Fecal sample biobanking for breast cancer research focused on the gut microbiome

Abstract

Background:

Gut microbiome is an emerging potentially modifiable contributor to breast health, including breast cancer (BCa). To advance prevention research in this area, we established a prospective biobanking cohort of cancer-free women.

Methods:

Eligible women were ≥40 years old, had no cancer history and no recent antibiotic use. Women were enrolled during screening mammography visits at three imaging centers in Florida (February 2021-June 2024), completed a BCa risk factor survey, and underwent body measurements. We collected digital mammograms and stool/urine/saliva samples. Optionally, women completed NIH’s Diet History Questionnaire and a neighborhood stress questionnaire. Mammographic breast density (MBD) was assessed using established computerized approaches.

Results:

We recruited 733 cancer-free women (49% Caucasian, 21% African American, 26% Hispanic, and 4% from mixed/other races). The average age was 60 years (range 40–92); the majority (68.3%) were postmenopausal. BCa risk factor, neighborhood stress and diet questionnaires were completed by 97%, 65% and 58% of participants, respectively. Urine, saliva, and mammograms were available for all women; 83% also returned stool samples.

Conclusions:

We have established a representative cohort of screen-aged women with comprehensive BCa risk factor data, biospecimen collection, and MBD.

Impact:

This unique resource provides opportunity for future gut microbiome-focused BCa prevention research.

Introduction

Breast cancer (BCa) remains by far the leading cancer diagnosis and the second leading cause of cancer mortality. While BCa treatments have improved in recent years [1] early detection and prevention remain key to reducing BCa incidence and the burden of disease. Several recent studies suggest that gut microbiome may represent an emerging potentially modifiable contributor to a woman’s risk for breast cancer.

The populations of microbes that inhabit human bodies have complex relationships with their hosts. Microbes inhabiting the gut mediate key metabolic, physiological and immune functions [2–5] and perturbations in these functions influence health and disease [4–10].

The intestinal microbiome contributes to important protective, metabolic, and structural functions relevant to breast cancer, including host immunity; metabolism of a wide range of compounds (steroid hormones, phytochemicals, nitrates, and xenobiotics); integrity of the epithelial barrier and absorption; and host energy balance and nutritional status [11–14]. As a result, the dynamic population of intestinal microbes can influence metabolism and absorption of compounds with carcinogenic properties, modify levels of endogenous factors implicated in breast carcinogenesis (for example, steroid hormones), and alter inflammation and host susceptibility to oncogenic factors [15]. In particular, variations in the gut microbiome’s genes that are capable of metabolizing estrogens could affect risk of estrogen-driven breast cancer and offer future gut microbiome-directed targeted interventions to reduce breast cancer risk [16–18].

In spite of compelling theoretical grounds for a role of the gut microbiome in breast cancer, epidemiological evidence is limited due in part to the exceptional challenges of studying this exposure in observational research. In retrospective case-control studies, gut microbiome profiles in cancer cases may reflect changes in microbial composition that occur as a result of the cancer diagnosis or treatment, with potential for reverse causality [19]. In prospective cohort studies, investigations of the gut microbiome are impractical in the near term due to the large numbers of subjects that must be enrolled, long follow-up, and repeat sampling that may be required to capture change in gut microbiome profiles over many years of follow up. An alternative strategy to avoid these challenges is to study relationships of the gut microbiome with two of the leading, well-documented risk factors for breast cancer, i.e. mammographic breast density and endogenous hormone profiles. Importantly, use of well-known intermediate phenotypes has been recognized as an efficient alternative for studying chronic diseases that could also provide valuable information for primary cancer prevention [20–24].

To advance knowledge in this understudied area, we established an infrastructure that allowed us to prospectively recruit cancer-free women undergoing routine mammographic screening for breast cancer in the State of Florida with collection of urine, stool, and saliva samples and comprehensive data on critical breast cancer risk factors. The Florida population is among the most diverse in the nation and includes a large segment of traditionally underrepresented persons in research, including those of African American/Hispanic race, and lower socioeconomic status with poor health care, offering unprecedented opportunity for expansion of research into these populations. This unique infrastructure represents an outstanding resource for research on the gut microbiome and implementation of novel integrative molecular epidemiology approaches to the studyof breast cancer etiology and disparities.

