Summary
Over the past century, soybean oil (SBO) consumption in the United States increased 100,000%. The main SBO fatty acid, linoleic acid (18:2), inhibits the growth of lactobacilli in vitro. Human-associated lactobacilli have declined in prevalence in Western microbiomes, but how dietary changes may have impacted their ecology is unclear. Here, we compared the in vitro and in vivo effects of 18:2 on Lactobacillus reuteri and L. johnsonii. In vitro, directed evolution led to strong 18:2 resistance with mutations in genes for lipid biosynthesis, acid stress, and the cell membrane or wall. In murine, microbiomes conditioned on a high-18:2 diet yielded L. reuteri isolates with improved resistance. Small-intestinal Lactobacillus populations in vivo were unaffected by chronic and acute 18:2 exposure, yet harbored 18:2- sensitive strains. This work shows that small intestinal beneficial microbes are protected from toxic dietary components via the gut environment and their own capacity to evolve resistance.
Introduction
During the 20th century, the greatest dietary change in the United States was in the consumption of soybean oil (SBO), which increased over 100,000%, from less than 0.001 kg/person/year to 12 kg/person/year (Blasbalg et al., 2011). Soybean oil, frequently labeled as ‘vegetable oil’, is a mixture of triglycerides composed of five long chain fatty acids (FAs), with linoleic acid (18:2) comprising over 50% of the FAs. When triglycerides are hydrolyzed by lipases active in the saliva, stomach, and upper duodenum, free FAs and monoglycerides are absorbed in the small intestine (Mansbach et al., 2000). The microbiota of the human small intestine is exposed to FAs during transit (El Aidy et al., 2015). Thus, a change in the concentration of specific dietary FAs will impact the small intestinal microbiota.
Linoleic acid and the other two major unsaturated FAs in SBO, oleic acid (18:1), and alpha-linolenic acid (18:3), are bacteriostatic and/or bactericidal to small intestinal bacteria as non-esterified (free) fatty acids in vitro at concentrations found in the small intestine (Kabara et al., 1972; Kankaanpaa et al., 2001; Kodicek, 1945; Nieman, 1954). The primary modes of killing include permeabilization of cell membranes (Greenway and Dyke, 1979) and interference with FA metabolism (Zheng et al., 2005). Affected microbes are predominantly Gram-positive bacteria including the genus Lactobacillus (Nieman, 1954). Lactobacilli are particularly important as they are considered beneficial members of the human small intestine (Walsh et al., 2008; Walter et al., 2011; Walter et al., 2007). They have been shown to be growth inhibited by the specific FAs present in SBO (Boyaval et al., 1995; De Weirdt et al., 2013; Jenkins and Courtney, 2003; Jiang et al., 1998; Kabara et al., 1972; Kankaanpaa et al., 2001; Kodicek, 1945; Raychowdhury et al., 1985). It is interesting to note that the human-associated L. reuteri underwent a population bottleneck that coincides with the increase in SBO consumption in the U.S. and is far less prevalent than it was in the past (Walter et al., 2011). Despite its decline, L. reuteri and other lactobacilli persist in the small intestine of Western individuals, suggesting mechanisms to counter the inhibitory effects of FAs in vivo.
While antibiotics can cause lasting alterations to the microbiome (David et al., 2014a; Dethlefsen et al., 2008; Dethlefsen and Relman, 2011; Jakobsson et al., 2010), dietary perturbations rarely do so (Sonnenburg et al., 2016). In humans and in mice, the gut microbiome can be quickly altered by diet but community composition generally recovers within days (Carmody et al., 2015; David et al., 2014a; David et al., 2014b; Zhang et al., 2012). Resilience to dietary perturbation may be direct, as gut microbes functionally adapt to diet, or indirect through buffering by the gut environment.
