Association between maternal breastmilk microbiota composition and rotavirus vaccine response in African, Asian, and European infants: a prospective cohort study

Background. Maternal breastmilk is a source of pre- and pro-biotics that impact neonatal gut microbiota colonisation. Since oral rotavirus vaccines (ORVs) are administered at a time when infants are often breastfed, breastmilk microbiota composition may have a direct or indirect influence on vaccine take and immunogenicity. Methods. Using standardised methods across sites, we compared breastmilk microbiota composition in relation to geographic location and ORV response in cohorts prospectively followed up from birth to 18 weeks of age in India (n = 307), Malawi (n = 119), and the UK (n = 60). Results. Breastmilk microbiota diversity was higher in India and Malawi than the UK across three longitudinal samples spanning weeks of life 1 to 13. Dominant taxa such as Streptococcus and Staphylococcus were consistent across cohorts; however, significant geographic differences were observed in the prevalence and abundance of common and rare genera throughout follow-up. No significant associations were identified between breastmilk microbiota composition and ORV outcomes including seroconversion, post-dose 1 vaccine shedding, and/or post-vaccination rotavirus-specific IgA level. Conclusions. Our findings suggest that breastmilk microbiota composition may not be a key factor in shaping trends in ORV response within or between countries.

CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint Given the additional amplification involved in library preparation for breastmilk samples, reads 154 were frequently detected in extraction controls (n = 56 individual or pooled controls with >10,000 155 reads after the filtering steps above). Several additional filtering steps were therefore included. First, 156 we retained RSVs if they were detectable at •0.1% abundance in •1% of breastmilk samples from at 157 least one country. Second, we applied prevalence-based filtering using the decontam package with a 158 p value threshold of 0.05 to exclude RSVs that were more common in extraction controls. Finally, we 159 removed samples if their mean Bray-Curtis distance (based on either weighted or unweighted 160 metrics) from breastmilk extraction controls was smaller than their mean distance from other 161 breastmilk samples collected from the same country (Supplementary Figure 1). CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. were detected with a prevalence of >5% in at least one of the groups being compared. We 188 supplemented cross-sectional analyses with longitudinal mixed-effects models of Shannon index 189 and taxon abundances (zero-inflated negative binomial models of genus-level sequence counts), 190 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint including week of life as a covariate and study ID as a random effect. Genera were included in 191 longitudinal models if they were present in 20% of samples in a given country.

217
. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint Figure 5). Genera underlying the predictive accuracy (based on mean importance scores) were 273 consistent with the discriminant taxa described above (Supplementary Table 1).

275
We also assessed alpha and beta diversity of breastmilk samples in relation to individual-level 276 variables measured in each cohort (Figure 2) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint
The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint discriminant genera were identified with respect to secondary outcomes based on longitudinal 299 models of genus abundance (Supplementary Table 2). 300

302
Breastmilk is a key source of pre-and pro-biotics that shape infant gut microbiota configuration. 303 This, in turn, plays a pivotal role in shaping immune development. We documented significant 304 differences in breastmilk microbiota composition between Malawi, India, and the UK across the 305 first 13 weeks of life. However, no consistent differences in breastmilk microbiota composition were 306 observed with respect to ORV response.  CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint breastmilk microbiota composition and ORV response in this cohort. Because of their low biomass, 354 breastmilk samples were subjected to extra rounds of PCR amplification to attain adequate material 355 for sequencing (35 cycles vs 25 used for stool), leading to amplification from extraction controls. We 356 accounted for this via stringent abundance-and prevalence-based filtering of potential 357 contaminants and excluded samples which clustered among extraction controls rather than other 358 breastmilk samples. Nonetheless, the potential contribution of contamination and site-specific batch 359 effects to the observed trends cannot be discounted. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 11, 2022. ; https://doi.org/10.1101/2022.11.09.22282115 doi: medRxiv preprint microbiota sequencing work. Above all, we are grateful to the families involved in the study. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  R  e  s  e  a  r  c  h  D  i  s  s  e  m  i  n  a  t  i  o  n  C  o  n  f  e  r  e  n  c  e  ,  B  l  a  n  t  y  r  e  ,  M  a  l  a  w  i  ,  2  4  t  h  -2  5  t  h  N  o  v  e  m  b  e  r  2  0