Virus–Bacteria Interactions: Implications for Prevention and Control of Human Enteric Viruses from Environment to Host

Infectivity and pathogenesis

One such study by Kuss et al. (2011) explored interactions of poliovirus (PV; serotype 1, Mahoney) with bacteria and bacterial cell components. The authors found that when PV is incubated in the presence of gram-negative and gram-positive bacteria, the virus had increased viability as determined by plaque assays. The largest increase in viability was seen with the gram-positive bacterium Bacillus cereus. Further investigation revealed that B. cereus increased adherence of PV to HeLa cells, thus aiding the infection process. In addition, Kuss et al. (2011) reported an increased yield of plaque-forming units of PVs in the presence of bacterial components, such as lipopolysaccharide (LPS) and peptidoglycan (PG).

Further research was conducted based on that of Kuss et al. (2011). Robinson et al. (2014) investigated the mechanisms leading to the increase in yield of PV. More specifically, after incubation of PV with LPS, the authors observed that LPS associated with PV binds directly to the PV receptor. As a result, PV associated with bacterial cell components had an increase in attachment to the host cells. It was also discovered that only a few sites on the viral capsid—specifically, the lysine amino acid at position 99 located in the surface-exposed BC loop region of viral protein 1 (VP1)—had to bind with LPS to lead to an increase in attachment.

Another study examined murine norovirus (MNV)—a hNoV surrogate—strain types 1 and 3, and the ability of MNV to infect B-cells in the presence of enteric bacteria (Jones et al., 2014). To begin, the authors investigated whether MNV infects B-cells, which then led them to determine whether hNoV (GII.4 Sydney) also infected B-cells. Once it was established that hNoV also infected B-cells, further investigation examined what occurs with the addition of the enteric bacteria, Enterobacter cloacae. It has been shown that hNoV can bind to histo-blood group antigen (HBGA)-like structures (Harrington et al. 2004), and E. cloacae is documented to possess the H-type HGBA that allows hNoV to bind (Miura et al., 2013). The results of the study by Jones et al. (2014) revealed that enteric bacteria, such as E. cloacae, can act as cofactors to aid in the virus attachment to and infection in B-cells. These two key studies demonstrate that the interactions between viruses and specific bacterial strains have the potential to increase infectivity during in vitro studies. However, do these observed interactions and enhanced infectivity translate to the infection process in the actual host?

The authors of the two cell culture studies above did in fact use the mouse host to provide further evidence. Both treated one group of mice with antibiotics to deplete the natural microbiota in the gut, and then challenged the mice with PV or MNV. In both studies, the group of mice treated with antibiotics had a reduction in viral replication (Kuss et al., 2011; Jones et al., 2014). These results further support the idea that virus–bacteria interactions can potentially impact the infectivity of viruses in a host.

Protection and competitive exclusion

A further review of gut microbiota and viruses indicates that these principles can translate into other hosts, including humans. In a study investigating the effectiveness of a vaccine against porcine rotavirus (pRV)—an enteric virus that infects swine—the researchers first inoculated gnotobiotic pigs with either healthy or diseased children's feces and then administered the pRV vaccine (Twitchell et al., 2016). The diseased feces were obtained from children in Nicaragua who demonstrated a high enteropathy score (i.e., an indication of intestinal inflammation and poor gastrointestinal health) and had previously received the human rotavirus vaccine. Next, the pigs were challenged with infectious pRV particles, and the pigs inoculated with healthy feces demonstrated a lower incidence of infection and a stronger adaptive immunity to the pRV vaccine when compared with the pigs inoculated with diseased feces. These results indicate that the intestinal microbiota affects the infection process of the virus for better (e.g., protective) or for worse (e.g., increased susceptibility) (Twitchell et al., 2016).

In an editorial by Iturriza-Gómara and Cunliffe (2017), the authors discuss the link between the gut microbiome and the efficacy of enteric virus vaccines used in areas with elevated morbidity due to infectious disease. Iturriza-Gómara and Cunliffe highlight the findings of Harris et al. (2017) who reported significant differences in the gut microbiota of infants who responded positively versus that of infants who responded poorly to an administered RV vaccine. While the difference could be strictly due to a decrease in the immune-modulating capacity of the LPS of the more abundant bacteria in the infants with low response to the RV vaccine, another theory is possible. More specifically, because the RV vaccine contains live attenuated virus, the gut bacteria in the positive response group could be expressing HBGA or other relevant glycans that are necessary for RV cell entry and replication; thus, these bacteria help elicit an immune response and future protection from infection with wild-type RV (Harris et al., 2017).