Here, we explored mechanisms of microbiome resistance to toxic dietary components with a focus on linoleic acid (18:2) toxicity to L. reuteri that was either chronic or acute in vivo and under laboratory conditions. Using an in vitro evolution assay, we observed that L. reuteri can develop 18:2 resistance upon exposure to increasing concentrations of 18:2. In mice, L. reuteri populations were unaffected by chronic dosing of 18:2 from diet and acute dosing via gavage; however, once isolated and tested in vitro, many strains were nevertheless 18:2-sensitive. The other predominant Lactobacillus in the mouse gut, L. johnsonii, behaved similarly. Together, these results show that although lactobacilli are capable of functional adaptation to chronic and acute 18:2, it is not required in the mouse gut to survive 18:2 exposure. Thus, although dietary FA toxicity is buffered in the small intestine, the ability of lactobacilli to evolve resistance constitutes an additional insurance against this potential stressor and may explain the variation in 18:2 resistance observed across strains and within human hosts.
Results
Strains of L. reuteri show variable resistance to 18:2 in vitro
To confirm the previously reported toxicity of long chain FAs towards L. reuteri, we performed disc diffusion assays with the individual free FAs of soybean oil (SBO) using the commonly studied L. reuteri ATCC 53608. We observed growth inhibition of this strain by free 18:1, 18:2, and 18:3 (Fig. 1A), and this inhibition occurred in the presence of completely hydrolyzed SBO (Fig. S1). The two saturated free FAs 16:0 and 18:0 and glycerol did not interfere with growth. To determine if the inhibitory concentration of 18:2 was comparable to concentrations in the mammalian digestive tract, we performed a cell permeability assay using propidium iodide with L. reuteri ATCC 53608 over a 10-fold dilution range from 0.01 to 1000 μg/ml of 18:2. We observed that 18:2 permeabilized the cells with an estimated inhibitory concentration 50 (IC50) of 20 μg/ml (p<0.001) (Fig. 1B). This IC50 concurs with our estimates of the concentration of 18:2 present in a mouse consuming a SBO diet (11 to 28 μg/ml for a mouse on a 7% by weight SBO diet) and with previous estimates of mammalian physiological relevant concentrations of unsaturated FAs (Kankaanpaa et al., 2001; Kodicek, 1945).
We next assessed 40 strains of L. reuteri for 18:2 resistance in liquid culture. These 40 strains were previously isolated from different sources - humans, pigs, rodents (mice, rats), birds (chicken, turkey), and sourdough, and stemmed from six different continents (Table S1) (Böcker, 1995; Oh et al., 2010). We quantified how the strains grew in 18:2 by taking the mean of the ratios for cells growing in 18:2 to cells growing in medium alone for each of the last three OD600 measurements at hours ∼4, 6, and 8 during the growth assay. L. reuteri strains have been shown to be host-specific and form host specific clades (Walter et al., 2011). The phylogenetic relationships of these strains indicate that the rodent-associated strains form clades that are basal to the other clades (Duar et al., 2017; Oh et al., 2010); therefore, if 18:2 resistance is vertically transmitted and had evolved in the human-associated strains, the rodent strains should display less 18:2 resistance than the human strains. In accord, we observed that the rodent-associated strains on average were inhibited by 18:2 more strongly than the other strains (t-test, p<0.001) (Fig. 2). However, we observed variation in L. reuteri 18:2 resistance regardless of source. For instance, the human-associated strains were not more resistant to 18:2 than the strains derived from pig, poultry, or sourdough. Moreover, within human strains, 18:2 resistance did not relate to L. reuteri isolation site (Fig. S2A). There also did not appear to be a clear relationship of 18:2 resistance with L. reuteri clades as defined by Oh et al. (2010) (Fig. S2B).
