Significant Correlation Between the Infant Gut Microbiome and Rotavirus Vaccine Response in Rural Ghana
JID
Significant Correlation Between the Infant Gut Microbiome and Rotavirus Vaccine Response in Rural Ghana
Vanessa C. Harris () 1 2
George Armah 5
Susana Fuentes 0
Katri E. Korpela 4
Umesh Parashar 3
John C. Victor 8
Jacqueline Tate 3
Carolina de Weerth 6
Carlo Giaquinto 7
Willem Joost Wiersinga 1
Kristen D. C. Lewis 1
Willem M. de Vos 0 4
0 Laboratory of Microbiology, Wageningen University
1 Center for Experimental and Molecular Medicine, Division of Infectious Diseases, Academic Medical Center, University of Amsterdam
2 Amsterdam Institute for Global Health and Development
3 Division of Viral Diseases, National Center for Immunization and Respiratory Diseases, Center for Disease Control and Prevention , Atlanta , Georgia
4 Department of Bacteriology and Immunology, and Immunobiology, University of Helsinki , Finland
5 Noguchi Memorial Institute for Medical Research, College of Health Sciences, University of Ghana , Legon
6 Behavioral Science Institute, Radboud University , Nijmegen , The Netherlands
7 Department of Paediatrics, University of Padova , Italy
8 PATH, Vaccine Access and Delivery , Seattle, Washington
(See the editorial commentary by Iturriza-Gómara and Cunliffe on pages 8-10.) Background. Rotavirus (RV) is the leading cause of diarrhea-related death in children worldwide and 95% of RV-associated deaths occur in Africa and Asia where RV vaccines (RVVs) have lower efficacy. We hypothesize that differences in intestinal microbiome composition correlate with the decreased RVV efficacy observed in poor settings. Methods. We conducted a nested, case-control study comparing prevaccination, fecal microbiome compositions between 6week old, matched RVV responders and nonresponders in rural Ghana. These infants' microbiomes were then compared with 154 age-matched, healthy Dutch infants' microbiomes, assumed to be RVV responders. Fecal microbiome analysis was performed in all groups using the Human Intestinal Tract Chip. Results. We analyzed findings in 78 Ghanaian infants, including 39 RVV responder and nonresponder pairs. The overall microbiome composition was significantly different between RVV responders and nonresponders (FDR, 0.12), and Ghanaian responders were more similar to Dutch infants than nonresponders (P = .002). RVV response correlated with an increased abundance of Streptococcus bovis and a decreased abundance of the Bacteroidetes phylum in comparisons between both Ghanaian RVV responders and nonresponders (FDR, 0.008 vs 0.003) and Dutch infants and Ghanaian nonresponders (FDR, 0.002 vs 0.009). Conclusions. The intestinal microbiome composition correlates significantly with RVV immunogenicity and may contribute to the diminished RVV immunogenicity observed in developing countries.
-
against severe RV gastroenteritis ranging from 48% to 64%
for both Rotarix and RotaTeq vaccines in Africa and Asia
[3–5]. Emerging effectiveness data in Africa provides similar
estimates of RVV protection [6]. This compares to an observed
efficacy of 85%–98% against severe RV in trials in wealthier
countries in Latin America and Europe [7–10].
Understanding the pathophysiologic mechanism driving this
diminished efficacy in developing countries is critical, because
even small improvements in vaccine efficacy could increase
the number of children’s lives saved by the vaccine by hundreds
of thousands over the next 15 years [11]. There are several
hypotheses as to why oral RVVs are underperforming in Africa
and Asia [12]. These include interference with the first dose
of coadministered oral poliovirus vaccine, RVV immune
response suppression through high prevaccination levels of
serum immunoblogulin (Ig) G, including transplacentally
derived IgG, high levels of breast milk–derived RV-specific IgA,
and HLA blood group antigen type [13–16]. However, none
of these explanations have adequately and sufficiently explained
the underperformance of RVV in developing countries, where
vaccine efficacy can dip even lower than 50% in some settings.
