ZIP Code to Genomic Code

Abstract

Objective:

To determine the association between objective (geospatial) and subjective (perceived) measures of neighborhood disadvantage (ND) and aggressive breast cancer tumor biology, defined using validated social adversity-associated transcription factor (TF) activity and clinical outcomes.

Background:

ND is associated with shorter breast cancer recurrence-free survival (RFS), independent of individual, tumor, and treatment characteristics, suggesting potential unaccounted biological mechanisms by which ND influences RFS.

Methods:

We quantified TF-binding motif prevalence within promoters of differentially expressed genes for 147 tissue samples prospectively collected on the protocol. Covariate-adjusted multivariable regression analyzed objective and subjective ND scores with 5 validated TFs of social adversity and aggressive biology—pro-inflammatory activity [nuclear factor-κB (NF-kB), activator protein 1 (AP-1)], sympathetic nervous system (SNS) activity [cyclic 3′−5′ adenosine monophosphate response element-binding protein (CREB)], and protective cellular responses [interferon-regulatory factor (IRF) and signal transducer and activator of transcription (STAT)]. To clinically validate these TFs as prognostic biomarkers of aggressive biology, logistic regression and multivariable Cox proportional-hazards models analyzed their association with Oncotype DX scores and RFS, respectively.

Results:

Increasing objective ND was associated with aggressive tumor biology (up-regulated NF-kB, activator protein 1, down-regulated IRF, and signal transducer and activator of transcription) and SNS activation (up-regulated CREB). Increasing subjective ND (eg, threat to safety) was associated with up-regulated NF-kB and CREB and down-regulated IRF. These TF patterns were associated with high-risk Oncotype DX scores and shorter RFS.

Conclusions:

In the largest human social genomics study, objective and subjective ND were significantly associated with TFs of aggressive biology and SNS activation. These TFs also correlated with worse clinical outcomes, implicating SNS activation as one potential mechanism behind ND survival disparities. These findings remain to be validated in a national cohort.

Despite significant advances in breast cancer (BCa) screening, diagnosis, and treatment, women living in disadvantaged neighborhoods compared with advantaged neighborhoods have higher BCa mortality.^1–3 The drivers of this survival gap are multifactorial and largely impacted by access to care barriers. However, recent work shows that survival disparities persist between advantaged and disadvantaged neighborhoods independent of sociodemographic, access to care, tumor, and treatment factors, warranting further inquiry into how neighborhood disadvantage (ND) influences BCa survival.¹

Neighborhoods reflect complex environments with unique cultural, physical, and economic attributes, and ND contributes to the creation and persistence of under-resourced neighborhoods. This is quantitated with objective (eg, neighborhood-level income/employment, education, housing, household characteristics) and subjective (eg, one’s perceptions of their neighborhood with respect to threat to safety, esthetic quality, and social cohesion) ND indicators which all have known effects on overall health, cancer rates, and survival.^4–9 High ND communities experience considerably higher levels of social adversity-associated stress which leads to biological “wear and tear”. ^8,10 The impact of social adversity from ND, therefore, warrants consideration as a significant biological risk factor when studying BCa survival disparities.

The paucity of translational research on ND and how it influences BCa biology limits the development of potential targeted interventions and therapies to reduce health disparities. Therefore, a clear biological mechanistic explanation for how ND biologically impacts BCa survival merits further inquiry. To better understand these biological mechanisms, a growing literature in human social genomics research has begun to analyze how everyday life circumstances influence human gene expression and tumor biology, thus offering a distinct approach to bridge this knowledge gap.^11–14 Social genomics-based research, initially established for social disparities in chronic disease and mortality, identified a recurrent pattern of heightened inflammatory transcription factor (TF) activity [nuclear factor-κB (NF-kB) and activator protein 1 (AP-1)] that accompanied down-regulated protective cellular immune response TF activity [interferon-regulatory factor (IRF) and signal transducer and activator of transcription (STAT)] in individuals exposed to various adverse life circumstances, such as chronic loneliness and social isolation.^9,12,15–18 This gene expression pattern was referred to as the “conserved transcriptional response to adversity” (CTRA) and its activation in blood leukocytes was found to be mediated through the sympathetic nervous system (SNS), which is activated as a result of one’s perceived adversity/threat in response to their environment.¹⁴ Within the tumor, these CTRA-induced immunological alterations contribute to aggressive tumor biology and tumor progression. As CTRA-conditioned monocytes, macrophages, and lymphocytes enter the tumor microenvironment, they up-regulate pro-inflammatory TFs and down-regulate TFs that mediate protective cellular immune responses.^{11,14,16–18} This pattern, previously validated with immunohistochemistry, creates an environment that fosters a pro-metastatic state, invigorating macrophage recruitment, angiogenesis, and tumor invasion, while inhibiting protective cellular immune responses and programmed cell death¹⁹ (Fig. 1).