Evolved 18:2 resistance is associated with mutations in lipid-related, acid stress, and cell membrane/wall genes
To directly test if 18:2 resistance could evolve in L. reuteri through exposure to 18:2, we isolated an 18:2-sensitive L. reuteri strain (LR0) from the jejunum of a conventionally-raised mouse fed a 16% by kcal fat (low fat) SBO diet (see Methods). We seeded five cultures with LR0 and passaged them twice daily from a growth-dampening concentration of 18:2 up to a growth-inhibitory concentration over a period of 6 weeks (Fig. 3A). We also evolved five cultures of L. johnsonii strain (LJ0) obtained from the same mouse. We selected L. johnsonii based on its high abundance in mouse small intestinal microbiota (see below). At the end of the passaging regime, all of the evolved lactobacilli populations showed smaller zones of inhibition around 18:2 and 18:3 in the disc diffusion assay compared to their respective starting strains (Fig. 3B and Fig. S3). Notably, populations LR2 and LJ4 had very small to no zones of inhibition around 18:2 and 18:3. We picked an isolate from each of these populations (LR2-1 from LR2 and LJ41072 from LJ4) for testing in liquid culture supplemented with 18:2 to confirm their 18:2 resistance. We observed that LR2-1 was resistant to the maximal concentration of 18:2 used in the evolution assay, whereas LJ41072 was somewhat inhibited by this concentration (Fig. 3C and 3D). We also observed that both isolates had increased growth rates compared to their parental strains.
To characterize the mutations these populations acquired, we sequenced all five of the L. reuteri populations at the population level, four of the five L. johnsonii populations (the fifth was lost), the evolved isolates LR2-1 and LJ4107, and the starting strains LR0 and LJ0, using 300 bp paired end sequencing on an Illumina MiSeq. For the populations, we achieved approximately 500X coverage, and for the isolates, 50X coverage (Table S2). Mutations were called in the populations and isolates by aligning sequencing reads derived from the populations or isolates to the assembled genome for the respective starting strain (LR0 or LJ0). After filtering mutations appearing at a minimum frequency of 10% in a population and confirming all mutations were not due to potential mismapping in repeated regions, we observed 30 mutational events in 15 genes across the five L. reuteri populations and 35 mutational events in 21 genes in the four L. johnsonii populations. On average, each population had 6 mutations in 6 genes (Table S3, S4); exceptionally, LJ5 had 17 mutations in 17 genes. The isolates LR2-1 and LJ41074 had three and five mutations, respectively. Overall, the total number of mutations acquired by all lactobacilli populations were similar.
In each population, a few mutations had swept the entire population (Tables 1, 2, S3, and S4). Both the L. reuteri and L. johnsonii populations bore high frequency variants (>60%) in genes relating to FA metabolism, ion transport, and the cell membrane/wall. In the L. Reuteri populations, we found high frequency variants in (i) FA biosynthesis transcriptional regulator FabT (Eckhardt et al., 2013), (ii) two related tyrosine-protein kinases involved in exopolysaccharide synthesis, EpsD, and EpsC (Minic et al., 2007), (iii) an HD family hydrolase, (iv) a hypothetical protein, and (v) in the region upstream of an ammonium transporter that may respond to acid stress (Wall et al., 2007). In the L. johnsonii populations, high frequency mutations were present in (i) two distinct intracellular lipases, (ii) a putative membrane protein gene, (iii) the potassium efflux system KefA/small-conductance mechanosensitive channel, which protects against growth defects in acidic conditions (Cui and Adler, 1996; McLaggan et al., 2002), (iv) the glycosyltransferase LafA, which affects the lipid content of the cell wall and membrane (Webb et al., 2009), (v) a TetR family transcriptional regulator, and (vi) the ribonucleotide reduction protein NrdI. All but two of the above mutations are non-synonymous or cause protein truncations. The other two mutations are in intergenic regions and thus may alter the expression of the downstream gene. Although the specific genes mutated differed between L. reuteri and L. johnsonii, they are associated with similar functions, suggesting that Lactobacillus species can evolve 18:2 resistance through changes relating to lipid metabolism, acid stress, and the cell wall/membrane.
L. reuteri can survive chronic and acute 18:2 exposure in the mouse
Given that 18:2 resistance can evolve in vitro, we asked if L. reuteri and L. johnsonii could survive 18:2 exposure in vivo. To this end, we assessed the response of L. reuteri to both a chronic and an acute exposure to 18:2 in vivo. For the chronic exposure, weaned (3 week old) male C57BL/6J mice from Jackson Laboratories were fed ad libitum for 10 weeks one of two diets, wherein all of the fat was derived from SBO. The diets differed in the percentage of calories derived from fat: 16% for the “low fat” (LF) diet and 44% for the “high fat” (HF) diet (Table S5). For the acute exposure, at the end of the 10 weeks, we gavaged (delivered to the stomach) 6 mg 18:2 per gram mouse weight (e.g., double the 18:2 consumed by mice daily on the LF diet) or vehicle control (saline). At 1.5 hours post-gavage, the time at which we observed gavaged 18:2 in the blood stream (Fig. S4A), mice were sacrificed, and the small intestine contents were collected (Fig. 4A).
