Gut virome in early life: origins and implications

Abstract

The human body is colonized by a multitude of bacteria, fungi, and viruses, which play important roles in health and disease. Microbial colonization during early life is thought to be a particularly important period with lasting consequences for health. Viral populations in the gut are particularly dynamic in early life before they stabilize in adulthood. The composition of the early-life virome is increasingly recognized as a determinant of disease later in life. Here, we review the development of the virome in healthy infants, as well as the role of the early-life virome in the development of disease states including diarrhea, malnutrition, and autoimmune diseases.

Introduction

The human gut hosts a diverse community of microbes, of which bacteria have been most prominently studied. The virome largely consists of viruses which infect bacteria (bacteriophage) and to a lesser extent viruses which infect other micro-organisms and human cells. Age is a major driver of variation in the composition and diversity of gut virus communities in Western individuals [1•]. Numerous factors have been reported to influence viral communities of the infant human gut including dietary, environmental, and maternal conditions. In contrast to the stable adult gut virome, the infant virome is quite dynamic [2•,3]. The period of microbial acquisition during early infancy is thought to be a critical window as these communities could predict, or possibly ordain, healthy development.

Viruses interact with each other, bacterial communities, and host cells [4]. Studies of gut viral populations associate specific patterns with various disease states. Here we review what is known regarding the development of the healthy infant gut virome and then discuss a variety of disorders that have been associated with altered viral communities. Studies on the role of the virome in inflammatory bowel disease are reviewed separately in this issue [5].

Healthy infant virome

Although the intrauterine environment was traditionally considered sterile, some studies have suggested that even healthy infants may be exposed to both bacteria and viruses prior to birth via colonization from the placenta or amniotic fluid [6,7]. However, subsequent evidence shows that the sequences detected were equivalent to those observed in negative control samples, suggesting that the observed sequences may be contamination of these extremely low microbial biomass samples [8–11]. This is supported by epifluorescent staining of stool samples, which shows that virus-like particles (VLPs) are generally absent in meconium samples but become prevalent within the first weeks of life [2•,12].

This data suggests that virus colonization of the infant gut is derived from postnatal exposures. This is supported by studies which find that a portion of phage detected in the infant gut can be mapped to identical viruses in the maternal gut [13–15] or breast milk [16,17]. However, the majority of sequences in the infant gut cannot be mapped back to these maternal sites, suggesting that the virome is likely from multiple sources including other maternal body sites, other individuals the infant has contact with, and environmental exposures. This initial phage community is largely composed of prophages induced from colonizing bacteria [2•] themselves obtained from multiple sources. Therefore, the role for a variety of contributors to virome composition is consistent with the development of bacterial communities in the infant gut, where multiple maternal body sites and environmental exposures together contribute to infant gut colonization [19,20]. Additionally, birth mode may influence virus colonization, as virome composition and diversity at one year of age differed significantly between infants delivered by Cesarean section compared to vaginally-delivered infants, confirming that very early-life exposure can influence virome assembly [18].

The diversity and composition of the healthy infant gut virome changes rapidly in the first years of life. This includes highly variable VLP abundances in the first two years of life, before stabilization in later childhood [21•]. In the first months of life, the gut virome is dominated by phage [22•]. Early in infancy, phage in the order Caudovirales predominate [2•,23,24], but this order decreases as phage in the family Microviridae increase over the first two years of life [3]. CrAss-like phages, which are abundant in adults [25], are rarely detected in the first months of life, but become more prevalent a few months after birth [2•,3]. These changes likely reflect the acquisition and changing proportions of host bacterial species with age [26•], as specific phage co-occur with their bioinformatically-predicted host species [27]. Changes in phage populations can also affect bacterial communities via phage predation [28,29], although most phage in the infant virome are predicted to be primarily lysogenic [2•,21,27], and may influence bacterial populations via phage-encoded genes in their host bacteria [26•].

Although phages are far more prevalent than eukaryotic viruses in the healthy infants, the latter population increases in abundance and diversity throughout development. Anelloviruses, which are commonly detected at many body sites in healthy adults [30], are rare early in life but increase in abundance with a peak between 6 and 12 months of age [2•,3]. Eukaryotic DNA viruses (i.e. viruses that infect eukaryotic cells) such as circoviruses [3,14], as well as RNA viruses such as enteroviruses [2•,3,31], parechoviruses [3,31], and picornaviruses [2•,3], and plant viruses such as members of the family Virgaviridae [23], are frequently found in infants, again mirroring populations commonly detected in the gastrointestinal tract of healthy adults [4]. Healthy infants also often shed potentially pathogenic viruses, such as rotaviruses, noroviruses, and sapoviruses [23,31–33], even in the absence of symptoms, although infection with these viruses may be inhibited by breastfeeding [2•] and increases with the start of daycare attendance [27], consistent with an ongoing role for environmental exposures in altering virome composition in children.