FIGURE 1.

“Social signal transduction” conceptual model describing how the effects of objective and subjective neighborhood disadantage (ND) lead to aggressive tumor biology. Social adversity from objective and subjective perceptions of ND trigger neural and endocrine responses through activation of the sympathetic nervous system (SNS) leading to the release of norepinephrine at nerve terminals, activation of β-adrenergic receptors on adjacent cells, and stimulation or repression of specific transcription factors (TFs)s in response to the intracellular cAMP/PKA signaling pathway.^6,20,21 cAMP-mediated activation of PKA subsequently phosphorylates TFs such as CREB. Collectively, this β-adrenergic signaling stimulates the activity of immune regulatory TFs such as AP-1 and NF-kB and inhibits the activity of interferon response factors (IRF and STAT), to induce a pattern of social adversity-associated gene expression consistent with the CTRA in circulating blood leukocytes. This blood leukocyte CTRA gene expression profile influences the circulating blood monocyte pool that subsequently infiltrates the tumor to form tumor-associated macrophages which promote aggressive tumor biology, marked by increased activity of pro-inflammatory TFs NF-kB and AP-1, leading to a pro-metastatic niche by downstream stimulation of macrophage infiltration, inflammation, angiogenesis, epithelial-mesenchymal transition and tumor invasion, and by inhibition of cellular immune responses and programmed cell death. cAMP indicates cyclic 3′−5′ adenosine monophosphate; PKA, protein kinase A.

Despite this rich body of social genomics research, many studies are preclinical or based on small samples, and there remains a critical knowledge gap as to whether these social genomic TF activity patterns translate to more aggressive BCa tumor biology and worse clinical predictors and outcomes in larger cohorts. The main objective of this study was to evaluate how objective and subjective measures of ND impact human BCa tumor biology based on predefined TF patterns of aggressive biology and SNS activation in a well-characterized and diverse prospective cohort. As a secondary objective, we evaluated the association between these TFs and predictors of clinical outcomes [Oncotype DX and TFs associated with response to treatment (ESR1 and BCL-6)], as well as recurrence-free survival (RFS). Our primary hypothesis is that objective and subjective measures of ND will be associated with up-regulation of SNS activation [cyclic 3′−5′ adenosine monophosphate response element-binding (CREB) protein] and aggressive biology (up-regulated NF-kB and AP-1 and down-regulated IRF and STAT). We also hypothesize that these TF activity patterns will be associated with high-risk Oncotype DX scores and shorter RFS, validating these social adversity-associated TFs as potential clinical biomarkers of aggressive BCa biology.

METHODS

Study Design and Population

We included a population-based sample of 147 non-Hispanic White, non-Hispanic Black, and Hispanic women diagnosed with stages I to III estrogen receptor (ER)+/human epidermal growth factor receptor 2 (HER2) or triple-negative breast cancer enrolled in our Miami Breast Cancer Disparities Study (MBCDS) between 2020 and 2022 (Supplemental Digital Content Fig. 1, http://links.lww.com/SLA/F59). Eligibility criteria for enrollment into the MBCDS, a prospective longitudinal cohort study, includes all English, Spanish, or Haitian Creole-speaking women > 18 years of age with BCa treated at our NCI-designated cancer center and sister safety-net hospital with fresh frozen tissue samples from surgery. All enrolled women were invited to complete the NSEAS survey (Supplemental Digital Content Table 1, http://links.lww.com/SLA/F59) and provide residential history. Women who were previously diagnosed with BCa were excluded from this study. The tissue collection protocol is detailed in the Supplemental Methods (Supplemental Digital Content, http://links.lww.com/SLA/F59).

To assess potential generalizability discrepancies between our hospital-based study and corresponding population-based data, we accessed the Florida Cancer Data System, which provided information on women diagnosed with ER⁺/HER2⁻ or ER−/HER2⁻ BCa between 2016 and 2017 (n = 2845) and residing in Miami-Dade County. This county serves as the primary reference area for the study catchment area. We evaluated race/ethnicity and stage (localized, regional, distant, unknown), as defined by the standards set forth by the North American Association Central Cancer Registries Association (Supplemental Digital Content Table 2, http://links.lww.com/SLA/F59).