Materials and Methods

Study population and recruitment

Recruitment took place between February 2021 and June 2024. Participants were enrolled at imaging centers affiliated with three Florida-based institutions: University of Florida (UF) in Jacksonville; Moffitt Cancer Center (MCC) in Tampa; and the University of Miami (UM). Collaborating centers are located in north, central and south Florida, respectively, and offer broad coverage of the state’s nearly 23 million residents. UF in Jacksonville serves the local community that comprises the state’s largest population of African American women. UM Sylvester Comprehensive Cancer Center, a NIH-designated cancer center serves the greater Miami area, along with south and central America, and is home to the state’s largest Hispanic community. MCC, a NIH-designated Comprehensive Cancer Center, serves a 15-county region on Tampa Bay in west central Florida comprised of more than 6 million persons (29% of the state’s population).

At MCC, participants were identified through the IRB-approved protocol that provides infrastructure for studies of lifestyle and genetic determinants of breast health measures in screening-aged women. Ethnic composition of MCC clinic patients mirrors that of the underlying population in Moffitt’s catchment area (~70% Caucasian, 15% Hispanic, and 15% African American women). At UF Jacksonville, participants were identified through the UF Jacksonville Imaging Centers. The Centers are accredited by the American College of Radiology and provide state-of-the-art screening and diagnostic mammography. Jacksonville is the largest (excluding the large metro areas of Miami and Tampa) and fastest-growing city in the state of Florida. Located in Duval County, Jacksonville has the state’s largest African American population. Ethnic composition of Jacksonville clinic patients undergoing screening mammography is similar to the distribution in the catchment area (50% African American, 40% Caucasian, and the remaining ~10% a mix of Asian, Native American, and other races). At UM, participants were identified through the Posner Breast Imaging Center at the NCI-designated Sylvester Comprehensive Cancer Center (SCCC). Miami is home to a diverse population with high concentrations of native and immigrant White and Black Hispanics from the Caribbean and Latin America. Racial composition of Miami clinic patients undergoing screening mammography is similar to the distribution in the catchment area (~20% Caucasian, ~70% Hispanic; and ~10% African American).

Women were recruited by study coordinators dedicated to this project in clinics during scheduled mammography appointments. IRB authorization for the project was maintained separately at each of the 3 project sites. Written informed consent was obtained from all study participants. Because of the infrastructure building nature of the project and implied access of resources to the research community throughout Florida and beyond, the subjects were made aware that 1) unnamed researchers would have access to their data and biospecimens in the future, 2) resources will support testing of currently unspecified hypotheses, 3) resources will be maintained indefinitely, 4) future research may involve genetic testing of samples and 4) no test results will be made available to participants, in addition to statements on the standard risk and benefits that apply in observational research studies with scope of the planned resource-building project. A $20 gift card was offered to women in appreciation of their time and effort upon completion of the in-clinic portion of the study. An additional $20 gift card was sent to the participant upon receipt of the stool sample.

Women were eligible for this study if they had no cancer history (other than non-melanoma skin). Further exclusions included: i) any oral/IV antibiotics within 30 days and/or, ii) more than two separate antibiotic regimens within the previous 3 months.

Data collection

Women were approached by study coordinators in study clinics at MCC, UF and UM, given an explanation of the study, and offered an opportunity to enroll. In addition, at UF we distributed information about the study via other primary care prevention offices affiliated with the UF Jacksonville and women interested in participation were then able to set up an appointment with the study coordinator for their participation. At UM, women had the option of completing study procedures in clinic or in their homes during scheduled in-home visits by study coordinators. Eligible women provided consent and completed an electronic survey covering demographics, established breast cancer risk factors (comprehensive reproductive history, including menopausal hormone use, detailed family history of breast cancer, alcohol consumption, and smoking); menopausal symptoms; medications (including recent use of antibiotics); quality of life and other psychosocial factors; and breast screening and medical history). A link to the electronic survey was emailed to women wishing to complete the survey at home. Additionally, women were offered with an opportunity to complete two questionnaires capturing diet and residential history. Information on usual diet in the preceding 12 months was collected via the Diet History Questionnaire food frequency questionnaire (DHQII) developed and administered through the National Cancer Institute that assesses diet and supplement use including probiotics during the previous year and produces correlations for energy, macronutrients, and several vitamins and minerals, consistent with other widely used FFQs [25]. DHQII has been successfully validated in previous studies [25,26]. Women agreeing to complete diet assessment were sent a link to the DHQII that includes an embedded unique sequence of digits that allows each subject’s data to be identified. Diet data is processed with Diet*Calc software developed by the NCI to generate specific nutrient and food group intake estimates. The data is maintained by the NCI and will be available for download for diet-related investigations. The Neighborhood Social Environment Adversity Survey captured information on neighborhood stress [27].

