Rotavirus A shedding and HBGA host genetic susceptibility in a birth community-cohort, Rio de Janeiro, Brazil, 2014-2018.

Recent studies have investigated whether the human histo-blood group antigen (HBGAs) could affect the effectiveness of the oral rotavirus vaccines, suggesting secretor positive individuals develop a more robust response. We investigated the Rotavirus A (RVA) shedding in association with the host susceptibility profile in children from a birth community-cohort in Rio de Janeiro, Brazil, from 2014 to 2018. A total of 132 children were followed-up between 0 to 11-month-old, stool samples were collected before/after the 1st/2nd RV1 vaccination doses and saliva samples were collected during the study. RVA shedding was screened by RT-qPCR and G/P genotypes determined by multiplex RT-PCR and/or Sanger nucleotide sequencing. The sequencing indicated an F167L amino acid change in the RV1 VP8* P[8] in 20.5% of shedding follow-ups and these mutant subpopulations were quantified by pyrosequencing. The HBGA/secretor status was determined and 80.3% of the children were secretors. Twenty-one FUT2 gene SNPs were identified and two new mutations were observed. The mutant F167L RV1 VP8* P[8] was detected significantly more in Le (a+b+) secretors (90.5%) compared to non-secretors and even to secretors Le (a-b+) (9.5%). The study highlights the probable association between RV1 shedding and HBGAs as a marker for evaluating vaccine strain host susceptibility.

In relation to the other non-G1P [8] genotypes detected in this cohort in acute diarrheic episodes (ADE), the G3P[9] genotype was observed 11 days after the 1 st dose of RV1 (RT-qPCR Ct value of 19) and G3P [8] was detected in one 6 month-old child (RT-qPCR Ct value of 32.5). The G12P [8] genotype was detected in one 4-month-old child (7 days after the RV1 2 nd dose, RT-qPCR Ct value of 39.1) and one 10-month-old child (RT-qPCR Ct value of 36.4).
Detection of VP8* P[8] gene mutation. The F167L VP8* gene mutation in the RV1 G1P [8] characterized strains (F167L RV1 VP8* P [8]) was detected in 20.5% (27/132) of children monitored with shedding and was present in 23 children after the 1 st dose, 2 children after the 2 nd dose, 1 child before the 1 st dose (persisting after the 1 st dose, and 1 child before the 1 st vaccine dose; see Supplementary Table S1). The F167L RV1 VP8* P [8] was detected from day 5 to day 31 (median of 8 days, except for 2 cases detected before the 1 st dose), with a median RT-qPCR Ct value of 32.9. The RV1 G1P [8] shedding period (days post-vaccination) for children releasing these particles either with or without the F167L mutation was similar (median of 7 and 8 days).
Other RVA genotypes isolated in ADE in this cohort were detected in the secretor children, these being G12P [8] in Le (a−b+) and Le (a+b+) and G3P [8] in Le (a+b+). The G3P