This study followed the “Strengthening the Reporting of Observational Studies in Epidemiology” reporting guidelines. The investigators obtained informed consent from each participant or each participant’s guardian and all human investigations were performed after approval by our local Human Investigations Committee and in accordance with an assurance filed with and approved by the Department of Health and Human Services where appropriate. This study was approved by the University of Miami IRB.

Data Collection

Objective Measure of Neighborhood Disadvantage: Area Deprivation Index

The Area Deprivation Index (ADI) was used to measure objective ND.²² It is a validated, neighborhood-level composite index reflecting 17 dimensions of social determinants of health within the domains of housing, income, employment, and education, captured in the American Community Survey and U.S. Census Survey data through principal components analysis methodology. The ADI state rankings range from 1 to 10, with disadvantage reflected by higher scores. The state ADI composite score was calculated at the census-block group level using the ADI mapping atlas and participant addresses.

Subjective Measure of Neighborhood Disadvantage: Neighborhood Social Environment Adversity Survey

To understand subjective, perceived social adversity associated with living in ND, the Neighborhood Social Environment Adversity Survey (NSEAS) was administered to patients before surgical intervention. The NSEAS is a validated, 10-item survey from the Multi-Ethnic Study of Atherosclerosis that encompasses dimensions of neighborhood social environment adversity: safety, esthetic quality, and social cohesion.^10,23 Participants report their levels of agreement on 4 and 5-point item scales (for items, Supplemental Digital Content Table 1; http://links.lww.com/SLA/F59). Responses were standardized as Z-scores, which set the mean at zero and measured the number of SDs from the mean, calculated for each dimension and a composite score by summing Z-scores. Items were reverse coded as necessary so that higher values indicated greater adversity. Residential history was provided by self-report as the number of years lived at each of their addresses over their lifetime.

Covariates

Covariates were selected a priori based on subject matter knowledge, causal diagrams, and the current literature and captured from the electronic medical record. We collected sociodemographic data (age, self-identified race, and ethnicity), tumor characteristics (ER/HER2) status, tumor grade, clinical stage (AJCC Eighth Edition I–III), and treatment (surgery, radiation, chemotherapy, and endocrine therapy). Race was considered a sociopolitical construct. All patients received guideline-concordant multidisciplinary treatment. We also accounted for comorbidities that may affect pro-inflammatory TFs [eg, body mass index (BMI), alcohol, and smoking].²⁴

Clinical Predictors and Outcomes: Oncotype DX Scores and Recurrence-free Survival

Oncotype DX scores were collected from the electronic medical record for those who clinically met the guidelines for Oncotype DX testing. RFS was defined by the STEEP 2.0 Guidelines,²⁵ includes invasive ipsilateral BCa, loco-regional, invasive, or distant recurrence, death from BCa, death from a non-BCa cause, and death from an unknown cause. Patients without recurrence or death at the time of analysis were censored for RFS at the date they were last known to be alive and disease-free based on medical chart review.

RNA Extraction and Sequencing

Fresh frozen tumor specimens from patients with BCa were obtained from the MBCDS cohort biospecimen bank at the Sylvester Comprehensive Cancer Center, University of Miami, Miami, FL. RNA was extracted and sequenced for downstream statistical analysis (Supplemental Methods, Supplemental Digital Content, http://links.lww.com/SLA/F59).

Statistical Analyses

Sociodemographic, Lifestyle, Behavior, Tumor, Treatment, and Neighborhood Social Environment Adversity Survey Characteristics

Patient, tumor, and treatment factors, along with NSEAS scores were detailed for the total cohort and dichotomized above and below the median ADI. Descriptive statistics with χ² and analysis of variance (ANOVA) for categorical variables and t tests for continuous variables were conducted using STATA, version 17. All P values were 2-sided, alpha level was 0.05.

Primary Outcome: Social Adversity-associated Transcription Factors of Aggressive Tumor Biology and Sympathetic Nervous System Activation

To evaluate whether we needed to control for clustering within our sample, we identified that our study cohort catchment area is predominantly distributed across 358 block groups and 147 cases, giving a mean of 0.42 cases/block group. Thus, we did not have sufficient clustering to use a multilevel regression model.