One underexplored hypothesis is that the intestinal microbiome
may be modulating an infant’s immune response to the enteric
RVV [17]. We hypothesized that the composition of the
intestinal microbiome is correlated with RVV response, that RVV
responders have different intestinal microbes as compared
with nonresponders and that these dissimilarities may
contribute to the decreased efficacy of RVV found in resource-poor
settings. To test these hypotheses, we conducted a nested,
casecontrol study in Navrongo, Ghana, comparing the differences
in intestinal microbiome composition and diversity between
RVV seroconverters and nonseroconverters after vaccination
with the Rotarix vaccine. We then compared these infants’
microbiomes with those from a large group of age-matched
healthy infant from the Netherlands, where RVV response is
postulated to be high, similar to responses observed in other
Northern European countries [10, 18].
Study Design and Participants
Ghanaian Infants
The original trial within which this trial was nested was
conducted in Navrongo, a rural setting in Northern Ghana where
>70% of the population belong to the lowest wealth quintile in
Ghana, in 2012. The neonatal and mortality rates are 24 and 46
deaths per 1000 live births, respectively [19].
All participating infants were healthy infants with a birth weight
> 2000 g and/or a gestational age >38 weeks. The infants were
enrolled at 6 weeks of age in a previously reported phase IV
randomized clinical trial conducted in 2012 in Navrongo to evaluate the
immunogenicity of the Rotarix vaccine after different dosing
schedules (at age 6 and 10 weeks, 10 and 14 weeks, or 6, 10,
and 14 weeks) (NCT01575197, clinicaltrials.gov) [20]. In this
trial, all infants received concomitant standard Expanded Program
on Immunization vaccinations, including the trivalent oral
poliovirus vaccine and pentavalent vaccine (diphtheria, tetanus,
wholecell pertussis, hepatitis B, and Haemophilus influenza type).
Only infants from the 6- and 10-week and the 6-, 10-, and
14week dose arms of the clinical trial were included in this
microbiome study. A serum samples was collected before the receipt
of the first dose of vaccine (at 6 weeks of age) and serum was
collected again approximately 4 weeks after the last dose of
vaccine (at age 14 or 18 weeks, depending on the study arm) for
anti-RV IgA antibody measurements. Fecal samples were
collected immediately before vaccination, at 6 weeks of age.
Infants were included in the study if during the original
study their guardians had consented to additional testing of
specimens in RV-vaccine related studies. Inclusion criteria
further mandated that a baseline fecal sample was available and
that there was no evidence of natural RV infection before
vaccination ( prevaccination IgA level ≥20 IU/mL). An IgA level
≥20 IU/mL was considered an indication of seroconversion
and a surrogate marker for RVV protection against severe RV
gastroenteritis [21].
Participating infants were then grouped as either RVV
responders ( postvaccination anti-RV IgA antibody ≥20 IU/mL)
or RVV nonresponders ( postvaccination anti-RV IgA antibody
<20 IU/mL) and matched by hand in a 1:1 ratio using the
following ranked variables: number and timing of doses of vaccine
received (at age 6 and 10 weeks or 6, 10, and 14 weeks), sex, age at
vaccination, RV season (defined as the date of the serum IgA level
obtained 28 days after the last vaccination, between 1 December
2012 and 1 March 2013), ethnicity, height and weight at
enrollment (including underweight, stunting, and wasting Z scores),
and whether the infant regurgitated the vaccine after
administration. The exact mode of delivery data and breastfeeding practices
data are not known for the original Ghanaian study population.
However, all infants in Ghana were delivered in the study hospital
where >95% of deliveries are vaginal, and almost all infants
delivered vaginally are breastfed. The nested trials were approved by
the institutional review board of the Noguchi Memorial Institute
for Medical Research and the research was conducted in
accordance with good clinical practice guidelines.
Dutch Infants
In parallel, we compared the microbiome of both the Ghanaian
responder and nonresponder infants with those in a cohort of
healthy, age-matched Dutch control infants. The Dutch infants
had not received RVV, but were assumed to be RVV responders,
in line with ample clinical trial data demonstrating a >90% RVV
seroconversion rate in Northern European countries [7, 10]. Fecal
samples for these control infants were collected at approximately
30 days of age as part of a previously published study (the Bibo
study) in which mothers and their children were followed up
beginning with the third trimester of pregnancy [22, 23]. Pregnant
women were recruited through midwife practices in Nijmegen
and surrounding areas in the Netherlands. All parents provided
written informed consent for participation in the study. The
study was approved by the ethical committee of the faculty of
social sciences at the Radboud University in Nijmegen. The infants’
microbiomes were analyzed in the study using an identical
protocol based on the Human Intestinal Tract Chip (HITChip)
phylogenetic microarray [22]. Consent had been given for further use
of the samples, which were entirely anonymous in our analysis.