Standard body composition measurements were taken during the clinic or at-home (for UM) visit. For each participant, we recorded height and weight, circumference of the waist and hips (to calculate a waist-to-hip ratio), and chest circumference. Vital signs (pulse and blood pressure) were also recorded for approximately half of the participants. All consents and data were collected and managed using centralized RedCap data management system developed and maintained by the Participant Research, Interventions, and Measurement Core at MCC.

Sample collection

A clean-catch spot urine specimen was collected at recruitment in study clinics. Samples were aliquoted and stored at −80°C according to standard operating procedures at each institution, 90% within approximately one hour of collection. Times of urine collection and freezer storage were recorded for all urine samples. The samples from UF and UM were periodically shipped to MCC for centralized banking. Salivary samples were collected via ‘Oragene’ kits as a source of germline DNA and shipped to MCC for long-term banking.

Subjects were provided with a stool collection kit with detailed instructions, a stool collection questionnaire, and postage paid mailing kits for return of the materials according to methods that are used successfully in ongoing funded research projects and pilot investigations [28,29]. Stool samples and stool collection questionnaires were returned by the participants directly to MCC where they were immediately stored in −80°C freezer. Times of sample collection and receipt by the Tissue Core at MCC were recorded for all stool samples. The median time between stool sample collection and receipt by MCC was 3 days (range 0–6) for samples returned via FedEX shipment; in a subset of women (74), most enrolled at UM or UF, shipment was delayed by a week or more after collection of the sample for a variety of circumstances

Women also provided consent for release of pathology results and tissue samples in the event of breast biopsy. Formalin-fixed histologic sections will be requested from pathology labs and stored at MCC as they become available.

Mammogram collection and mammographic breast density assessment

Full field digital mammograms were obtained for all women. All three clinics have Hologic Selenia 3-dimensional (3D) Digital Breast Tomosynthesis units. UF-Jax and UM transferred zipped (compressed) digital image data to MCC using a secure cloud content management system. Upon receipt, DICOM images from all sites were inspected for quality and stored on a dedicated secure server at MCC. Anonymization procedures were performed to remove PHI from stored image data. As the primary measure, percent breast density was determined using methods developed by Heine et al. [30–33] that compute percent of the breast occupied by the dense (epithelial/stromal) tissue relative to the total breast area [30,34]. PD is regarded as the de facto standard for MBD determination [35] as it has been shown repeatedly to produce a validated measure of risk regardless of the mammographic method [36–39]. Because breast density of the right and left breast for a woman are strongly correlated [40], the average density of both breasts was used as the final mammographic breast density measure for each woman.

Data Sharing

The authors recognize and support the principles of resource sharing that have been endorsed by the NIH. Access to data and biospecimens (urine, residual stool sample and germline DNA) from the study described in this report is extremely valuable for future research. Therefore, investigators are encouraged to apply for use of these resources. Access decisions will be based on the following general principles: 1) Scientific merit, 2) Overlap with other approved protocols, and 3) Submission of a research plan appropriate to answer the study question(s). Review of applications and final decisions will be made by a steering committee comprised of the overall project PIs (Drs. Egan, Yaghjyan, and Goel) and other members of the research team. The committee will consider scientific merit, ethical and legal factors, and any potential conflicts of interest on the part of the applicant, with fair and transparent deliberation. Upon acceptance, de-identified samples and/or data will be released with Committee approval, and after establishing a Data Sharing or Material Transfer Agreement between institutions. The investigators have not budgeted costs for distribution of resources to end users. Hence, costs for data downloads or retrieval/ shipment of samples will be borne by researchers with approved protocols making the request. Further information can be obtained from the corresponding author.