The primary outcome analyses tested for 5 validated social-adversity TFs (chosen a priori) of aggressive biology and SNS activation: up-regulated NF-kB and AP-1 families of TFs, down-regulated IRF and STAT families of TFs, and up-regulated CREBs which mediates B-adrenergic SNS signaling [with parallel analyses of a distinct stress-related signaling pathway through the glucocorticoid receptor as a specificity control] as quantified by the established validated Transcription Element Listening System for downstream analysis (Supplemental Methods, Supplemental Digital Content, http://links.lww.com/SLA/F59).^26,27 These outcomes were characterized by mean percent fold-differences in TF binding motif (TFBM) prevalence, which represents the percent amount of increase or decrease in prevalence of RNA sequences that are bound by these TFs with more positive or more negative values meaning more common or less common than a comparison standard, respectively. Multivariable linear regression analysis quantified the relationship between (log2) transcript abundance and objective ND (ADI: 1–10) and subjective ND (NSEAS) controlling for age, race/ethnicity, smoking, alcohol, BMI, clinical stage, subtype, grade, neoadjuvant treatment, and assay batch.

Secondary Outcomes: Clinical Predictors and Outcomes

Logistic regression was used to evaluate the association between TFs and Oncotype DX score which was dichotomized based on clinical relevance: low-risk (0–25) and high-risk (26–100) from the TAILORx and RxPONDER clinical trials.²⁸ Multivariable Cox proportional hazard models evaluated the association between TFs and RFS controlling for age, race and ethnicity, smoking, alcohol, BMI, clinical stage, subtype, grade, treatment, and assay batch. We also evaluated the association between subjective and objective ND with TFs of BCa treatment response based on the literature (ESR1 and BCL-6), which were also represented as mean percentage fold differences in TFBM. Multiple hypothesis testing was corrected with the 5% false discovery rate for all TF data. All analyses were conducted in R (www.r-project.org/).

RESULTS

Sociodemographic, Lifestyle, Behavior, Tumor, Treatment, and Neighborhood Social Environment Adversity Survey Characteristics by Objective Neighborhood Disadvantage

The final analysis included 147 patients. The median ADI for the entire cohort was 4 (interquartile range: 2–6). Non-Hispanic Black had a higher proportion of patients living in high ND (ADI: 5–10) compared with all other racial and ethnic groups (P < 0.001). Patients in the high ND compared with the low ND group were more likely to have stage III disease (21.5% vs 7.3%, P = 0.03) and triple-negative breast cancer (29.2% vs 11.0%, P = 0.005; Table 1). The average number of years lived at their current address was 17.9 ± 14.2 for the total cohort, 15.6 ± 14.8 for high ADI, and 19.7 ± 14.0 for low ADI groups. In the total population, the mean (SD) composite NSEAS score (higher indicates more perceived ND) was −0.12 (4.80), with the subscale scores as follows: (1) threat to safety: −0.05 (1.54), (2) social cohesion: −0.06 (2.81), and (3) esthetic quality: 0.01 (1.73; Table 1).

TABLE 1.

Cohort Sociodemographic, Lifestyle, Behavioral, Tumor, Treatment, and NSEAS Characteristics by ND^*

