Overview of the Microbiome Among Nurses study (Micro-N) as an example of prospective characterization of the microbiome within cohort studies.

Finally, the BIOM-Mass Data Portal ( http://portal.biom-mass.org ) is provided by the Harvard Chan Microbiome in Public Health Center (HCMPH) to manage and share microbiome profiles, sample, and population information from microbiome epidemiology studies carried out through the HCMPH BIOM-Mass platform, including Micro-N. It supports both open- and controlled-access dissemination of microbiome multi-omics (16S rRNA gene amplicon profiles, metagenomes, metatranscriptomes, metabolomes, etc.), raw and processed data products (sequences, taxonomic profiles, functional profiles, etc.), and sample and subject covariates (phenotypes, demographics, biometrics, technical protocols, etc.). Data can be shared publicly, controlled-access, or securely protected on a project-specific basis. Only microbial information (i.e. non-human genetic material) are shared, and sensitive covariates can be stripped, linked from an external database such as dbGaP, or secured for individual projects by Google Cloud Platform authentication. The Data Portal builds on technology from the Human Microbiome Project Data Coordinating Center ( http://ihmpdcc.org ) 1 and the Genomic Data Commons ( https://gdc.cancer.gov ) 28 and is integrated with the Terra platform ( https://terra.bio ) for ‘omics data dissemination. The BIOM-Mass Data Portal is particularly tailored to provide raw and processed microbiome epidemiological profiles and accompanying phenotypes and covariate annotations, including for large, controlled-access projects such as Micro-N.

After return via U.S. mail, kits are received and their components (questionnaires and specimens) tracked using the matched, affixed barcodes. This allows technical considerations such as kit incompleteness or damage to be recorded as well. For all received kits, standardized questionnaires are set aside to be coded, scanned, and programmed with existing algorithms validated in the parent NHS II study. Stool and oral specimens are loaded onto a laboratory information management system (LIMS) (LabVantage, Somerset, NJ)-integrated, Hamilton (Franklin, MA) Microlab STAR liquid handling robot ( Figure 2 ). This permits pre-separation into multiple ~500uL aliquots per sample prior to storage, preventing the need for freeze-thaw cycles prior to assays at various times in the future. By matching specimen, questionnaire, LIMS, and aliquot cryovial barcodes, batched aliquot racks are then automatically loaded into a Hamilton BiOS robotic freezer for long-term storage at −80°C. This storage environment permits automated retrieval of any aliquot subset based on sample type, collection date, subject characteristics, or other information recorded in the Micro-N LIMS or the parent NHS II study. Retrieval is executed and tracked automatically by the freezer, and successfully retrieved aliquots can be transferred directly for experimental work (e.g. for culturing or gnotobiotics), or shipped to molecular data generation facilities (for amplicon sequencing, metagenomics, metatranscriptomics, and/or metabolomics). Critically, this infrastructure permits long-term parent and child sample linkage with their contributing cohort participants and associated information. For example, should a population-level study result in the isolation of a microbial strain of interest, the isolate, its parent specimen, associated molecular data, and the medical history of its original donor are all available through the integrated biorepository, NHS II cohort infrastructure, and BIOM-Mass data portal.

The advent of inexpensive, widely available microbial community assays (particularly high-throughput, next-generation sequencing-based approaches) has underscored the extent to which prospectively banked human microbiome specimens would benefit existing long-running cohorts. With Micro-N, BIOM-Mass, the associated protocols, and the resources of the HCMPH, we aim to future-proof the NHS II and other large-scale epidemiologic studies for future developments in microbiome science, while also driving near-term discoveries in mechanism, causality, and public health. Although long-term (and especially early life) exposures are viewed as dominant driving forces for the microbiota’s substantial inter-individual variation, 29 – 31 short-term exposures also have critically important consequences (e.g. antibiotics), 32 , 33 and the specific effects of many such exposures on the microbiome remain to be established. Addressing these significant knowledge gaps requires prospective investigations with high-quality exposure data collected over the life course, with microbial composition and function assessed prior to disease onset. The Micro-N project’s large-scale microbiome specimen collection within the NHS II will provide unparalleled opportunities for future prospective studies to interrogate the role of the microbiome in human health and disease, as well as the interplay between environmental factors, genetics and other host characteristics, and the microbiome in disease development and progression. It has also served to establish a scalable, generalizable, and validated protocol for microbiome epidemiology, implemented by and available from the HCMCC, BIOM-Mass, and the HCMPH. Any such investigations can take advantage of the efficient sampling design of nested case-control studies, in which incident cases diagnosed after specimen collection and their matched controls are identified for microbiome assessments. These data can then be pooled to study the long-term influence on the microbiome of lifestyle, genetic and environmental exposures over the life course. While the NHS II is limited by its inclusion of female participants only, the generalizable protocol we are using allows for future pooled analysis of data from other cohorts and different study populations. 34 – 36 Moreover, although the current protocol covers sample collection at a single time-point, the established infrastructure in the NHS II cohort allows for future repeated collections from these participants. Finally, 27,706 children of the NHS II participants between the ages of 9 and 17 have been enrolled and followed up since 1996 in another prospective cohort, the Growing Up Today Study (GUTS), 37 offering opportunities for transgenerational studies. Therefore, we anticipate that the Micro-N project will provide the ideal setting to elucidate how host, environment, and the microbiome interact with each other to influence health, facilitating development of microbiota-targeted preventives, diagnostics, and therapeutics.