Data availability

The authors recognize and support the principles of resource sharing that have been endorsed by the NIH. Access to data and biospecimens (urine, residual stool sample and germline DNA) from the study described in this report is extremely valuable for future research. Therefore, investigators are encouraged to apply for use of these resources. Access decisions will be based on the following general principles: 1) Scientific merit, 2) Overlap with other approved protocols, and 3) Submission of a research plan appropriate to answer the study question(s). Review of applications and final decisions will be made by a steering committee comprised of the overall project PIs (Drs. Egan, Yaghjyan, and Goel) and other members of the research team. The committee will consider scientific merit, ethical and legal factors, and any potential conflicts of interest on the part of the applicant, with fair and transparent deliberation. Upon acceptance, de-identified samples and/or data will be released with Committee approval, and after establishing a Data Sharing or Material Transfer Agreement between institutions. The investigators have not budgeted costs for distribution of resources to end users. Hence, costs for data downloads or retrieval/ shipment of samples will be borne by researchers with approved protocols making the request. Further information can be obtained from the corresponding author.

Results

By study close, we prospectively recruited 733 women in our biobanking study (348 from MCC, 153 from UF, and 232 from UM). The average age of women was 60 (range 40–92), with majority of the women (68.3%) being postmenopausal. This diverse sample included 49.2% Caucasian, 21.0% African American, 25.6% Hispanic, and 4.1% women from mixed or other races. Risk factor data collection was complete for 96.9% of women, with 57.8% and 64.8% of women also completing the optional diet and residential history questionnaire, respectively. Urine and saliva samples as well as mammograms were available for all women and stool samples along with the accompanying questionnaire were returned by 83% of participants.

Distribution of BI-RADS breast density categories from mammography reports in our sample (15.3% for BI-RADS A [almost entirely fatty], 46.7% for BI-RADS B [scattered areas of fibroglandular density], 34.4% for BI-RADS C [heterogeneously dense], and 3.7% for BI-RADS D [extremely dense]). Table 1 describes distribution of some of the breast cancer risk factors in our study population, overall, by study site, and by race/ethnicity.

Table 1.

Characteristics of the study population, by study site and race/ethnicity¹

Characteristic	Overall	MCC	UF	UM	White	African American	Hispanic	Other
Mean
Age (years)	59.68	61.00	57.67	59.04	61.68	59.42	56.92	54.34
Age at menarche (years)	12.41	12.48	12.28	12.39	12.52	12.34	12.27	12.30
Body Mass Index (kg/m2)	29.27	28.27	31.79	29.09	28.21	32.70	28.95	26.36
Age at menopause (years)	47.33	47.31	44.75	49.00	48.22	43.82	48.18	49.58
Alcohol, drinks/week	2.69	2.96	2.80	2.22	3.32	1.34	2.31	4.58
Percentages
Parity/age at first birth
Nulliparous	23.02	24.04	19.18	24.00	25.43	24.16	18.23	17.86
Parous, age<25 years	34.46	29.97	52.05	29.33	26.86	56.38	33.70	17.86
Parous, age≥25 years	42.51	45.99	28.08	46.67	47.71	19.46	48.07	64.29
Family history of breast cancer (yes)	23.52	30.06	19.59	16.37	28.65	22.15	16.30	14.29
History of benign breast disease (yes)	40.99	52.08	30.97	31.08	47.28	28.19	40.22	35.71
Smoking
Never smoked	67.84	70.15	59.46	69.91	62.93	77.18	67.39	82.14
Past smoker	25.25	26.87	24.32	23.45	31.32	14.77	24.46	10.71
Current smoker	6.91	2.99	16.22	6.64	5.75	8.05	8.15	7.14
Menopausal status/MHT use
Premenopausal	31.73	28.19	33.11	36.16	26.22	29.14	40.44	57.14
Postmenopausal/never used MHT	46.12	42.73	50.00	48.66	41.79	56.29	46.99	39.29
Postmenopausal/past MHT	12.98	18.40	6.08	9.38	18.16	8.61	8.20	3.57
Postmenopausal/current MHT	9.17	10.68	10.81	5.80	13.83	5.96	4.37	0.00
BI-RADS breast density
Almost entirely fat (A)	15.27	11.49	14.97	21.12	13.61	20.53	16.04	3.45
Scattered areas of fibroglandular density (B)	46.63	43.10	50.34	49.57	45.00	56.29	44.39	31.03
Heterogeneously dense (C)	34.39	41.67	32.65	24.57	37.78	23.17	32.62	62.07
Extremely dense (D)	3.71	3.74	2.04	4.74	3.61	0.00	6.95	3.45