There is now an emerging literature demonstrating high inter-individual diversity of the gut virome and some patterns of typical gut virome assembly, defined by a rapid colonization of phage populations induced from bacterial microbiota, followed by colonization with eukaryotic viruses. Although the role of some factors, such as birth mode and breastfeeding, have begun to be defined, many questions remain with regards to other influences on the infant gut virome. For example, there is relatively little understanding on the role of geography and urbanization, antibiotics or other medications, the built environment, or host genetics and immune function on infant virome composition. Future longitudinal studies that collect relevant metadata from participants from a variety of backgrounds will enable a better understanding of the dynamics of infant gut virome assembly and implicate factors that may influence its development.

Virome in disease

Diarrhea

Diarrhea is a leading cause of death in children less than five years old, leading to about 500 000 deaths per year [34]. Although eukaryotic viruses such as rotavirus, norovirus, and adenoviruses play a key etiologic role in diarrhea [35], a significant portion of acute diarrhea is of unknown etiology [36]. Metagenomic studies of diarrheal samples characterize the entire viral population in a sample, permitting the identification of both known and previously unknown RNA and DNA viruses [37]. Examples of novel viruses detected in cases of diarrhea include astrovirus (MLB 1–3 and VA/HMO 1–5) [38–43], cosavirus [44,45], saffold virus [46], klasse/salivirus [47–49], polyomavirus (MW and STL) [50–52], bufavirus [53], tusavirus [54], bocavirus [55–57], recovirus [58], gyrovirus [59], pecovirus [60,61], and smacovirus [62]. However, many viruses can be detected in combination with one another, making it difficult to attribute causality to one or the other agent [63–65]. It should be noted that although these viruses were found in children with diarrhea, it remains to be determined whether they are truly agents of diarrhea or not, as many have also been detected in stool from asymptomatic children [36,37].

Multiple studies report a diversity of both eukaryotic viruses and phage in stool from people with diarrhea. A study of young Ethiopian children found that eukaryotic viral richness is positively correlated with increasing watery consistency [66]. A study examining the differences in eukaryotic viral communities in children with diarrhea in different geographic areas found that children from Northern Territory, Australia had more viral families per sample than samples from Melbourne [67]. This highlights the potential importance of environmental factors on the composition of the stool virome. However, Yinda et al. performed a viral metagenomic study of human stools, largely from individuals with gastroenteritis with varying degrees of bat contact, to assess if any viruses of animal origin could be identified. There was no evidence of interspecies transmission of the common diarrhea-associated viruses [68].

Malnutrition/enteropathy

Childhood malnutrition is an important cause of increased morbidity and mortality worldwide [69]. Severe acute malnutrition can present as either wasting (marasmus) or kwashiorkor, which is characterized by generalized edema and hepatic steatosis from insufficient protein intake. A comparative metagenomic study of the fecal DNA viromes of healthy twin pairs during the first 3 years of life and twins who became discordant for kwashiorkor or marasmus, plus their mothers and older siblings, identified age discriminatory phage and eukaryotic viruses [70]. Additionally, viral contigs could discriminate discordant from concordant healthy pairs. These data suggest that viromes are associated with, but perhaps not mediators of, malnutrition.

Stunting, defined as height for age greater than two standard deviations below the mean, is strongly associated with cognitive deficits, poor school performance, and lower economic productivity [71]. Stunting is likely a multifactorial condition caused by poor nutrition, repeated infections, and gut inflammation. Two recent studies identified specific bacteriophages associated with childhood stunting, as well as differences in the relationship between bacteriophage and bacterial communities in children with poor linear growth [21•,72].

Type 1 diabetes

Type 1 diabetes (T1D) is an autoimmune disease resulting from the destruction of insulin-secreting pancreatic β-cells. T1D is preceded by islet autoimmunity, and characterized by autoantibodies targeted against pancreatic autoantigens [73]. Enterovirus infection has long been considered a potential trigger for the development of disease in genetically susceptible individuals [74]. Although evidence exists both in favor of and against that hypothesis [75], a metanalysis of 24 studies showed a significant association between enterovirus infections, detected by targeted molecular methods, and the development of islet autoimmunity and T1D [76]. Human β-cells express receptors sufficient for enterovirus infection [77], and enteroviruses have been identified in pancreatic tissues of recently diagnosed T1D patients [78]. Although the mechanism of this interaction has not been fully elucidated, studies performed in vitro and in mice suggest that enteroviruses can diminish the β-cell insulin response and trigger hyperglycemia [79,80].