Laboratory Evaluations
Assays
Anti-RV IgA antibody was measured in serum using an
enzyme-linked immunosorbent assay described elsewhere, with
values expressed in international units per milliliter [24, 25].
Fecal Microbiome and Enteric Pathogen Analysis
In Ghana, fecal samples were collected by community health
workers in infants’ homes, transported in a cool box, and frozen
to −20°C within 24–48 hours of collection. All samples were
stored in 3% glycerol in frost-free freezers. Routine temperature
monitoring did not indicate any freeze-thaw cycles. In the
Netherlands, fecal samples were collected by parents at home
and immediately stored at −20°C. The samples were
transported in coolers with freezing cartridges or dry ice for further
storage at −80°C. These procedures are both considered adequate
for fecal microbiome analysis [26].
The fecal samples were analyzed by means of HITChip
microarrays in duplicate, as described elsewhere [27]. In brief, total
DNA was extracted from the fecal material by a repeated bead
beating procedure using a modified protocol for the QiaAmp
DNA MiniStool Kit (Qiagen), also as described elsewhere
[28]. The 16S ribosomal RNA gene was amplified using primers
that enabled incorporation of T7 promoter sequence at the
5terminus of the amplicon. RNA was transcribed with
aminoallyl modified nucleotides that were later coupled to cyanine
(Cy) 3 or Cy5 dyes. Labeled RNA was fragmented, and 2
samples, each carrying a different dye, were hybridized in duplicate
to the HITChip microarrays.
The HITChip microarray is a comprehensive and highly
reproducible phylogenetic microarray that enables the parallel
profiling and the semiquantitative analysis of >1100 phylotypes
representing all major intestinal phyla grouped in 130 genuslike
groups described for the human intestinal microbiome [27].
This high-throughput technique has been benchmarked with
ultradeep pyrosequencing of 16S ribosomal RNA amplicons
and next-generation parallel sequencing of intestinal
metagenomes [29, 30]. After microarray hybridization, a sample was
accepted only if 2 independent hybridizations (labeled with
Cy5 and Cy3) correlated significantly (>95% Pearson
correlation). This highly reproducible microbiome analysis allowed
for direct comparison of all samples described in this study.
Statistical Analysis
A χ2 test was used to determine statistically significant differences
in baseline characteristics between Ghanaian RVV responders
and nonresponders. The Shannon diversity index was used to
measure the diversity of the microbiome per sample, including
richness and evenness, using the hybridization signal of all probes
included in the HITChip microarray [31]. Paired 2-tailed Student
t tests were used to evaluate statistical significance.
Comprehensive multivariate statistical analyses were
performed using Canoco 5.0 software for Windows [32].
Genuslevel principal coordinate analysis and redundancy analyses
were performed to evaluate differences in the overall microbial
composition between the Ghanaian study groups (Ghanaian
RVV responders and nonresponders). The 130 genuslike
bacterial groups targeted by the HITChip microarray were used as
biological variables, and the matching variables named above
were used as background variables and visualized by means of
inverse distance-weighted interpolation of the variable values
over the component space. Monte Carlo permutation testing
was used to assess the significance of the effect of these variables
in the data set.
Generalized linear models were used to identify individual
bacterial phyla and genuslike groups (class for Firmicutes)
associated with RV seroconversion, as differing significantly either
between Ghanaian nonresponders and responders or between
the Ghanaian nonresponders and Dutch infants. P values
corrected for the false discovery rate (FDR) were used to correct for
multiple testing [33]. Bacterial taxa, whose log-transformed
relative abundance was differed significantly (FDR-corrected
P < .10) between the Ghanaian nonresponders and responders
or between the Ghanaian nonresponders and Dutch infants,
were considered correlated with RVV immunogenicity. These
statistical analyses were performed with the program R [34].
To evaluate the similarity between Ghanaian infants
(nonresponders and responders) and Dutch infants, we generated
Pearson correlation scores and then performed an analysis of
variance between Ghanaian nonresponders and responders.