Investigations have also explored the role probiotics may play in host protection from and/or during viral infection as observed in the Twitchell et al. (2016) study discussed previously. More specifically, Rubio-del-Campo et al. (2014) explored the interaction of hNoV (GI.1 and GII.4) P-particles—the protruding domain of the VP1 capsid protein—with lactic acid bacteria, including Lactococcus lactis and nine types of Lactobacillus sp. along with Escherichia coli strain Nissle 1917. The authors observed varying degrees of ability to bind hNoV P-particles among all eleven bacteria assayed with the best and worst binding observed for Lactobacillus casei BL23 and the gram-negative E. coli Nissle 1917, respectively.

After confirmation of bacterial cell binding, the investigators explored the effects of bacteria on the binding of hNoV GI.1 P-particles to HT-29 enterocyte cultures. These studies revealed that total inhibition of P-particle binding to HT-29 cells was achieved in the presence of high concentrations (OD₅₅₀ ≥ 0.5) of E. coli and less so with L. casei BL23. Of more interest, however, is the observation of this inhibitory effect only through competitive exclusion (i.e., simultaneous inoculation of bacteria and P-particles), and not when HT-29 cells were pretreated with bacteria or when P-particles were already attached to the cells. In the latter scenario, the addition of bacteria to the cells with P-particles already attached actually enhanced P-particle retention on the enterocytes by up to fourfold depending on the bacteria type and density. Rubio-del-Campo et al. hypothesized that during competitive exclusion hNoV GI.1 P-particle association with bacterial cells may limit binding to HT-29 cells; however, this simple association may not fully explain the inhibitory mechanism, especially in the case of E. coli Nissle 1917. It is plausible that this probiotic strain of E. coli could prevent hNoV GI.1 P-particle binding to enterocytes through a nonmicrobicidal substance as was previously shown for preventing invasion of intestinal cells by bacterial pathogen (Altenhoefer et al., 2004).

Along with Rubio-del-Campo et al. (2014), additional studies on the role of probiotics in both hNoV and its surrogates binding to host cells have been published. Li et al. (2016) investigated the effect of Bifidobacterium adolescentis against both MNV and hNoV virus-like particles (VLPs). The authors determined that B. adolescentis primarily decreased MNV replication in the murine macrophage cell line (RAW 264.7 cells) as opposed to denaturing the MNV protein capsid through lactic acid production or inhibition of host cell binding. With respect to hNoV VLPs, they observed that B. adolescentis actually did impact the binding of VLPs to the cells. More specifically, hNoV GI.1 VLP binding to Caco-2 cells was decreased significantly, whereas binding to HT-29 cells was marginally impacted. Interestingly, hNoV GII.4 VLP binding to Caco-2 cells was not impacted by the presence of B. adolescentis. Additional investigations by Shearer et al. (2014) and Aboubakr et al. (2014) also explored probiotic interactions with hNoV surrogates—specifically Tulane virus (TV) and MNV as well as feline calicivirus, respectively. However, these studies primarily consider cell-free spent media from probiotic culture for the purpose of viral inactivation.

Based on this evidence, researchers have recently considered the role gut microbiota may play in hNoV infection. Before 2016, the hNoV research community relied on surrogates and limited human volunteer studies to understand the mechanisms behind hNoV infection processes; however, Ettayebi et al. (2016) published the first evidence of reproducible hNoV replication using an ex vivo human intestinal enteroid (HIE) model. Following the lead of Jones et al. (2014) who reported MNV as well as hNoV infection of B-cells in the presence of enteric bacteria, Ettayebi et al. claimed that hNoV did not require bacterial cofactors for infection nor did LPS promote replication. However, the investigators acknowledge that hNoV replication within the HIEs varied greatly by strain type as well as HIE origin (i.e., FUT2 secretor status of the patient from which the biopsy was taken).