	N (%)
	Total population (N = 147)	Below median ADI (1–4); N = 82 (55.8%)	Above median ADI (5–10); N = 65 (44.2%)	P
Age, mean (SD)	55 (12)	57 (11)	54 (14)	0.13
Race/ethnicity; n (%)				< 0.001
Asian	4 (2.7)	3 (3.7)	1 (1.5)
Hispanic	97 (66.0)	58 (70.7)	39 (60.0)
NHB	21 (14.3)	3 (3.7)	18 (27.7)
NHW	25 (17.0)	18 (22.0)	7 (10.8)
Tobacco use; n (%)				0.74
Never	113 (76.9)	63 (76.8)	50 (76.9)
Former	29 (19.7)	17 (20.7)	12 (18.5)
Active	5 (3.4)	2 (2.4)	3 (4.6)
Alcohol use; n (%)				0.08
Never	25 (17.0)	16 (19.5)	9 (13.8)
Former	82 (55.8)	39 (47.6)	43 (66.2)
Active	40 (27.2)	27 (32.9)	13 (20.0)
BMI (kg/m2), mean (SD)	29.2 (5.8)	28.7 (6.4)	29.7 (4.9)	0.30
Clinical stage; n (%)				0.03
I	41 (27.9)	27 (32.9)	14 (21.5)
II	86 (58.5)	49 (59.8)	37 (56.9)
III	20 (13.6)	6 (7.3)	14 (21.5)
Histology; n (%)				0.19
Invasive ductal	130 (88.4)	70 (85.4)	60 (92.3)
Invasive lobular	17 (11.6)	12 (14.6)	5 (7.7)
Subtype; n (%)				0.005
ER+/HER−	119 (81.0)	73 (89.0)	46 (70.8)
ER−/HER2−	28 (19.0)	9 (11.0)	19 (29.2)
Tumor grade; n (%)				0.42
I	22 (15.1)	13 (16.0)	9 (13.8)
II	74 (50.7)	44 (54.3)	30 (46.2)
III	50 (34.2)	24 (29.6)	26 (40.0)
Treatment; n (%)
Neoadjuvant chemotherapy	42 (28.6)	15 (18.3)	27 (41.5)	0.002
Adjuvant chemotherapy	62 (42.2)	28 (34.1)	34 (52.3)	0.020
Adjuvant radiation	82 (55.8)	40 (48.8)	42 (64.6)	0.034
Adjuvant endocrine therapy	114 (77.6)	71 (86.6)	43 (66.2)	0.004
NSEAS scores (N = 54)†
Composite Scores	−0.12 (4.80)	−0.46 (5.13)	0.47 (4.23)	0.52
Threat to safety domain scores	−0.05 (1.54)	−0.20 (1.49)	0.21 (1.62)	0.94
Social cohesion domain scores	−0.06 (2.81)	−0.11 (3.07)	0.03 (2.38)	0.22
Aesthetic quality domain scores	0.01 (1.73)	−0.14 (1.64)	0.28 (1.88)	0.32
Years lived at current address; mean (SD)	17.9 ± 14.2	19.7 ± 14.0	15.6 ± 14.8	0.09

10 being the most disadvantaged neighborhood and 1 being the least disadvantaged neighborhood, median ADI score for the cohort was used to dichotomize the population into high and low neighborhood disadvantage subgroups.

^*Neighborhood disadvantage is defined using the Area Deprivation Index as a continuous number from 1 to 10.

^†All NSEAS scores are represented as standardized Z-scores.

NHB indicates non-Hispanic Black; NHW, non-Hispanic White.

Objective Neighborhood Disadvantage is Associated With Aggressive Biology and Sympathetic Nervous System Activation

Analysis of covariate-adjusted objective ND identified that higher ADI (more ND) was associated with up-regulated pro-inflammatory TFs NF-kB and AP-1 (Fig. 2; NF-kB: 1.72 mean fold-difference TFBM prevalence among promoters of disadvantage-related transcripts, ± 0.26 fold-difference SE, P = 0.022; AP-1: 1.34 ± 0.11, P = 0.004) and with down-regulated protective TFs involved in cellular immune responses and programmed cell death (Fig. 2; IRF: 0.65 ± 0.12, P = 0.011, STAT: 0.75 ± 0.09, P = 0.0083).¹⁰ Higher ADI was also associated with up-regulated CREB TFs that transduce catecholamine signaling through β-adrenergic receptors from the SNS (2.32 ± 0.24, P = 0.0002). By contrast, results showed no indication of differential activity for the glucocorticoid receptor that mediates HPA-axis cortisol signaling (1.10 ± 0.11, P = 0.32). Higher ADI was also associated with down-regulated TFs linked to BCa treatment response [Fig. 2; ESR1: (0.43 ± 0.15, P = 0.006) and BCL-6: (0.63 ± 0.06, P < 0.0001)].

FIGURE 2.

Transcriptional factor regulation of breast tumors as a function of ND (N = 147). Data represent covariate-adjusted mean (SE) % difference in promoter TFBM prevalence for 201 gene transcripts up-regulated >2-fold over the range of continuous neighborhood ADI decile scores (compared with 220 genes equivalently down-regulated), for pathways involved in immune regulation (inflammation: NF-kB, AP-1; type I interferon response: IRF, STAT), stress-related signaling from the SNS (CREB) and the hypothalamus-pituitary-adrenal axis (GR), and BCa treatment response (ESR1, BCL-6). Promoter TFBM prevalence was quantified in RefSeq core promoter sequences using TRANSFAC position-specific weight matrices, quantified as a % difference from TFBM prevalence in down-regulated gene transcripts, and tested for statistically significant deviation from the null hypothesis ratio of 0% difference using a z test based on standard errors (SD of difference sampling distribution) from 200 cycles bootstrap resampling of gene expression data cases (accounting for any correlation among transcripts within individuals) as previously detailed.²⁷ *P < 0.05; **P < 0.01. GR indicates glucocorticoid receptor.