Ghanaian Infants
A total of 234 infants (of which 74 were RVV responders) with
prevaccination fecal samples were available from the original
Ghana clinical trials and eligible for this study. A total of 52
RV responders were successfully matched to 52 nonresponders
in a 1:1 ratio based on the predefined matching criteria. Of
those, a total of 78 samples (equaling 39 per-protocol matched
pairs) had sufficient DNA of good enough quality to be
successfully analyzed using the HITChip pipeline.
No significant differences between the vaccine responders
and nonresponders in Ghana were identified for any of the
measured variables and baseline characteristics as described in
Table 1. The diversity index (Shannon index) of the intestinal
microbiome did not differ between infants who were RVV
responders and those who were nonresponders (mean [SD],
4.40 [0.24] vs 4.41 [0.25], respectively; P = .87).
In Ghana, a high abundance of the Bacteroidetes phylum
(FDR, 0.003), particularly several bacteria related to Bacteroides
and Prevotella species, were significantly correlated with a lack
of RVV response (Figures 1 and 2). Conversely, the Bacilli
phylum (FDR, 0.027) correlated significantly with RVV response,
specifically bacteria related to Streptococcus bovis (FDR = 0.008)
(Figures 1 and 2).
We further evaluated the high abundance of bacteria in the
Bacteroidetes phylum in nonresponder infants. Recent
literature has suggested that many bacteria in the Bacteroidetes
phylum have a less immunogenic lipopolysaccharide (LPS) than
Enterobacteriacae, such as Escherichia coli [35]. We therefore
calculated an enterobacteria-Bacteroidetes ratio to provide a
rough estimate of the presence of toxigenic LPS in the
microbiome and compared the ratios between all groups. The
RVV Nonresponders and Responders as Determined With χ2 Tests
Infants, No. (%)a
Abbreviations: RV, rotavirus; RVV, RV vaccine.
a Data represent No. (%) of infants unless otherwise specified.
enterobacteria-Bacteroidetes ratio was significantly higher in
the Ghanaian responder infants (P = .04) than in nonresponder
infants (Figure 1C).
Spearman correlation analyses were performed to assess
whether the actual titer of the IgA—as opposed to RV
seroconversion as a binary variable (≥20 IU/mL)—was correlated with
specific bacterial groups. The Spearman correlation results
showed that 33 genuslike bacterial groups were significantly
correlated with RVV response (defined as an FDR <0.200). As
with the results obtained when using IgA as a binary variable,
bacteria related to Streptococcus bovis (FDR, 0.070) were the
only bacteria whose abundance were correlated significantly
(FDR, <0.1) with increased RVV titers. All other bacterial
groups were significantly associated with lower RVV titers,
and findings in only 1 bacterial group (bacteria related to
Xanthomonadaceae) differed from those of the statistical analysis
using IgA as a binary variable. Therefore, the Spearman
correlation analysis using absolute IgA titers both mirrored and
confirmed the findings attained when evaluating IgA as a binary
variable with 20 IU/mL as a cutoff point. (Supplementary
Figure 1).
Finally, the genuslike principal coordinate analysis indicated
that the main variable differentiating the microbiome
composition between the Ghanaian responders and nonresponders was
RVV seroconversion, as calculated with Monte Carlo
permutation testing (P = .01; FDR 0.12). (Supplementary Figure 2).
None of the other environmental variables that were tested
correlated with significant microbiome differences.
Ghanaian Infants Compared With Dutch Infants
The Ghanaian infant microbiome compositions were then
compared with the fecal microbiomes of 154 healthy Dutch infants.
Of these Dutch infants, 62% were still receiving breast milk at 4
weeks of age, 42% were female, and 93% had been delivered
vaginally; their mean (SD) birth weight was 3619 (457)
g. When the Dutch infants were compared with the Ghanaian
infants, their overall microbiome composition was significantly
more similar to that in the Ghanaian RVV responders than to
that in nonresponders (P = .002; Figure 1B).
We then evaluated which phyla and bacterial genus groups
differed significantly between Dutch infants and Ghanaian
RVV nonresponders (see Supplementary Table 1) and
subsequently assessed which of these bacterial groups also differed
significantly in the comparison between Ghanaian RVV
responders and nonresponders (Figure 3). At the phylum level,
bacilli were significantly more abundant in both Ghanaian
RVV responders and Dutch infants than in nonresponder
infants (FDR for Dutch vs Ghanaian RVV nonresponders,
0.0002 ). The Bacteroidetes phylum was also significantly
more abundant in Ghanaian nonresponders than both the
Ghanaian responders and the Dutch infants (FDR, 0.006).