Abbreviations: MCC – Moffitt Cancer Center; UF – University of Florida; UM – University of Miami

¹Among 733 participants, the results reflect the numbers based on completed questionnaire data

Discussion

As described above, over an approximately 3-year period we successfully enrolled 733 women with breast cancer risk factor data, urine, stool, and saliva samples, and mammogram collection. Of note, funding of the project coincided with emergence of the Covid19 pandemic such that recruitment was low during the initial project year and gradually increased to the originally projected levels of 8–10 participants per week as clinic operations returned to normal. Enrollment continued at the three project sites through June 2024 with a final sample size of 733 participants. This valuable resource offers unique opportunities supporting the study of the gut microbiome and breast cancer etiology.

Our imaging center-based recruitment efforts have generated a sample whose demographics mirror those of the state of Florida. We have been particularly successful in recruiting African American and Hispanic women who have been traditionally less likely to participate in research. We recognize, however, several challenges in covering broader recruitment areas and engaging additional imaging centers that could enable us to increase the cohort size. Thus, we are now working with additional facilities in the state to identify appropriate supplemental recruitment venues to further increase representation of African American and other underrepresented women in the study.

There are several advantages to our recruitment and data/sample collection procedures. The study consent, and most of the data collection, and urine/saliva collection were completed in the clinic thus allowing us to achieve collection of urine samples from all enrolled women as well as risk factor survey data from the great majority (97%) of participants. Collection of saliva for future genotyping studies is less invasive and time consuming than blood sample collection, thus improving participation. While stool sample collection is challenging and requires more commitment on the part of the participant, we were able to achieve a relatively high rate of stool sample collection (~83%). Incorporation of $20 incentive payment for stool return likely encouraged participation.

Recruitment processes aimed to minimize each woman’s time commitment while maximizing the amount of collected information. Women who did not have sufficient time to complete survey in the clinic, were offered an opportunity to do so at home or were followed up by the study coordinators who offered coordinator-assisted survey completion by phone, if preferred. We believe these approaches allowed us to maximize the amount of collected information and ensured high rates of completion.

In the initial months of sample collection using a USPS sample return delivery protocol, shipment times lagged especially for UF and UM. Therefore, we adopted FedEx-assisted delivery which reduced return times by several days on average. Previous studies show reasonable levels of stability of the gut microbiome under field conditions even with large variation in timing of collection and fluctuation of temperatures prior to freezer storage [41]. Urine samples were aliquoted and stored at −80°C at each institution within variable amount of time (on average 55 minutes for MCC, 49 minutes for UF, and 2 hours for UM). Our previous pilot investigation with a similar recruitment protocol for a study of gut microbiome and estrogens has demonstrated stable concentrations of estrogen metabolites with varying collection-to-storage times.\

Conclusions

We established a cohort of screening-aged women with collection of breast cancer risk factor data, urine, stool, and saliva samples, and mammographic breast density in the State of Florida. By using this unique resource, investigators have the opportunity to explore lifestyle factors as well as circulating biomarkers in relation to various health end-points in women and to examine contributions of gut microbiome to breast health via related intermediate phenotypes, such as circulating estrogens and mammographic breast density. Investigators encourage collaboration with cancer researchers in Florida and beyond.

Funding Statement

This project was funded by the Florida Department of Health Bankhead-Coley Cancer Research Program Infrastructure Grant.