Subjective Neighborhood Disadvantage is Associated With Aggressive Biology and Sympathetic Nervous System Activation

Analysis of covariate-adjusted subjective ND identified that higher composite NSEAS scores and specifically higher threat to safety subscores (eg, “Violence is a problem in my neighborhood” and “My neighborhood is not safe from crime”) were associated with up-regulated NF-kB (1.39 ± 0.09, P = 0.002), down-regulated IRF (0.73 ± 0.09, P = 0.003), and up-regulated CREB (1.34 ± 0.11, P = 0.0058; Fig. 3).

FIGURE 3.

Transcriptional Factor Regulation of Breast Tumors as a Function of Subjective NSEAS Scores (N=54). Data represent covariate-adjusted mean (SE) % difference in promoter TFBM prevalence for NSEAS score-associated gene transcripts up-regulated in association with continuous (A) composite, (B) threat to safety, (C) social cohesion, and (D) aesthetic quality survey scores, for pathways involved immune regulation (inflammation: NF-kB, AP-1; type I interferon response: IRF, STAT), stress-related signaling from the SNS (CREB) and hypothalamus-pituitary-adrenal axis (GR). Higher scores are indicative of higher subjective ND. Promoter TFBM prevalence was quantified in RefSeq core promoter sequences using TRANSFAC position-specific weight matrices, quantified as a % difference from TFBM prevalence in down-regulated gene transcripts, and tested for statistically significant deviation from the null hypothesis ratio of 0% difference using a z-test based on standard errors (SD of difference sampling distribution) from 200 cycles bootstrap resampling of gene expression data cases (accounting for any correlation among transcripts within individuals) as previously detailed.²⁷ *P < 0.05; **P < 0.01. GR indicates glucocorticoid receptor.

Pro-inflammatory Transcription Factors Are Associated With High-risk Oncotype DX Scores

Sixty-two patients met the clinical criteria for sending Oncotype DX scores. Fifty-one (82.3%) had low-risk and 11 (17.7%) had high-risk scores. High-risk scores were significantly associated with up-regulated pro-inflammatory activity (NF-kB: 1.89 ± 0.15, P < 0.001).

Transcription Factors of Aggressive Biology and Sympathetic Nervous System Activation Are Associated With Shorter Recurrence-free Survival

A total of 16 patients had a recurrence or died. The most common event was a recurrence (n = 15). At 12 and 24 months, the cumulative RFS was 85% and 76%, respectively. After controlling for covariates, shorter RFS was associated with aggressive biology and SNS activation (Fig. 4; NF-kB: 2.18 ± 0.19, P < 0.001, IRF: 0.42 ± 0.18, P < 0.001, STAT: 0.74 ± 0.07, P < 0.001, CREB: 2.03 ± 0.20, P < 0.0001).

FIGURE 4.

TF control pathway activity is associated with shorter RFS (N = 147). Data represent covariate-adjusted mean %-difference in bioinformatically inferred transcription control pathway activity for genes over-expressed in breast tumors from patients with shorter RFS, for pathways involved immune regulation (inflammation: NF-kB; Type I interferon response: IRF, STAT), and adversity-related signaling from the SNS (CREB). Promoter TFBM prevalence was quantified in RefSeq core promoter sequences using TRANSFAC position-specific weight matrices, quantified as a % difference from TFBM prevalence in down-regulated gene transcripts, and tested for statistically significant deviation from the null hypothesis ratio of 0% difference using a z-test based on standard errors (SD of difference sampling distribution) from 200 cycles bootstrap resampling of gene expression data cases (accounting for any correlation among transcripts within individuals) as previously detailed.²⁷ *P < 0.05; **P < 0.01.