At the genus level, all bacterial groups that differed
significantly between Ghanaian nonresponders and responders
(FDR, <0.1) also differed significantly between Ghanaian
nonresponders and Dutch infants. Bacteria related to Streptococcus
bovis were more abundant in both Dutch infants and Ghanaian
responders than in the Ghanaian infants without an RVV
immune response (FDR, 0.0018). In parallel, several bacteria from
the Bacteroidetes phylum (FDR, 0.002), particularly the
Bacteroides and Prevotella genera, were more abundant in the
Ghanaian nonresponders than in the Dutch infants (all Figure 3).
Finally, the enterobacteria-Bacteroidetes ratio was also
significantly higher (P < .01) in Dutch infants than in Ghanaian
infants without an RVV immune response (Figure 1C).
This study of Ghanaian infants demonstrates that the
prevaccination intestinal microbiome differs significantly—on a
genuslike, phylum, and overall composition level—between RVV
responders and nonresponders in a rural, low-income setting
in sub-Saharan Africa and that these microbiome differences
are robustly recapitulated when Dutch infants are compared
with RVV nonresponders.
RVV response was positively associated with the Bacilli
phylum, specifically bacteria related to Streptococcus bovis. Several
bacterial groups were significantly associated with a lack of
RVV response—namely, bacteria belonging to the Bacteroidetes
phlyum, especially several bacteria related to species from the
Bacteroides and Prevotella genera.
The high abundance of Bacteroidetes in the microbiome of
infants not responding to RVV is intriguing. Certain species
from that phylum have been shown to express a form of LPS
that is functionally and structurally different from the LPS
expressed in some proteobacteria, such as E. coli [35, 36]. LPS is
present in the outer membrane of most gram-negative bacteria
and is a strong immunogenic stimulator of the innate immune
system [37]. LPS structure varies from species to species, and
variation—particularly in the lipid A core of LPS—can
influence LPS’ immunostimulatory capacity [38]. LPS derived from
several Bacteroidetes species has recently been shown to have an
impaired or even immune-inhibitory capacity to stimulate
inflammatory cytokines in vitro when compared with LPS derived
from E. coli [35].
Although our study could not measure infants’ microbiome
expression of a less toxigenic LPS, we chose to roughly evaluate
this phenomenon by comparing the ratio of all enterobacteria to
Bacteroidetes in each infant group. We did not compare specific
enterobacteria or Bacteroidetes species because of some
crosshybridization at the species level in the HITChip microarray.
This enterobacteria-Bacteroidetes ratio was significantly
increased in both the Ghanaian RVV responders and Dutch
infants compared with the nonresponders, raising the possibility
that early microbiome colonization with bacteria expressing a
less toxigenic or perhaps inhibitory form of LPS might be
down-regulating innate immune responses to the live
attenuated RV contained in the vaccine. Alternatively, more toxic LPS
might be having an adjuvant effect on RVV responses in those
Ghanaian infants capable of eliciting an immunogenic response
to RVV.
Interestingly, in the Ghanaian responders, bacteria related
to Streptococcus bovis was significantly correlated with RV
response. The Streptococcus bovis group belongs to the
S. bovis–Streptococcus equinus complex, which, like bacteria
with toxigenic LPS, such as the E. coli, Serratia, and
Klebsiella groups, can cross from being commensals to being
opportunistic pathogens and whose cell surface proteins can
trigger inflammatory responses [39]. Although speculative,
these S. bovis–S. equinus complex bacteria could also be
priming the immune system or acting as natural vaccine
adjuvants.
An alternative hypothesis, given that the vaccines contain live
attenuated virus, is that these bacteria might be expressing blood
group antigens or glycans needed for RV replication, as has
recently been demonstrated with norovirus and RV [40, 41].
Unfortunately, FUT2 secretor status, which is epidemiologically
associated with RV gastroenteritis, is unknown among these infants.
This study has some limitations, restricting the strength of
our findings. One of the most important is that, for the cohort
of Dutch infants, we do not have RVV immunogenicity data
and are predicting RVV response from randomized control
studies showing >90% RVV efficacy in Europe [7, 10]. Because
these were all healthy, at-term infants, we do predict that they
would mount immune responses to RVV but are unable to
demonstrate this. In addition, the study lacks specific data on
breastfeeding and delivery practices for the infants in Ghana,
and these may be confounders in our study results.