In the area of hNoV–bacteria interactions within this new culture model, more work is needed to characterize the individual requirements for infection of each hNoV genotype. This is especially apropos given the nearly parallel (in time) publication on the role E. cloacae plays in the shedding of hNoV by gnotobiotic pigs inoculated with the hNoV GII.4/200b variant (Lei et al., 2016). The authors support the conclusions of Ettayebi et al. (2016) with respect to enterocytes specifically being the site of infection. However, Lei et al. (2016) also reported that pigs colonized with E. cloacae inhibited hNoV infection by reducing both the concentration of hNoV in the feces and the duration of shedding compared with the control group. Similarly, Rodríguez-Díaz (2017) reported that individuals with a greater abundance of certain bacterial families—for example, Ruminococcaceae bacteria—might have lower susceptibility to infections with RV and hNoV. However, limitations linked to the sample population and the interdependency of gut microbiota composition and secretor status are not conducive to generalizability of results to the greater population.

Role in recombination

Besides directly impacting virus infectivity, virus–bacteria interactions in vivo may indirectly play a role in recombination events that viruses can undergo. Recombination happens as viruses interact with other viruses during the replication process within the host, and this allows the virus to acquire new genes (Worobey and Holmes, 1999). These newly acquired genes can lead to viral evolution and a potential increase in virulence (Bull et al., 2007). These recombination events can happen in a variety of ways and settings, including during animal production. For instance, Mattison et al. (2007) examined swine and cattle fecal samples and retail meat (raw chicken, beef, and pork) for the presence of noroviruses—both animal and human. It has been established that swine and bovine-specific NoV strains are present in these animals and can infect their respective hosts (Scipioni et al., 2008). For this reason, the authors were interested in knowing whether hNoV strains could simultaneously be present in livestock, and thus possibly cause indirect zoonotic transmission through fecal contamination of retail meat products. They reported the detection of human-like GII.4 (genogroup II, cluster 4) NoV in cattle and swine fecal samples alongside GIII (bovine) and GII.18 (swine). In addition, one raw pork meat sample tested positive for a hNoV in the GII.4 cluster. Since hNoV strains were found to be in the presence of NoV strains infectious to cattle and swine, Mattison et al. (2007) suggested the opportunity for recombination of the virus along with its new virulence factors. More recently, Sisay et al. (2016) confirmed the presence of hNoV GII.1 in collected swine fecal samples—demonstrating both zoonotic and viral evolutionary potential.

Other studies have investigated the whole virome—a collection of viruses that make up a viral community within a given ecosystem. A study by Shan et al. (2011) looked specifically at the virome of food production animals. Here, the authors explored the virome associated with the feces of healthy and diarrheic piglets on high-density farms. The majority (68%) of classified sequences in the piglets' intestines were viruses with 99% of those being mammalian RNA viruses from the families Picornaviridae (kobuviruses, enteroviruses), Astroviridae, Coronaviridae, and Caliciviridae (sapoviruses). Shan et al. posit that the level of presumed coinfection of diverse viruses observed in their study presents favorable conditions for viral recombination and viral evolution.

In the context of virus–bacteria interactions, what role do these interactions play in coinfections and the possibility of allowing accelerated viral evolution? Fortunately, an in vitro study by Erickson et al. (2018) took this step forward. The authors investigated the bacterial strains that aid in coinfection of cells, and found that, when coinfection occurs, the bacteria (1) aid in recombination events and (2) prevent deleterious mutations from occurring, ultimately causing an advantageous impact on the fitness of the virus and viral population diversity. The authors established this using PV (serotype 1 Mahoney) and 41 bacterial isolates recovered from the feces of healthy mice. Through experimental procedures, Erickson et al. were able to observe a 4.6-fold increase in recombination in the presence of coinfection aiding bacteria over the control group that contained no bacteria.

The research presented above provides evidence that virus–bacteria interactions can increase viability and virulence by allowing coinfection and recombination of viruses. There are still many questions on how these interactions affect human viruses since most research is completed using virus surrogates. These surrogates represent the human enteric virus well but do not behave completely like the human strains.

Food contact surfaces

Food products have natural microflora—some may be pathogenic, while others are naturally occurring (Wang et al., 2017). As food is processed, the ingredients and products encounter nonporous surfaces where microorganisms can be transferred resulting in adherence to the surface and possibly biofilm formation. There have been several studies that look at the development of biofilms with bacterial foodborne pathogens such as Staphylococcus aureus, Salmonella, Listeria monocytogenes, and E. coli O157:H7 (Di Bonaventura et al., 2008; Yang et al., 2009; Dourou et al., 2011; da Silva Meira et al., 2012). Because viruses do not propagate outside of a host, studies on viruses and fomite surfaces have been limited to investigating their persistence under varying conditions.