The Micro-N project is designed to integrate microbiome characterization into epidemiologic research within the NHS II. Specifically, it will enable us to prospectively characterize the determinants and health effects of the microbiome, as well as its interactions with environmental and genetic factors in disease development and progression. In Micro-N, stool and tongue swab samples are collected for gut and oral microbiome analysis from a targeted subset of 20,000 women under active follow-up ( Figure 2 ). To maximize the scientific yield, we have prioritized historically underrepresented women and women who have previously contributed other biospecimens (e.g., blood and urine) for genotyping and other profiling assays (e.g., metabolomics) to allow for systematic interrogation of host-microbiota interactions. The study was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health. In 2017, on the biennial questionnaire, we asked participants if they would be interested in providing a stool and oral sample if the collection was simple and relatively hygienic. Among the 83,695 participants who responded, 55,215 (66%) women answered yes, among whom 42,093 (50%) said they would definitely participate and the other 13,122 (16%) would possibly participate. These responses support the feasibility to recruit and collect microbiome samples from at least 20,000 participants. The feasibility of this large-scale collection is further supported by two previous studies, the Men’s Lifestyle Validation Study (MLVS) 9 , 10 and the Mind Body Study (MBS). 11 The MLVS included 308 men from a parallel cohort, the Health Professionals Follow-up Study (HPFS), who provided up to four stool samples (two per week separated by 6 months) with additional data collected from two seven-day diet records, two physical activity monitors, and two FFQs. The MBS included 233 women from the NHS II who completed a detailed psychosocial assessment and were asked to self-collect stool samples from two consecutive bowel movements (1−3 days apart) at two time points six months apart. In the MBS, 213 (91%) women returned the first set of kits and 206 (88%) returned the second set. Among women from the main NHS II cohort who expressed their willingness to participate, we conducted, via email, an enrollment survey to collect updated mailing and email addresses and to obtain electronic consent ( Figure 2 ; Supplementary Methods 1 ). We also asked participants whether they had had any overnight hospital visits or lower endoscopic evaluations within the last 2 months, because of the known influence of hospitalization 12 and bowel preparations 13 , 14 on the gut microbiota. Participants who reported a hospitalization or colonoscopy in the past 2 months were held in the collection queue and approached for collection 2 months later. The collection kits with prepaid return postage were shipped by the U.S. Postal Service (kit components are summarized in the Supplementary Table ). We launched the collection in February 2019 with the support of an infrastructure grant from the Massachusetts Life Sciences Center (MLSC). Our collection began with a phased, three-month ramp-up period in which we monitored both the rate of enrollment and consent as well as the rate of kit return. In this period, kits were sent to 2321 women, and 88% returned a kit within 2 months. As of January 22, 2021, we have sent kits to 17,464 women who had consented to participate and received the kits back from 14,731 women. We currently project to complete the collection from 20,000 women by mid-August 2021. Based on the age distribution and disease rate in the NHS II, we projected the number of incident cases of selected disease outcomes in 5, 10 and 15 years after the microbiome specimen collection in Micro-N ( Table 2 ). Leveraging the ongoing follow-up of the NHS II cohort, these estimations highlight the potential of Micro-N for future prospective studies elucidating the role of the microbiome in disease incidence.