Discussion
This prospective study focused mainly on assessing the RVA shedding in association to the HBGA profile in a birth community-cohort in Rio de Janeiro, Brazil. Through monitoring the RV1 G1P [8] shedding, it was possible to detect the F167L RV1 VP8* P [8] from stool samples of children from prior to the 1 st dose and during the RV1 immunization period. The evaluation of these children's host susceptibility profile showed that secretor as well as Lewis b positive children were significantly more likely to shed RV1 G1P [8] with occurrence of the F167L VP8* mutation.
Via VP8* P [8] Sanger sequencing, 88.5% of the G1P [8] RVA detected were identified as RV1, mainly after the 1 st dose (96.5%), at 2 months of age, similar to previously documented in clinical trials [34][35][36] . The G1P [8] genotype was detected in eight non-vaccinated children (before the 1 st dose) and due to the low viral load, only three cases could be sequenced and determined as RV1 G1P [8]. Despite the VP8* P [8] Sanger sequencing having been used to differ between RV1 and non-RV1 shedding 14,15 , we should strictly consider the possibility of the RV1 G1P [8] shedding here characterized containing a RVA constellation differing from the source RV1. Therefore, this study shows the importance of characterizing the RV1 via the RVA constellation method before a vaccine batch can be used to produce RV1 doses.
The horizontal transmission of vaccine viruses has been previously demonstrated between vaccinated and unvaccinated infants or other close contacts 35,36 . RV1 G1P [8] shedding was reported in two diarrhea cases after the 1 st dose, with the F167L mutation being detected in one of them. In this study we observed a low diarrhea incidence, mainly in the first 4 months of age, which could be related to RV1 protection and the effect of exclusive breastfeeding (and transferred maternal antibodies) in these first months of life 37 .
In the few diarrhea cases detected in this study, a rare G3P[9] was identified in one 2-month-old child after the 1 st dose, with the VP8* P[9] gene exhibiting 97% nucleotide similarity to an AU-1-like Brazilian strain (KJ820906) collected from an inpatient 2-year-old, indicating a possible feline/canine-to-human interspecies transmission 38 . The effectiveness of the RVA vaccines against genogroups 1 and 2 (Wa-like and DS-1-like strains, respectively) has been proven, but still remains unclear against the genogroup 3 (AU-1-like) 39,40 . As to this rare genotype detected in Manguinhos, further studies are being conducted to determine the complete genomic constellation.
Through our molecular characterization of the RV1 G1P [8] shedding, it was possible to observe a high frequency of the mutation in the 167 position of the VP8* P [8]    www.nature.com/scientificreports www.nature.com/scientificreports/ analyzed because it is directly related to interactions with HBGAs. In the previous studies with the G1P1A [8] rotavirus vaccine candidate 89-12, the precursor to RV1, Ward et al. 41 identified five aa changes in the VP4 gene (G51D and L167F in VP8* domain; D331Y, D385Y and N695I in VP5* domain) that occurred during vaccine attenuation. It was proposed that these alterations could reduce cross neutralizing antibody responses. Indeed, according to Gozalbo-Rovira et al. 42 , the 167 position is placed at the bottom of the sugar-binding pocket, and Phe residue contained in RV1 G1P [8] could influence the interaction between the vaccine strain and the sugar of the cellular receptor, due to hydrophobic residue decreasing the binding affinity. 32.9% of the RV1 G1P [8] shedding samples detected in this birth-cohort presented the mutation in the 167 position, with a mixture of virus subpopulations of both amino acids, Leu (CTT, TTA, CTA) and Phe (TTT), in different percentages. We could hypothesize that this suggest an initial process of reversion of attenuation or a selective pressure, through the one and/or two alteration(s) in this codon (positions nt 499, 501), favoring the highest amount of circulation of strains containing the Leu residue, which has a greater avidity for the HBGA's binding site in the secretor profile. Zeller et al. 43 also reported the mutation in position 167 in the 3 rd position of the codon (nt 501), in one vaccine derived strain (BE00048) detected in Belgium, in 2009.
Positive secretor status was the most prevalent (80.3%) in this community-cohort, and the Le (a+b+) profile (59.8%) frequency was similar to that observed in younger children from the Amazon (58.8%) as presented in our previous study 44 . The rs1047781 (385 A > T) SNP, responsible for the weak genotype in the Asian population, was not detected in Le (a+b+) children in the Manguinhos community, Rio de Janeiro state. The rs281377 (357 C > T), a synonymous mutation, was the most frequent SNP detected in this Le (a+b+) secretor profile, and this SNP has been detected in the Brazilian population 44,45 . According to Ferrer-Admetlla et al. 19 , the rs281377 (357 C > T) SNP has been detected in a worldwide distribution profile in the natural population, and the se 357/385 haplotype is frequently detected in the Asian population. We could explain the high frequency of the Le (a+b+) phenotype detected in children in Rio de Janeiro, Brazil under an evolutionary view that, at some point in the molecular clock, the se 357/385 haplotype had diverged and separated from the rs1047781 (385 A > T) SNP. However, epistatic interactions between rs281377 (357 C > T) and other SNPs could affect the expression of the FUT2 gene. The failure in detecting the rs1047781 (385 A > T) SNP, which has been attributed as responsible for the weak secretor phenotype detected in Le (a+b+), in the children in the Manguinhos community in the state of Rio de Janeiro could be explained by the presence of new SNPs in conjunction with rs281377 (357 C > T). Further studies encompassing larger populations, including Le (a − b + ) secretors will be needed to evaluate whether particular SNPs in FUT2 gene can explain the phenotypes Le (a+b+) vs Le (a − b + ) (or similar).
Our study identified SNPs in the FUT2 gene in children from the Manguinhos community, in Rio de Janeiro as rs1800021 (40 A > G), rs48703160 (171 A > G), rs681343 (216 C > T), rs281377 (357 T > C), rs601338 (428 G > A), rs602662 (739 G > A) and rs485186 (960 A > G), also reported by Vicentini et al. 45 in children from a Quilombola community (black population, slave descendants) in Espírito Santo (borders with the state of Rio de Janeiro, both states in Southeastern Brazil).
Le (a+b+) phenotype is very common in infants and children under 2 years-old, usually being a transient status in most children 46 . In the infancy period, the Lewis phenotypes have not yet fully matured 47 , due to the activity of the fucosyltransferases (FUT2 and FUT3 genes) not yet reaching the normal levels (enzyme activity of Le > Se) 20 .
In relation to the non-secretor phenotype, a unique child (0Le (a−b−)) did not present any SNP in the FUT2 gene, i.e., not showing the nonfunctional allele se 428 (rs601338) 21 . This child shed the RV1 G1P [8] strain virus in both doses (6 days post-1 st dose; 7 days post-2 nd dose); however, only in the 2 nd dose was the F167L mutation detected, with mixed virus subpopulations of 63.5% Phe/36.5% Leu. Possibly, other genetic factors could be related to this profile and vaccinal response.
The P[9] RVA genotype interacts with A-type HBGA antigen 48 , and the AB0 enzyme assay was performed in a saliva sample from the unique child that was infected with this genotype (G3P[9]) in this cohort, presenting positivity for type A antigen (ALe (a−b−) secretor) (data not shown), thus in line with the proposed susceptibility of A to genotype P[9].
The Manguinhos cohort showed a high prevalence of secretors, being a common profile in Latin America 30,31,44,45,49 . In this study, we were able to investigate RVA shedding and a common mutation was identified in many samples characterized as RV1 according to VP8* Sanger sequencing criteria. This mutational event suggests that the vaccine could be replicating and therefore a more robust immune response may be produced in these infants. In summary, our study demonstrated that Le (a+b+) secretors can influence vaccine replication.
This study highlights the detection of the new mutations in the FUT2 gene detected in the community of Manguinhos, Rio de Janeiro, Brazil and the association between vaccine strain shedding and HBGAs as a marker for evaluating vaccine efficiency. The results in this study indicate the importance of the full molecular monitoring of the RV1 vaccine virus stool shedding, mainly if the F167L VP8* mutation is frequent in other populations, and how the host susceptibility profile can influence this viral selective pressure, besides the questions about circulation and fluctuations of the emergent RVA genotypes and ensuring vaccine efficacy in a post-RVA vaccine era.