For instance, Escudero et al. (2012) examined viral persistence on food contact surfaces (stainless steel, ceramic, and formica), and reported that hNoV GI.1 (Norwalk strain), GII.2 (Snow Mountain strain), and MNV (type 1) were able to survive on surfaces for 42 days. These results have been substantiated by other researchers as reviewed by Kotwal and Cannon (2014). Unfortunately, most published studies investigate viruses in isolation as opposed to in complex microbial systems, such as biofilms, that are present in the real world. This paucity of published data related to interactions between viruses and bacteria on surfaces was also previously noted by Vasickova et al. (2010).

Recently, Schumacher et al. (2016) investigated the spread of porcine epidemic diarrhea virus (PEDV)—an animal coronavirus—within an animal food manufacturing facility. The authors reported that one batch of feed contaminated with PEDV distributed the virus to both animal and nonanimal food contact surfaces throughout the facility. Moreover, the control measures typically employed for the prevention of cross-contamination of bacterial contaminants were not adequate for the control of PEDV. While the authors did not specifically look at the interaction of PEDV with bacteria, research has shown that—once diffusion through the biofilm occurs—viruses can utilize the protective aspects of the biofilm to avoid environmental stressors (Habimana et al., 2011; Bridier et al., 2015). It can be speculated that specific associations of viruses with bacteria may allow for easier entry of virus particles into the biofilm, resulting in a reservoir of viruses that are as difficult to remove and inactivate as their bacterial counterparts (Belessi et al., 2011; Corcoran et al., 2014; Coughlan et al., 2016). However, specific studies on virus–bacteria interactions on food contact surfaces are nearly nonexistent, and it is an area that needs to be further explored.

Water resources—biofilms

Biofilms in our water conveyance systems are not a novel occurrence and have been investigated for years. A review by Skraber et al. (2005) examined how viruses in water distribution systems can cause health concerns. Another review by Wingender and Flemming (2011) discussed research on the ability of pathogenic bacteria to persist in drinking water biofilms and act as reservoirs for a variety of pathogenic microorganisms. These reviews point to similar references, such as Quignon et al. (1997). The authors of this seminal study demonstrated that viruses can incorporate into biofilms within water distribution systems. The researchers evaluated how PV-1 (Sabin strain) behaved in a water distribution system and found that the virus was always recovered at a higher percentage from the biofilms than from the water alone. The main concern within the water industry is that sloughing off of the biofilm can occur and result in pathogenic microorganisms reaching the consumer (Ashbolt, 2015). This transmission of pathogens through water to the consumer could occur either directly or indirectly; directly from drinking the contaminated water while indirectly through consumption of contaminated food products that have come in contact with the water through irrigation or processing (Lynch et al., 2009).

There are several types of microbes that have been detected in irrigation waters, and Uyttendaele et al. (2015) recently published a thorough review of irrigation water quality in the fresh produce industry. A study conducted in Belgium monitored microbes not only on the surface of the produce but also in the irrigation water of several farms (Holvoet et al., 2014). The authors found that within the irrigation water E. coli demonstrated a regular occurrence with positive detection in 75% of the samples, and that Campylobacter spp. was occasionally detected with a 30.9% presence in samples. As indicated by Holvoet et al., the prevalence of both Campylobacter and E. coli was quite high and comparable with previous reports. Of the farms that were sampled, six used open wells and two used borehole water for irrigation, and the samples were collected either from the water source or if able, from the outlet of irrigation. While this study targeted only bacteria, it provides evidence of the susceptibility of irrigation water sources to human pathogens including viruses as reported by Kokkinos et al. (2017).

Regarding viruses, Kokkinos et al. (2017) investigated the presence of enteric viruses in irrigation waters within leafy green and berry production chains in multiple countries. The researchers reported Hepatitis E virus and hNoV GII in 1 of 20 and 4 of 28 samples within leafy green production, respectively. In berry production, norovirus GII was detected in 2 of 56 samples. Here, the water samples were collected from a variety of systems in which water was most often pumped directly to the produce, while some production water sources were stored in open basins. In these instances, the contamination could be introduced through direct fecal contamination, or even association with and detachment from biofilms within the water pipes.

Pachepsky et al. (2012) focused on the effect of biofilms on aluminum irrigation pipes, and observed that the concentration of E. coli was always greater in the biofilm rather than in the water. Moreover, E. coli concentrations were higher in the sprinkler water, or irrigation output, than in the intake creek water—indicating the release of microbes from the biofilms. Given that viruses can associate with E. coli along with other bacteria within biofilms, one can speculate that viruses could enter the irrigation water just as easily as bacteria, especially if physically associated with bacteria during biofilm detachment.