The specimen and metadata collection protocols developed for Micro-N have been formalized under the Harvard Chan Microbiome Collection Core (HCMCC) component of BIOM-Mass, a platform that we developed, which also provides customizable fee-for-service implementations of these kits and processes for other large or small studies. As such, we sought to ensure that both fecal and oral specimen collections were reliable, flexible, and scalable, and that a minimum set of standardized participant information accompanies each collection. Although immediate freezing of stool samples at −20°C or below is often preferable for microbiome preservation, it is not feasible for such large-scale field studies, leading to the development of alternative preservation methods for self-collected samples. Prior studies from our group and others have tested several different preservation methods, including RNAlater (Thermo Fisher, Waltham, MA), 90+% ethanol, 70% ethanol, OMNIgene•GUT (DNA Genotek Inc., Ottawa, Ontario, Canada), Zymo DNA/RNA Shield (Zymo Research, Irvine, CA), fecal occult blood test (FOBT) cards, fecal immunochemical test (FIT) tubes, and Whatman®FTA cards (GE Healthcare, Chicago, IL). 15 – 21 The predominant findings of these studies indicate that, for the subset of preservative methods meeting a minimum threshold for microbial fixation (i.e., the cessation of substantial metabolic activity), the remaining effects of different preservatives and temperature regimes on the microbial community profiles are small compared to inter-individual variability. 15 – 22 Through the use of preservatives, most of the collection methods are able to prevent major compositional changes in fecal microbial community when exposed to temperature fluctuations from 4°C −40°C over as many as 8 weeks, making them appropriate for home collection and a variety of shipment conditions. 15 – 22 For Micro-N, stool and oral samples are thus self-collected by participants using the provided kits. Table 3 summarizes the applicable downstream assays and stability evidence for the collection kits used in Micro-N. For stool collection, to accommodate substantial biomass for long-term biobanking and a variety of different downstream assays, we provide three different sample tubes. One tube includes 95% ethanol, the second is the commercial OMNIgene•GUT kit, and third is a cryovial pre-filled anaerobically with liquid dental transport medium (LDTM) (Anaerobe Systems, CA, USA). Together, these preservative options enable amplicon, shotgun metagenomic and metatranscriptomic sequencing, stool metabolomic profiling, in addition to future culture and gnotobiotic animal model studies. Participants are asked to collect samples from the same bowel movement for all 3 tubes ( Supplementary Methods 2 ). Briefly, once the toilet accessory has been affixed to the commode, participants are asked to use the included disposable spatula to transfer a small amount of stool into each collection tube up to a clearly marked, specified target level. After collection, participants are asked to shake the three tightly sealed tubes for 30 seconds to allow for adequate mixture of the samples with the stabilizing liquids. Upon finishing this collection, participants complete a brief stool sample questionnaire ( Supplementary Methods 3 , and see next section for further details) and affix barcode labels to the questionnaire and each tube. We have validated our ethanol-based self-collection protocol in the context of fecal metagenome and metatranscriptome profiling in a pilot study of men enrolled in the HPFS. 23 Participants self-collected stool at home using both the Human Microbiome Project 1-validated protocol (fresh frozen) and our collection kits using the U.S. postal service for sample return. Consistent with 16S rRNA-based studies, 15 – 20 we demonstrated that self-collected stool using the preservatives in our sample tubes (one tube with 95% molecular biology grade ethanol and one OMNIgene•GUT tube) provided statistically near-identical metagenomic data to frozen samples. 21 The second iteration of the Human Microbiome Project, the Integrative Human Microbiome Project (iHMP), and specifically the Inflammatory Bowel Disease (IBD) cohort within iHMP, went on to use a similar protocol for self-collected stool, additionally allowing for untargeted metabolomic profiling in ethanol-preserved aliquots. 24 For the accompanying oral sample collection, we use tongue swabs from the OMNIgene•ORAL kit (DNA Genotek Inc., Ottawa, Ontario, Canada). The tongue microbiome can have somewhat greater microbial-to-human nucleotide ratio variability between individuals but has decreased measurement variability than other self-collectable oral locations (e.g. buccal swab, saliva). 25 Participants are asked to provide a tongue swab sample immediately upon waking on the day following the stool collection and avoid eating, drinking, smoking, using mouthwash, or brushing teeth prior to sample collection ( Supplementary Methods 2 ). Participants use the swab to gently rub the tongue for a minimum of 30 seconds, immediately insert the swab into the bottom of the tube, and then snap the shaft off at the break point, while leaving the swab tip in the tube of liquid. To avoid contamination, participants are instructed not to touch the swab tip to any other surface. Once the tongue swab collection is finished, participants complete an oral sample questionnaire ( Supplementary Methods 3 , and see next section for further details) and mail all stool and oral collection kits and questionnaires back to our laboratory using the pre-paid shipping box. Micro-N maintained a dedicated phone helpline and e-mail address for participants who had specific questions, needed to ask for replacement components, or were concerned they were not eligible for participation because of a recent hospital procedure, disease diagnosis, or change in medication use. Among ~12,000 participants who were consented and returned the collection kits by March 2020, we estimate that we received contact queries from approximately 1,200 (10%) women. While Micro-N collection kits include four types of collection tubes, along with redundant copies of consumables (toilet accessories and spatulas), the HCMCC BIOM-Mass’s generalized protocol allows cost-effective, modularized configuration of these kit components to enable study customization.