To assess how the gavage impacted the microbiome of the jejunum, where the bulk of fat absorption occurs (Mariadason et al., 2005), we sequenced the V4 region of 16S rRNA gene amplicons derived from one half of each jejunal luminal sample. In addition to live cells, the “total biomass” assessed in this way included cells permeabilized by 18:2 and dead cells, all of which may contain DNA that will be sequenced in this analysis. To assess the diversity of “live-only” biomass, we removed the DNA of cells with compromised membranes with the use of propidium monoazide (PMA) (Chiao et al., 2014; Exterkate et al., 2015; Nocker et al., 2006; Rogers et al., 2013) applied to the other half of each sample. This approach allowed us to gauge which taxa were still alive after the 18:2 treatment.
The effect 18:2 treatment on the microbial community was evident from analysis of the live cells but not the total biomass: microbiomes within a diet-group (HF or LF) could be distinguished by gavage treatment (18:2 or saline) only when live-only biomass was analyzed (live-only; weighted UniFrac, n = 22 for LF diet: adonis, pseudo-F=4.78, 15% of variance explained, p<0.05; n = 24 for HF diet: adonis, pseudo-F= 7.84; 28% of variance explained, p<0.005; also see Figs. S4B and S4C). Furthermore, we observed that for both diets, although the control gavage did impact the difference between live-only and total-biomass diversity, this difference was greater when 18:2 was administered (Fig. 4B). We note that this difference (beta-diversity distance) was greater for the HF than the LF diet samples (Fig. 4B), suggesting that the HF-diet conditioned microbiome was disrupted to a greater extent by 18:2 than the LF-diet microbiome.
In the total biomass, OTUs 692154 and 592160, taxonomically assigned to L. reuteri and L. johnsonii, respectively, were the two most abundant lactobacilli OTUs. These OTUs were at comparable relative abundances in the two diets (total biomass; t-test and ANOVA on a linear mixed model to include cage effects, p values>0.05, Fig. 4C and D). Note that when we detected L. reuteri OTU 692154, it was at very low levels in the microbiota of mice housed in three out of six LF diet cages and in two out of six HF diet cages (Fig. S4D). Additionally, these L. reuteri and L. johnsonii OTUs were present in the 18:2, live-only microbiota in both sets of mice (Fig. 4E and F), suggesting these taxa survived the 18:2 acute treatment regardless of the dietary fat content. Comparison of the relative abundance of these two OTUs in the total and live-only microbiota revealed these lactobacilli (with the exception of L. johnsonii in the LF diet) as enriched (2- to 4-fold, t-test, p values <0.01) in the total population after 18:2 gavage. Furthermore, the live-only microbiota of HF diet mice had an enrichment of 11 lactobacilli OTUs after 18:2 gavage (5- to 9-fold enrichment compared to control gavage, Kruskal-Wallis, FDRs <0.1 Fig. S4E). Similar enrichment of live lactobacilli after the 18:2 gavage was observed for the LF diet, although no OTU passed our significance threshold. These observations suggest that lactobacilli resist acute 18:2 exposure, particularly if they are conditioned to a high-18:2 diet.
To confirm that the Lactobacillus population was not reduced by the 18:2 gavage and that any changes in their relative abundances were due to die-offs of other susceptible bacteria, we quantified their population levels in total and live cell fractions by qPCR (see Methods). We determined the difference in the copy number of Lactobacillus 16S rRNA sequences in the total and live-only samples. This difference was normalized to the equivalent difference for total Eubacteria. We observed no difference between the saline (control) and 18:2 gavage samples for either diet (Fig. S4F, t-test, p values > 0.1). All live-only to total relative copy numbers were close to 1, as expected if the Lactobacillus population was not reduced by 18:2 exposure.