Nevertheless, because >95% of deliveries are vaginal and all infants
delivered vaginally are breastfed in this cohort, we do not expect
mode of delivery and breastfeeding practices to be significant
confounders in this study population. We also were not able
to match our infants in Ghana for maternal antibodies to RV,
such as breast milk anti-RV IgA and transplacentally acquired
anti-G1 RV IgG antibody. High levels of maternally derived
antibodies have been shown to diminish RVV immunogenicity,
which could be contributing to the lack of response observed
in our cohort and confounding the results [14, 15].
As another limitation, we used anti-RV IgA response as a
correlate of vaccine protection, which can be an imperfect
surrogate for vaccine efficacy [21]. Clinical vaccine efficacy would
require larger sample sizes and follow-up. In addition, the
intestinal microbiome is a complex ecosystem, and bacterial
populations are in constant interplay with other intestinal inhabitants
and pathogens, such as archae and eukaryotic microbes. This
analysis of the microbiome did not evaluate the intestinal
virome, fungiome, or parasites, and these could be immunologic
mediators and influence bacterial populations. Finally, the
associations between microbiome and RVV response are correlative
and not causative and taken at a single time-point—directly
before vaccination in a very young infants with high variability in
their microbiome population. Nevertheless, our study used an
identical, standardized, and reproducible technique to measure
the intestinal microbiome in 2 geographic locations, and
matching for numerous demographic variables in Ghana helps
minimize possible study cofounders. As a consequence, our work is a
springboard to understanding the mechanistic relationships
between the intestinal microbiome and oral RVV responses in
vulnerable infant populations and a call for more research to
inform the design of evidence-based interventions to improve
RVV efficacy worldwide.
Supplementary Data
Supplementary materials are available at http://jid.oxfordjournals.org.
Consisting of data provided by the author to benefit the reader, the posted
materials are not copyedited and are the sole responsibility of the author, so
questions or comments should be addressed to the author.
Acknowledgments. This study is dedicated to Professor Joseph (Joep)
Lange, who died before its completion. It would not have been possible
without his initiative, support, and inspiration. We acknowledge and
thank all the families who participated in this study and the staff members
of the RVV trial team in Ghana for their work in conducting this study;
Monica McNeal and her staff at the Laboratory of Specialized Clinical
Studies, Cincinnati Children’s Hospital, who performed all the immunoglobulin
testing for the original dosing studies; Duncan Steele for his contribution to
the original dosing study in Ghana; and Lauren Gazley for her contribution
to the original dosing study in Ghana and data management support for this
research in Ghana.
Disclaimer. The findings and conclusions in this report are those of the
authors and do not necessarily represent the official positions of policies of
PATH, the Bill & Melinda Gates Foundation, the US Centers for Disease
Control and Prevention, or the Netherlands Organization for Scientific
Research.
Financial support. This publication is based on research funded in part
by PATH through the Bill & Melinda Gates Foundation (grant OPP
1017334 to PATH) and the Netherlands Organization for Scientific
Research (Spinoza grant 2008 to W. M. d. V.).
Potential conflict of interest. C. G. has received honoraria for
participating in advisory boards and speaking at international conferences from
GlaxoSmithKline, Merck, and Sanofi PasteurMSD. All other authors report
no potential conflicts. All authors have submitted the ICMJE Form for
Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider
relevant to the content of the manuscript have been disclosed.
1. Tate JE , Burton AH , Boschi-Pinto C , et al. 2008 estimate of worldwide rotavirusassociated mortality in children younger than 5 years before the introduction of universal rotavirus vaccination programmes: a systematic review and meta-analysis . Lancet Infect Dis 2012 ; 12 : 136 - 41 .
2. Fischer Walker CL , Black RE . Rotavirus vaccine and diarrhea mortality: quantifying regional variation in effect size . BMC Public Health 2011 ; 11 (Suppl 3): S16 .
3. Armah GE , Sow SO , Breiman RF , et al. Efficacy of pentavalent rotavirus vaccine against severe rotavirus gastroenteritis in infants in developing countries in subSaharan Africa: a randomised, double-blind, placebo-controlled trial . Lancet 2010 ; 376 : 606 - 14 .
4. Madhi SA , Cunliffe NA , Steele D , et al. Effect of human rotavirus vaccine on severe diarrhea in African infants . N Engl J Med 2010 ; 362 : 289 - 98 .