Over the past decade, population-scale human microbiome studies have provided tremendous evidence linking microorganisms in the gut and body-wide with various conditions from gastroenterological and periodontal illnesses to cardiovascular, neoplastic, respiratory, and neurologic disorders. 1 However, the majority of human data are limited to cross-sectional studies, making it inherently challenging to differentiate cause from effect and are prone to reverse causation. Therefore, it is crucial to establish a biobank of samples collected prior to the onset of disease from well-characterized young and middle-aged populations with long-term longitudinal follow-up. 2 Leveraging ongoing large cohort studies provides both a relatively economical option and an ideal setting for the development of such microbiome biobanks, owing to the established infrastructure for biospecimen collection and outcome ascertainment, deep characterization of relevant risk factors, and complementary archival biospecimens that enable future linkage with the microbiome data. Developing and deploying such platforms for microbiome epidemiology requires the establishment of reliable sample collection methods and associated metadata generation instruments. 3 , 4 Here, we have accomplished this in the context of the Nurses’ Health Study II (NHS II), a leading epidemiologic cohort for studying risk factors and underlying mechanisms for chronic diseases among women. 5 To extend our knowledge about the role of the microbiome in health, we have recently launched a large-scale prospective collection of fecal and oral microbiome samples from 20,000 women in the NHS II, known as the Microbiome among Nurses (Micro-N) project. In this Perspective, we provide an overview of the Micro-N project and the methods it adopts to leverage generalizable population-scale microbiome collection protocols, which have subsequently been incorporated into the Harvard Chan Microbiome in Public Health (HCMPH) Center’s BIOM-Mass (Biobank for Microbiome research in Massachusetts) platform. This platform provides generalizable, scalable, and cost-effective methods for the conduct of prospective studies of the microbiome. We focus on the considerations for several key elements in microbiome specimen collection, including collection kit design, development of a comprehensive questionnaire for assessing potential major determinants of the microbiome, as well as sample shipment, handling, and storage protocols.

To provide a minimum set of essential proximal exposure, outcome, covariate, biometric, and technical information accompanying each sample, we developed a set of standardized questionnaires (for stool and oral collections, respectively) to accompany specimen collection kits. The scannable (Scantron, Omaha, NE) questionnaires were developed through collaboration with investigators from multiple institutions across the U.S. that participated in a joint microbiome working group for Micro-N (see Acknowledgements). To avoid over-burdening participants and facilitate implementation in similarly time-limited settings, we limited questionnaire length to two pages for stool collection and one page for oral swab collection. We employed an iterative process for questionnaire development. First, we identified major domains for assessment through literature review of microbiome determinants, complemented by reference to existing questionnaires that were shared within the working group. Seven domains were identified for the stool questionnaire, including: the timing of collection, stool consistency, bowel movement pattern, diet, major lifestyle factors, medication use, and medical history. For the oral questionnaire, five domains were identified, including: the timing of collection, use of the oral hygiene products, natural teeth, history of dental cleaning and surgery, and periodontal disease history. Then, for each of the domains we developed one or more potential questions that were subsequently discussed among the group to refine wording and response options. These discussions also identified the time frames most relevant to the microbiome for each question. After completing the draft questionnaires, we presented them to the working group and made further modifications based on feedback. In the end, a total of 12 overarching questions were included in each of the stool and oral questionnaires (some contained relevant sub-questions, e.g. diet and medications; the questionnaires are presented in Supplementary Methods 3 ; the rationales and considerations for questionnaire development are summarized in Supplementary Discussion ). Of note, the questions did not include those that have already been queried in the main follow-up questionnaires in the NHS II or do not change over time. Major categories of these previously-ascertained factors are early-life driving forces for the establishment of the microbiome, including: mode of delivery, breast feeding, and household exposures (e.g., siblings and pets) 26 , 27 , which have been included on the generalized forms of these questionnaires supported by HCMCC for other population studies.

Supplementary Methods 1. Enrollment survey. Supplementary Table. Kit components included in the collection in Micro-N. Supplementary Methods 2. Stool and oral sample collection instructions for Micro-N. Supplementary Methods 3. Micro-N questionnaires for stool and oral microbiome specimen collection. Supplementary Discussion. Major rationales and considerations for each of the questions on the stool and oral microbiome questionnaires