DISCUSSION

To our knowledge, this is the largest prospective social genomics cohort study in oncology patients. It is well-established that ND is associated with shorter RFS, independent of patient, tumor, and treatment factors, suggesting unaccounted mechanisms that may include more aggressive tumor biology. We, therefore, took a social genomics approach to evaluate the association between ND and aggressive biology. We discovered that living in higher ND (defined both objectively and subjectively) is independently associated with aggressive tumor biology, defined using validated TFs from the field of social genomics. These TF patterns (up-regulated pro-inflammatory TFs (NF-kB and AP-1) and down-regulated TFs for protective cellular responses and programmed cell death (IRF and STAT) reflect more aggressive BCa biology and are based on 25 years of preclinical and clinical social genomics research mapping the biological mechanisms through which social adversity risk factors influence host biology.^{9,11,12,14,19,27,29–31} We also identified up-regulation of CREB, which mediates the gene-regulatory effects of β-adrenergic signaling from the SNS.^32–34 This suggests the association between higher ND and aggressive tumor biology might be mediated by “fight or flight” SNS activity. These findings are highly consistent with findings from preclinical mouse models linking chronic stress-induced SNS activation to epithelial to mesenchymal transition, M2 polarization, and lymphangiogenesis in primary breast tumors and with findings from previous studies documenting increased SNS activity and pro-inflammatory gene activation in socially isolated humans.^19,33 Importantly, we discovered that these TF patterns correlate with more worse clinical features such as high-risk Oncotype DX scores and shorter RFS, validating their association with clinically relevant aggressive biology.

These novel findings add to prior social genomics literature by introducing neighborhood-level disadvantage as a source of social adversity. This is important as prior social genomics research has focused on individual-level social adversity such as chronic loneliness and social isolation,^9,12,15–18 social threat or instability,^35,36 and racial or sexual discrimination.^37,38 In addition to objective ND, we found that a patient’s subjective experience of their ND (NSEAS) was similarly associated with aggressive tumor biology. Interestingly, responses related to increased neighborhood crime and violence were most associated with up-regulated NF-kB, IRF, and CREB, suggesting that one’s perceptions of threat to safety may influence aggressive tumor biology and SNS activation. Collectively, these findings of up-regulated SNS activity (CREB), with both objective and subjective ND, suggest the SNS as a potential mechanism of how ND impacts aggressive tumor biology.¹⁴ Anticipation of threat (“fight or flight”) activates the SNS, and this response does not decay over time.¹⁴ This chronic, repetitive SNS activation, versus an acute stressor, is theorized to influence tumor biology and is corroborated by cancer neuroscience research.³⁹ Our patient population, on average, resided in their same neighborhood for 18 years, indicating that their exposure to both objective and subjective ND was a chronic rather than acute source of social adversity.¹⁴

The TFs also largely correlated with predictors of clinically relevant outcomes reflective of aggressive tumor biology. Specifically, Oncotype DX score > 26 was associated with up-regulated pro-inflammatory NF-kB. While Oncotype DX was not associated with other transcriptional control pathways such as CREB or IRF, this is expected since the other measures reflect SNS and cytotoxic T-cell activity, both of which Oncotype DX is not designed to assess. Oncotype DX assesses a tumor’s intrinsic aggressive biology, not the additional tumor-extrinsic characteristics that might also affect disease progression. Our findings therefore provide insight into clinically relevant tumor-extrinsic factors associated with more aggressive tumor biology. Moreover, a recent study using the National Cancer Database by Goel et al⁴⁰ found that patients with BCa living in higher ND had higher odds of a high-risk ( > 26) Oncotype DX score compared with those living in the most advantaged neighborhoods after accounting for sociodemographic and tumor characteristics. This supports our findings that higher objective ND is associated with high-risk clinical features, including Oncotype DX and shorter RFS.

The overarching importance of our findings lies in the potential to inform policy, advocate for healthy equity, and guide future behavioral and therapeutic interventions to improve BCa RFS among those living in ND. First, acknowledging that ND is entrenched in structural racism and must be addressed on multiple levels including: (1) addressing redlining practices that have historically led to disparities in neighborhood living conditions, access to resources, and environmental hazards; investing in community-based public safety models and addressing systemic biases in the legal system, (2) ensuring equitable medical care is delivered to marginalized communities, including bias training for medical professionals, equitable resource allocation, and community-based health initiatives, (3) educating the public and health care professionals about the impact of structural racism on health, (4) enacting laws and regulations that specifically address the root causes of these disparities, and (5) engaging with and empowering communities affected by racial health disparities.^41,42 While addressing the root causes of ND through changes in policy and the disassembly of structural disparities cannot be understated, behavioral and pharmaceutical interventions may also address modifiable risks to impact patient outcomes and improve equity in a timelier manner.⁴¹ One approach is through behavioral interventions such as stress reduction effective in reducing SNS activity for women with BCa living in ND. Antoni et al^43,44 have developed a cognitive behavioral stress management intervention that decreased blood leukocyte CTRA gene expression and improved BCa RFS in a randomized clinical trial. This intervention teaches a combination of relaxation and cognitive behavioral therapy techniques to address stress processes, including threat/anxiety, and reverses social adversity-associated aggressive biology and improves RFS.²⁰ Our NSEAS data, therefore, inform future cognitive behavioral stress management/cognitive behavioral therapy interventions to reduce SNS activity due to stress from threat to safety concerns.