5. Zaman K , Dang DA , Victor JC , et al. Efficacy of pentavalent rotavirus vaccine against severe rotavirus gastroenteritis in infants in developing countries in Asia: a randomised, double-blind, placebo-controlled trial . Lancet 2010 ; 376 : 615 - 23 .
6. Baar-Zeev N , Kapanda L , Tate JE , et al. Effectiveness of a monovalent rotavirus vaccine in infants in Malawi after programmatic roll-out: an observational and case-control study . Lancet Infect Dis 2015 ; 15 : 422 - 8 .
7. Ruiz-Palacios GM , Pérez-Schael I , Velázquez FR , et al. Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis . N Engl J Med 2006 ; 354 : 11 - 22 .
8. Linhares AC , Velázquez FR , Pérez-Schael I , et al. Efficacy and safety of an oral live attenuated human rotavirus vaccine against rotavirus gastroenteritis during the first 2 years of life in Latin American infants: a randomised, double-blind, placebo-controlled phase III study . Lancet 2008 ; 371 : 1181 - 9 .
9. Vesikari T , Karvonen A , Prymula R , et al. Efficacy of human rotavirus vaccine against rotavirus gastroenteritis during the first 2 years of life in European infants: randomised, double-blind controlled study . Lancet 2007 ; 370 : 1757 - 63 .
10. Vesikari T , Matson DO , Dennehy P , et al. Safety and efficacy of a pentavalent human-bovine (WC3) reassortant rotavirus vaccine . N Engl J Med 2006 ; 354 : 23 - 33 .
11. Atherly D , Dreibelbis R , Parashar UD , Levin C , Wecker J , Rheingans RD . Rotavirus vaccination: cost-effectiveness and impact on child mortality in developing countries . J Infect Dis 2009 ; 200 (suppl 1): S28 - 38 .
12. Clarke E , Desselberger U. Correlates of protection against human rotavirus disease and the factors influencing protection in low-income settings . Mucosal Immunology 2015 ; 8 : 1 - 17 .
13. Emperador DM , Velasquez DE , Estivariz CF , et al. Interference of monovalent , bivalent, and trivalent oral poliovirus vaccines on monovalent rotavirus vaccine immunogenicity in rural Bangladesh . Clin Infect Dis 2016 ; 62 : 150 - 6 .
14. Moon S-S , Groome MJ , Velasquez DE , et al. Prevaccination rotavirus serum IgG and IgA are associated with lower immunogenicity of live, oral human rotavirus vaccine in South African infants . Clin Infect Dis 2016 ; 62 : 157 - 65 .
15. Chilengi R , Simuyandi M , Beach L , et al. Association of maternal immunity with rotavirus vaccine immunogenicity in Zambian infants . PLoS One 2016 ; 11 : e0150100.
16. Nordgren J , Sharma S , Bucardo F , et al. Both Lewis and secretor status mediate susceptibility to rotavirus infections in a rotavirus genotype-dependent manner . Clin Infect Dis 2014 ; 59 : 1567 - 73 .
17. Isolauri E , Joensuu J , Suomalainen H , Luomala M , Vesikari T. Improved immunogenicity of oral D x RRV reassortant rotavirus vaccine by Lactobacillus casei GG . Vaccine 1995 ; 13 : 310 - 2 .
18. Ruiz-Palacios GM , Guerrero ML , Bautista-Márquez A , et al. Dose response and efficacy of a live, attenuated human rotavirus vaccine in Mexican infants . Pediatrics 2007 ; 120 : e253 - 61 .
19. Ghana Statistical Service, Ghana Health Service, and ICF International. Ghana demographic and health survey , 2014 . Rockville, MD: GSS, GHS, and ICF International , 2015 . Available at: https://dhsprogram.com/pubs/pdf/FR307/FR307.pdf. Accessed 20 June 2016 .
20. Armah G , Lewis KDC , Cortese MM , et al. A randomized controlled trial of the impact of alternative dosing schedules on the immune response to human rotavirus vaccine in rural Ghanaian infants . J Infect Dis 2016 ; 213 : 1678 - 85 .
21. Patel M , Glass RI , Jiang B , Santosham M , Lopman B , Parashar U. A systematic review of anti-rotavirus serum IgA antibody titer as a potential correlate of rotavirus vaccine efficacy . J Infect Dis 2013 ; 208 : 284 - 94 .