Early studies depended on targeted methods to test the link between enteroviruses and T1D, but high-throughput next-generation sequencing enables an untargeted characterization of the virome before onset of T1D. As such, a variety of prospective studies have utilized virome sequencing on fecal samples from children at increased genetic risk of T1D to assess whether virome changes precede islet autoantibody or T1D development [81•]. Although an early study performing next-generation sequencing failed to detect any differences between children who developed islet autoimmunity and controls [82], analysis of other birth cohorts identified a variety of virus families that were discriminatory based on disease status [83–85]. These metagenomic analyses in part recapitulated earlier findings, with different enterovirus types enriched in cases versus controls [84,85]. Intriguingly, prolonged shedding with enteroviruses, rather than short-term infections, is associated with islet autoimmunity [85]. However, these prospective cohort studies also identified additional viruses differentially detected by disease status, including members of the Circoviridae family and human mastadenovirus F, as well as a variety of bacteriophage families, which may warrant additional pursuit [83–85]. Altogether, these untargeted studies support the link between enteroviruses and T1D development, but additionally suggest that the dynamics of infection, as well as additional virus families, may have a role in disease outcome.

Celiac disease

Celiac disease is an autoimmune small intestinal enteropathy in response to dietary gluten, associated with autoantibodies targeted against tissue transglutaminase. Patients diagnosed with celiac disease possess high-risk HLA haplotypes, but this genetic risk factor alone is insufficient for disease presentation, as a substantial portion of healthy individuals carry these same alleles [86]. A variety of environmental factors, including virus exposure, have been studied in the context of disease development. A higher incidence of early-life infections of any type is associated with the subsequent diagnosis of celiac disease [87,88]. Infection with individual enteric viruses has been implicated by targeted molecular or serological methods, including enteroviruses [89] and rotaviruses [90,91]. Although the mechanistic link between viral infection and celiac disease has not been established in humans, a murine model suggests that viral infection promotes an inflammatory response and breaks oral tolerance to gluten in genetically susceptible mice [92].

Metagenomic sequencing of stool samples from a prospective cohort study has allowed for the untargeted analysis of viral composition prior to celiac disease diagnosis. Interestingly, early-life enterovirus colonization again appeared to be associated with the subsequent development of autoimmunity, with enteroviruses more prevalent in 1–2-year-old children who later developed celiac disease versus those that remained healthy [93]. The association between enterovirus colonization and later celiac disease was not evident in stool samples collected from children 6 months to 1 year old [93], suggesting that infections that occur during breast feeding do not predispose to later celiac disease, a finding also reported in a previous cohort study [89]. In contrast to the differences in virome composition before celiac disease diagnosis, there were minimal differences between the gut eukaryotic virome of patients with celiac disease after diagnosis and healthy children, with most differences in virome composition being driven by phage [94]. These studies together suggest that exposure to eukaryotic viruses in early life plays a role in the development of disease, although the viruses may no longer be detectable at the time of diagnosis.

Conclusions

The human early-life virome appears to have a broad role in future health and disease. Although specific gut viruses are well-recognized for their roles in diseases such as gastroenteritis, metagenomic sequencing has enabled the identification of additional viral agents that might contribute to symptomatic infection. However, as even healthy children shed a large number of viruses in their stool, carefully defined case-control studies are necessary to define whether these viruses are truly etiological agents.

A key challenge in implicating the infant virome in specific diseases is that many human studies are primarily designed to correlate virome composition with disease outcomes, rather than to demonstrate causality. To move from correlation to mechanistic studies is particularly challenging, as many viruses do not have established culturing conditions and, in the case of phage, the host bacterial species may not be known.

Currently, there are scant experimental systems to manipulate the virome and assess consequent microbiome and biological impacts. Gnotobiotic mouse models, in which mice are colonized with a known set of bacteria and/or viruses, have previously been used to study the effects of specific viruses on the host. These models have been used to show how individual viruses alter intestinal immune responses and gene expression [95] and specific bacteriophage affect the gut metabolome and metabolite levels, which can in turn affect host health [96]. Introducing viruses to gnotobiotic neonatal mice may make it possible to mechanistically link early-life exposure to specific viruses to later health outcomes. However, these models cannot account for the full complexity of the microbiota interacting with the introduced viruses. To overcome the limitation of uncultivatable virus, filtered fecal transplants, in which the entire viral portion of a fecal sample is isolated and provided to another host, have been studied. Altering the virome in this fashion leads to improved diabetes outcome in mice [97], reduced necrotizing enterocolitis in piglets [98], and reduced Clostridium difficile symptoms in humans [99]. These studies allow for the analysis of a complex virome, although implicating a specific virus may be more challenging in this context.

Nevertheless, a variety of prospective longitudinal cohort studies have helped to define the role of the early-life virome in the later development of disease. This area of research is particularly exciting, as it suggests the possibility for therapeutic intervention in early life. Further research to define the virome in healthy infants, as well as in those who go on to develop disease, will be essential to understand the viral communities that are central to promoting health and the factors that influence them.