Our findings also lay the foundation for future pharmacologic therapeutic interventions targeting the harmful TF pathways that may be up-regulated in women living in higher ND. Specifically, medications such as propranolol, a β-adrenergic antagonist, may reverse the downstream effects of ND on aggressive tumor biology since social genomic studies, including ours, have identified up-regulation of β-adrenergic SNS activity with aggressive tumor TF activity.^14,33 A randomized clinical trial identified that pretreatment of patients with propranolol²¹ and 2 small phase 2 randomized controlled biomarker studies identified favorable effects of propranolol on genome-wide transcriptional profiles of resected BCa.^45,46 Moreover, we identified down-regulated TFs associated with improved BCa treatment response (ESR1 and BCL-6) which may help risk-stratify treatment protocols for patients from ND. Taken together, these studies show promise in identifying targets for behavioral and pharmacologic therapeutics and treatment risk-stratification and prognostication that may reduce the impact of ND-associated adversity on aggressive biology and ultimately survival.

Our study is limited to 2 institutions from South Florida, and these findings need to be validated in a larger, national cohort. Because of our location, our population is predominantly Hispanic, which may also limit external population validity. Nevertheless, the data reflect local population-based data with respect to race and ethnicity and clinical stage distribution when compared with the Florida Cancer Data System BCa data (Supplemental Digital Content Table 2, http://links.lww.com/SLA/F59). Beyond the “majority-minority” racial and ethnic diversity of our cohort, our study also capitalizes on our institution’s urban location in Miami-Dade County—comprised of the only NCI-designated cancer center in South Florida and one of the nation’s largest safety-net hospitals—where we treat a socioeconomically diverse population.⁴⁷ Thus, our findings may be generalizable to urban, minority populations, despite differences in underlying cultural and/or other “lived experiences.”

Another strength of our study was our ability to control for granular-level confounders of objective and subjective ND, aggressive BCa tumor biology, and inflammatory pathways such as age, race and ethnicity, smoking, tobacco use, BMI, tumor stage and subtype, and treatment. Also, given the study focuses on our catchment area, we inherently controlled for air pollution given similar PM_2.5, PM₁₀, and Ozone levels across South Florida, which may serve as another potential confounder of the association between ND and aggressive tumor biology.^48,49 However, we recognize that additional individual-level factors associated with stress and adversity, that are unaccounted for in this study, may also influence the association between ND and tumor biology. Also, the NSEAS tool was originally studied in the context of atherosclerosis, and while conceptually this should not differ in other disease models, it requires further construct and content validation in larger cancer populations. Despite these limitations and the need for validation in a more nationally representative cohort, this study leverages our novel genomic-epidemiologic MBCDS cohort infrastructure to conduct this important social genomics research to discover an association between ND and aggressive BCa tumor biology in humans.

CONCLUSIONS

Overall, this prospective cohort study discovered that both objective and subjective ND were independently associated with TFs reflecting aggressive BCa tumor biology after controlling for sociodemographic, lifestyle, behavioral, tumor, and treatment characteristics. These TFs are also associated with high-risk Oncotype DX scores and shorter RFS, validating their clinical relevance. Moreover, our findings implicate social adversity-associated stress signaling from the SNS as a potential neurobiological mechanism of such effects. Taken together, our findings advance the current understanding of the social determinants of BCa outcomes by harnessing a social genomics approach to translational epidemiologic analyses. These findings need to be validated in a larger, national cohort to further establish modifiable features of ND that most adversely impact tumor biology and survival outcomes to develop targeted interventions that reduce BCa disparities in survival amongst women living in these environments.

Supplementary Material

Supplemental Digital Content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal’s website, www.annalsofsurgery.com.