Another less obvious reservoir of human pathogens in water resources used in food production are those found in freshwater sediments. Interestingly, sediments contain their own biological compartments (i.e., biofilms) and if disturbed through heavy rains, increased flow, or activities occurring within the water body, these sediments can significantly contribute to the microbial population of the water column (Pachepsky and Shelton, 2011). A study by Yakirevich et al. (2013) observed the prolonged release of E. coli after artificial high water flow events even when water levels returned to base flow, indicating continued detachment from sediments. Unfortunately, this study did not measure levels of pathogens. For some perspectives on the potential contributions of the sediments to microbial load, Pachepsky and Shelton (2011) described sediment densities of E. coli ranging from 1 to 500,000 colony-forming units or most probable number per gram of dry weight sediment from an analysis of >20 published studies.

It is also well known that viruses associate with particulates in the environment, including aquatic environments (Gerba, 1984). Although specific to coastal and estuarine sediments, Hassard et al. (2016) reviewed the reported abundance of enteric viruses in these sediments and listed levels ranging from nondetect to >6,000,000 viruses per 100 g of weight wet sediment. Research in this area has also revealed that protection from degradation is conferred to viruses when associated with sediments (Hassard et al. 2016). Therefore, it is conceivable that microbial settling and resuspension—including bacteria-associated viruses—are essential processes driving microbial contamination of freshwater, including water sources used for irrigation purposes.

Specialty crops

As with other natural environments, specialty crops such as fresh produce have their own unique microflora. Several studies have investigated the microbial diversity present on the phyllosphere—the total aboveground portions of plants—of a variety of fresh produce. Leff and Fierer (2013) observed that, while each produce type has a distinct microbial community, the majority of the microorganisms belonged to the family Enterobacteriaceae in the case of sprouts, spinach, lettuce, tomatoes, peppers, and strawberries. Meanwhile, Jackson et al. (2013) reported that Pseudomonas spp. were ubiquitous in leafy greens by both culture-dependent and culture-independent analyses.

As reviewed by Deng and Gibson (2017), numerous types of microorganisms inhabit leafy green phyllospheres, including viruses, some of which may be pathogenic to humans. Baert et al. (2011) investigated the prevalence of hNoV on a variety of fresh produce: leafy greens, red fruits, cucumbers, and tomatoes. Of 850 samples, 216 (25.4%) tested positive for hNoV (GI or GII) by real-time, reverse transcription polymerase chain reaction; however, these presumptive positives could not be confirmed through sequencing. Similarly, Stals et al. (2011) reported that 18 of 75 (24%) fruit samples tested positive for hNoV (GI and/or GII) and also could not confirm their results.

Looking beyond hNoV, Aw et al. (2016) were the first to characterize the virome of lettuce. The researchers collected samples of romaine and iceberg head lettuce from a produce distribution center, and then conducted viral metagenomic analysis. The authors observed that human and animal viruses—rotavirus and picobirnavirus, respectively—were present on the samples before retail distribution. Aw et al. also confirmed the presence of numerous viruses that require other hosts such as plants, bacteria, invertebrate, amoeba, fungi, and alga. Along these lines, the interactions of viruses with fresh produce, specifically leafy greens, in the presence of both biotic and abiotic (i.e., flooding, heat stress, mechanical stress) factors, have been investigated (Esseili et al., 2015; Gao et al. 2016). Deng and Gibson (2017) described various interactions that may be occurring, including specific binding, nonspecific binding, internalization, and microbial-assisted binding.

As discussed in the Water Resources—Biofilms section, irrigation water can transport and harbor microorganisms, and deliver them to crops; thus, irrigation waters also effect the microbiome on the surface of fresh produce. Jongman et al. (2017) applied next-generation sequencing to characterize the bacterial composition of both irrigation waters and leafy greens in South Africa by targeting the V1–V3 hypervariable region of the 16S ribosomal RNA (rRNA) gene found in prokaryotes. The authors examined several variables and determined that the bacterial microbiome of the fresh produce sample in their study was influenced by water quality, similar to the findings of Kokkinos et al. (2012). With both bacteria and viruses being present on fresh produce, this could result in interactions as previously speculated (Deng and Gibson, 2017).