counted, and mean TILs per high power field (HPF) were calculated by dividing the total number of TILs by 5. All cases were scored independently by 2 gastrointestinal pathologists blinded to clinical and molecular data (25). Cutoffs for TIL densities were previously determined in N0147 tumors based on their association with patient (disease-free survival [DFS]) and categorized as low ( 3 per HPF) vs high (>3 per HPF) TILs (25).

DNA Mismatch Repair (MMR) Status and KRAS/BRAF Mutation Analysis
Prospectively collected tumor tissues were analyzed for MMR status by analysis of MLH1, MSH2, and MSH6 proteins using immunohistochemistry. Tumors with deficient MMR were defined as having absent expression of 1 or more MMR protein. Testing for a BRAF (c.1799T>A V600E) mutation in exon 15 and KRAS mutations in codons 12 and 13 of exon 2 was previously described (26).

Statistical Analysis
Adiponectin and 25(OH)D were analyzed as continuous variables in the primary analysis. As binary variables, adiponectin was dichotomized at the median, and 25(OH)D was dichotomized at 30 ng/ml with lower levels regarded as insufficiency (12). The association of adiponectin with TILs was prespecified as the primary analysis and was analyzed by the Kruskal Wallis test. Associations between adiponectin and 25(OH)D and with clinicopathological features were assessed by Fisher exact, Pearson v 2 , t test, and Kruskal-Wallis tests as appropriate. The association between adiponectin and clinical outcomes were as follows: DFS, time from randomization to recurrence or death from any cause; overall survival (OS), time from randomization to death from all causes; and time to recurrence (TTR) as time from randomization to recurrence. Distributions of DFS, OS, and TTR were estimated by the Kaplan-Meier method. Biomarkers were analyzed in relationship to clinical outcome as continuous variables using a relative risk model with cubic splines and as binary variables in multivariable Cox regression. Multivariable models included variables that were considered to be potential confounders. Interaction tests were performed. A 2-sided P value of less than .05 was considered statistically significant and was not adjusted for multiple comparisons. SAS software version 9.4 and R version 3.6.2 were used (SAS Institute, Cary, NC).

Association of Plasma Adiponectin and 25(OH)D With Patient Characteristics and TILs
The frequency distribution and median levels of plasma adiponectin and 25(OH)D are shown in Figure 1. We found a statistically significant decrease in the level of adiponectin in patients whose tumors had high vs low TIL densities (median ¼ 6845 vs 8984 ng/mL; P ¼ .04) ( Table 1). Furthermore, a statistically significant and inverse association between adiponectin level and body mass index (BMI) category was observed whereby obese patients had a lower median adiponectin level (median ¼ 6608 ng/mL) than did those of normal weight (median ¼ 10 693 ng/mL) or underweight (median ¼ 18 010 ng/mL; P < .001). A statistically significant decrease in adiponectin was observed in men vs women (median ¼ 8185 vs 11 567 ng/mL; P < .001), in those aged 65 years or younger (median ¼ 7663 vs 9279 ng/mL; P ¼ .007), and in Blacks vs Whites or Asians (median ¼ 6412 vs 8847 vs 7858 ng/mL; P < .03) ( Table 1). Findings by race were not explained by differences in BMI, which was similar by race. Adiponectin levels differed by N stage in that a lower level was associated with fewer regional lymph node metastases (N1 vs N2: median ¼ 7768 vs 9253 ng/mL; P ¼ .01). Adiponectin was not associated with a statistically significant correlation with 25(OH)D (Spearman r ¼ À.0038; P ¼ .93) or with T stage, DNA MMR, or the mutational status of KRAS or BRAF genes.
A statistically significant reduction in plasma 25(OH)D, as a continuous variable, was found in Blacks vs Whites or Asians (P < .001) ( Table 2). Differences by sex were statistically, but not clinically significant. A statistically significant decrease in 25(OH)D levels was found in the relatively few patients with poor performance status. Insufficiency of 25(OH)D (<0 ng/mL) was detected in 49% (291 of 600) of study participants. Consistent results were found for the dichotomous 25(OH)D by patient race (P < .001) and sex (P < .03) (Supplementary Table 1, available online). Specifically, insufficiency of 25(OH)D was more common in Blacks vs Whites or Asians (73.6% vs 45% vs 53.6%) and in women vs men (53.2% vs 44.3%), with both achieving statistical significance. 25(OH)D was not statistically significantly associated with BMI or adiponectin level. Neither the association of the continuous nor of the dichotomous 25(OH)D level with TIL density achieved statistical significance (Table 2;  Supplementary Table 1, available online).