22. de Weerth C , Fuentes S , Puylaert P , De Vos WM . Intestinal microbiota of infants with colic: development and specific signatures . Pediatrics 2014 ; 131 : e550 - 8 .
23. Zijlmans MAC , Korpela K , Riksen-Walraven JM , De Vos WM , de Weerth C. Maternal prenatal stress is associated with the infant intestinal microbiota . Psychoneuroendocrinology 2015 ; 53 : 233 - 45 .
24. Bernstein DI , Smith VE , Sherwood JR , et al. Safety and immunogenicity of live, attenuated human rotavirus vaccine 89-12. Vaccine 1997 ; 16 : 381 - 7 .
25. Ward RL , Bernstein DI , Shukla R , et al. Protection of adults rechallenged with a human rotavirus . J Infect Dis 1990 ; 161 : 440 - 5 .
26. Tedjo DI , Jonkers DMAE , Savelkoul PH , et al. The effect of sampling and storage on the fecal microbiota composition in healthy and diseased subjects . PLoS One 2015 ; 10 :e0126685.
27. Rajilić-Stojanović M , Heilig HGHJ , Molenaar D , et al. Development and application of the Human Intestinal Tract Chip, a phylogenetic microarray: analysis of universally conserved phylotypes in the abundant microbiota of young and elderly adults . Environ Microbiol 2009 ; 11 : 1736 - 51 .
28. Salonen A , Nikkilä J , Jalanka-Tuovinen J , et al. Comparative analysis of fecal DNA extraction methods with phylogenetic microarray: Effective recovery of bacterial and archaeal DNA using mechanical cell lysis . J Microbiol Methods 2010 ; 81 : 127 - 34 .
29. Claesson MJ , O'Sullivan O , Wang Q , et al. Comparative analysis of pyrosequencing and a phylogenetic microarray for exploring microbial community structures in the human distal intestine . PLoS One 2009 ; 4 : e6669 .
30. Arumugam M , Raes J , Pelletier E , et al. Enterotypes of the human gut microbiome . Nature 2011 ; 473 : 174 - 80 .
31. Haegeman B , Hamelin JEROM , Moriarty J , Neal P , Dushoff J , Weitz JS . Robust estimation of microbial diversity in theory and in practice . ISME J 2013 ; 7 : 1092 - 101 .
32. Leps J , Smilauer P. Multivariate analysis of ecological data using CANOCO . Cambridge, United Kingdom: Cambridge University Press , 2003 .
33. Benjamini Y , Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing on JSTOR . J R Stat Soc Ser B 1995 ; 57 : 289 - 300 .
34. The R Project for Statistical Computing. 2016 . https://www.r-project.org.
35. Vatanen T , Kostic AD , d'Hennezel E , et al. Variation in microbiome LPS immunogenicity contributes to autoimmunity in humans . Cell 2016 ; 165 : 842 - 53 .
36. Hofstad T , Sveen K , Dahlén G . Chemical composition, serological reactivity and endotoxicity of lipopolysaccharides extracted in different ways from Bacteroides fragilis, Bacteroides melaninogenicus and Bacteroides oralis . Acta Pathol Microbiol Scand B 1977 ; 85 : 262 - 70 .
37. Alexander C , Rietschel ET . Bacterial lipopolysaccharides and innate immunity . J Endotoxin Res 2001 ; 7 : 167 - 202 .
38. Needham BD , Carroll SM , Giles DK , Georgiou G , Whiteley M , Trent MS . Modulating the innate immune response by combinatorial engineering of endotoxin . PNAS 2013 ; 110 : 1464 - 9 .
39. Jans C , Meile L , Lacroix C , Stevens MJA . Genomics, evolution, and molecular epidemiology of the Streptococcus bovis/Streptococcus equinus complex (SBSEC). Infect Genet Evol 2015 ; 33 : 419 - 36 .
40. Jones MK , Watanabe M , Zhu S , et al. Enteric bacteria promote human and mouse norovirus infection of B cells . Science 2014 ; 346 : 755 - 9 .
41. Payne DC , Currier RL , Staat MA , et al. Epidemiologic association between FUT2 secretor status and severe rotavirus gastroenteritis in children in the United States . JAMA Pediatr 2015 ; 169 : 1040 .