Association of Plasma Adiponectin and 25(OH)D Levels With Patient Clinical Outcome
Univariately, plasma adiponectin was not statistically significantly associated with DFS as a dichotomous variable (cut at median; hazard ratio [HR] ¼ 0.98, 95% confidence interval [CI] ¼ 0.74 to 1.29; P ¼ .88) shown in a Kaplan Meier plot (Figure 2, A) or a continuous variable shown by cubic spline (Figure 2, C). No association was found between adiponectin and OS (HR ¼ 0.90, 95% CI ¼ 0.66 to 1.22; P ¼ .49) or TTR (HR ¼ 1.03, 95% CI ¼ 0.75 to 1.34; P ¼ 1.00) (data not shown). We then constructed a multivariable model that included adiponectin and potential confounding variables based on observed associations (Table 1) and the literature. The association of adiponectin with patient DFS did not achieve statistical significance as a continuous (adjusted HR [HR adj ] ¼ 1.01, 95% CI ¼ 0.98 to 1.04; adjusted P [P adj ] ¼ .57) or a dichotomous (data not shown) variable in a multivariable model adjusted for treatment arm, race, BMI, TILs, and N stage (Table 3). In this model, statistically significantly poorer DFS was observed for tumors with low vs high TILs (HR adj ¼ 1.77, 95% CI ¼ 1.11 to 2.81; P adj ¼ .02) and N2 vs N1 stage (HR adj ¼ 2.08, 95% CI ¼ 1.44 to 2.99; P adj < .001) ( Table 3). Adiponectin was not statistically significantly associated with OS or TTR as either a continuous or dichotomous variable.
We constructed 3-way interaction models that examined the association of adiponectin and 25(OH)D (adjusting for race and BMI) by TILs and N stage subgroups. No statistically significant relationships were found indicating that the observed associations were not interdependent. Of note, TIL density did not differ by BMI category (P ¼ .50) (data not shown). Furthermore and given observed differences in adiponectin and 25(OH)D levels by patient sex (Tables 1 and 2), analysis of 25(OH)D with clinical outcome variables (TTR, DFS, OS) by sex or TIL density failed to reveal any statistically significant associations after adjustment for covariates.

Discussion
We analyzed adiponectin and 25(OH)D in postsurgical blood samples from patients with stage III colon cancer treated in a phase III trial of adjuvant chemotherapy. Based on preclinical data suggesting that adiponectin may influence the host inflammatory response to cancer, we determined the association of adiponectin with TIL density in the tumor immune microenvironment. Lower adiponectin levels were associated with a statistically significant increase in TIL density, indicating an inverse relationship between adiponectin and antitumor immunity. This result is consistent with data in a murine model where adiponectin deficiency was associated with tumors with an increased inflammatory infiltrate compared with tumors in nonadiponectin-deficient mice (27). Adiponectin has been shown to reduce T-cell and B-cell recruitment, induce production of anti-inflammatory cytokines such as IL-10 and an inhibitor of metalloproteinase-1, and inhibit pro-inflammatory chemokines such as IL-6 and TNF-a (5,28). Furthermore, adiponectin was shown to impede inflammation-induced tumorigenesis (29). In contrast to adiponectin, we did not find a statistically significant association between the level of plasma 25(OH)D and TIL density. In another report, higher plasma 25(OH)D levels were associated with a lower risk of CRCs with a high density of CD3 þ T cells (18). We found that lower adiponectin levels were statistically significantly associated with cancers showing fewer regional lymph node metastases (N1 vs N2 stage). Relevant to this finding is our prior observation of statistically significantly higher TIL densities in N1 vs N2 stage colon cancers (P ¼ .0001) (25), yet no interaction was found between TIL densities, N stage, and adiponectin in the current study. Importantly, we confirmed the reported inverse relationship of adiponectin level to BMI in our dataset (30,31). A statistically significant and inverse relationship between adiponectin and BMI category was observed, with the lowest adiponectin levels found among obese patients. Although adiponectin is exclusively secreted by adipocytes, its paradoxical reduction in obesity may reflect reduced secretion from visceral fat rather than subcutaneous fat. Hypertrophic adipocytes synthesize monocyte chemotactic protein-1 leading to infiltration of macrophages in white adipose tissue accompanied by elevated local TNF-a and increased free fatty acid concentrations that suppress adiponectin secretion (32). Other factors may include dysregulation of the adiponectin gene (ADIPOQ) in obesity (33). We observed differences in adiponectin by sex whereby men had statistically significantly lower levels compared with women. This observation may be explained by higher serum androgens in men (34), and the finding of increased adiponectin levels in older vs younger men may be due to an age-related decline in testosterone levels. In this regard, adiponectin levels were shown to decrease after testosterone administration in hypogonadal men (35). In women, adiponectin has been shown to decrease with the transition to menopause (36). We found statistically significantly lower adiponectin levels among Blacks compared with Whites or Asians, which was not explained by differences in BMI by race. However, lower adiponectin levels among Blacks vs Whites was attributed to adiponectin's association with BMI in the population-based race-ethnic Northern Manhattan Study (37). Differences in adiponectin levels may also have a hereditary component, and further analysis of single nucleotide polymorphisms in adiponectin-related genes may provide further insight (38).
Whereas plasma adiponectin was associated with vitamin D levels in nonCRC patients (11), no such relationship was observed in our CRC cohort. We observed that nearly one-half of patients were vitamin D insufficient (<30 ng/mL), which far exceeds the 8% prevalence of vitamin D insufficiency in the US general population (39). We found a higher rate of vitamin D insufficiency among Black vs White or Asian. In addition to dietary factors and sunlight exposure, race can influence vitamin D status in that melanin in skin reduces the rate of vitamin D biosynthesis (12,40). Furthermore, racial differences in the prevalence of common genetic polymorphisms can influence vitamin D levels; for example, Blacks are more likely to have the T allele at rs7041 and less likely to have the A allele at rs4588 vs Whites (41). Differences by sex for the continuous vitamin D variable were modest; however, vitamin D insufficiency was statistically significantly more common in women vs men, which is consistent with an increased prevalence of hypovitaminosis D in women especially after menopause (42).
Despite the inverse association of adiponectin levels and TIL densities that we observed, adiponectin was not statistically significantly associated with patient DFS, OS, or TTR univariately or after adjustment for potential confounders. To our knowledge, these are the first data for postsurgical adiponectin and clinical outcome in CRC patients. Adiponectin at diagnosis of CRC was also not prognostic in a prospective study in 344 consecutive cases (10). Regarding prediagnostic adiponectin, 2 large observational cohorts found a statistically significant and independent association of higher levels with reduced CRCspecific and OS that was more evident in patients with metastatic disease (9). A potential explanation may be an elevation of adiponectin during weight loss, which is a known predictor of poor prognosis in cancer patients. Adiponectin is an insulinsensitizing hormone, and low levels are associated with an increased risk of type 2 diabetes independent of other risk factors (43). Adiponectin mediates its insulin-sensitizing effect through activation of AMP-activated protein kinase (AMPK) and peroxisome proliferator-activated receptor-alpha pathways with resultant suppression of tumor growth in animal models (44). As with adiponectin, plasma 25(OH)D was not statistically significantly associated with clinical outcome variables in our cohort. However, a study of resected stage III colon cancer patients found that a higher predicted 25(OH)D score was associated with improved survival outcomes (16). Of note, 25(OH)D scores were calculated based on serial patient questionnaire data, but not measurements in blood samples. Another study of 1598 patients with stage I to III CRC found that patients with the highest vs lowest tertile of postoperative plasma 25(OH)D levels had lower CRC-specific and all-cause mortality that achieved statistical significance (45). 25(OH)D levels, however, were not prognostic in patients with stage IV CRC (n ¼ 515) from a clinical trial cohort (46). We interpret our results and those in stage IV cancers to suggest that any effect of vitamin D on prognosis is attenuated or lost once the disease has metastasized either to regional lymph nodes (stage III) or to distant sites (stage IV). In contrast, a statistically significant association of TILs with prognosis was found in our cohort with low vs high TILs being associated with poorer patient survival as we previously reported (22,25).
Strengths of our study include same stage patients with uniform treatment in a clinical trial with long-term and meticulous follow-up data. Blood was collected from all patients prior to initiation of chemotherapy, such that the time period between biomarker sampling and initiation of adjuvant treatment was relatively constant. Whereas other studies have used cancer prediagnosis measurements (47), our study is the first to examine postdiagnostic plasma adiponectin levels in relation to survival outcomes in CRC patients. Study limitations include retrospective biomarker analysis that was not prespecified and the single assay timepoint. Importantly, measurement of 25(OH)D at a single timepoint was shown to provide an accurate assessment of an individual's long-term vitamin D status (48). Similarly, adiponectin levels were shown to remain stable over time (49). We cannot exclude the possibility that the modest sample size of our study may have contributed to the failure to detect a prognostic effect for either biomarker.
In conclusion, we identified a novel inverse association of adiponectin with the intratumoral antitumor immune response indicated by TIL density. Validation of this association in an independent cohort of patients with CRC is warranted. In contrast to TILs, neither adiponectin nor 25(OH)D was found to be prognostic.
The study was also supported by NCI R01CA210509 (to FAS).

Notes
Role of the funders: The funders had no role in the design of the study or in the data collection, analysis, or interpretation. Furthermore, the funders did not participate in the writing of the manuscript or in the decision to submit the manuscript for publication.
Disclosures: FAS is a co-inventor of intellectual property with Roche Ventana Medical Systems (Tucson, AZ) and may receive royalties paid to Mayo Clinic. No other relevant conflicts related to the subject matter of this manuscript are reported by the study authors.

Data Availability
Data from published Alliance or North Central Cancer Treatment Group (NCCTG) trials can be accessed by submission of an Alliance Data Sharing Request Form (concepts@allianceNCTN. org). Requests to utilize biomarker data published in this manuscript should also be directed to the corresponding author.