We repeated the chronic 18:2 exposure with two additional sets of mice originating from Taconic Biosciences and an F2 generation of mice from Jackson Laboratories (see Methods). In these two additional sets of mice, 16S rRNA gene sequence diversity analysis of jejunal contents showed that the same two OTUs annotated as L. reuteri and L. johnsonii were again the predominant lactobacilli. In mice from Taconic, L. reuteri and L. johnsonii were detected in the jejunum after 10 weeks on both HF and LF diets (Fig. S5A and B). In F2 Jackson mice, L. johnsonii alone was detected after 10 weeks on both HF and LF diets (Fig. S5C), whereas L. reuteri was only present in the mice on the LF diet (Figs. S5D). L. reuteri, however, was not observed in the fecal samples of any mouse at time 0 (Figs. S5E). We detected L. reuteri in LF diet-fed mouse fecal samples from weeks 5 and 10 (Figs. S5E). While it appears in the F2 Jackson mice that L. reuteri may have invaded the mice on the LF diet at a later time point, we cannot rule out the possibility of L. reuteri existing below detection. It seems unlikely that only the LF mice were exposed to L. reuteri as the mice were handled in the same manner, the diets were sterilized, and the mice were all bred in the same facility. Hence, if all cages of mice were exposed to L. reuteri then only the mice on the LF diet were capable of being invaded. Previously, Kanaanpaa et al. (2001) observed that 18:2 decreased the ability of Lactobacillus GG to adhere to Caco-2 cells, and 18:2 inhibited L. casei Shirota adhesion to human infant mucus in vitro. Hence, it is possible that a high soybean oil diet prevents L. reuteri from stably associating in a mouse, potentially preventing adhesion to the forestomach epithelium (see Discussion), but overall the results from these mice support the notion that L. reuteri and L. johnsonii populations are minimally impacted by chronic dosing of 18:2.
L. reuteri isolated from SBO diet mice are sensitive to 18:2, but HF diet isolates show increased 18:2 resistance
Our results in mice suggested that L. reuteri and L. johnsonii survived chronic and acute exposure to 18:2 either directly, by 18:2 resistance, or indirectly, through buffering by the mouse gut environment. To assess the resistance of these lactobacilli to 18:2 in vitro, we established a collection of L. reuteri and L. johnsonii isolates derived from the upper ileum of mice on both HF and LF diets. Specifically, we determined the ability of these isolates to grow in liquid culture with a concentration of 18:2 that weakly inhibited their growth. After excluding 7/120 isolates that failed grow in medium lacking 18:2, we measured growth in 18:2 for 113 isolates of L. reuteri derived from 15 mice housed in 8 of 12 cages (5 HF diet cages and 3 LF diet cages; note that L. reuteri was not detected by 16S rRNA gene amplicon sequencing in several of the cages, Fig. S4D). While most isolates were sensitive to 18:2, we observed that L. reuteri isolates recovered from the HF-diet fed mice were on average more resistant to 18:2 than L. reuteri isolated from the LF-diet fed mice (t-test, p <0.05, Fig. 4G and S6A). This observation is consistent with the hypothesis that chronic exposure to a diet high in 18:2 promotes resistance in the resident L. reuteri population.
Lactobacilli population-level 18:2 resistance in vivo does not predict the resistance of isolates in vitro
Next, we sought to relate the in vitro resistance of the L. reuteri isolates to the in vivo changes in L. reuteri populations before and after acute 18:2 exposure. To do so, we assessed the enrichment of L. reuteri OTU 692154 in the live jejunal biomass post 18:2 gavage. We compared the sequence counts for this OTU in the live-only biomass (i.e., in PMA-treated samples) in mice gavaged with 18:2 to those obtained for the same OTU in the saline gavage group (see Methods). A resulting “18:2 gavage ratio” greater than 0 indicates that live L. reuteri OTU 692154 had greater relative abundance counts in mice gavaged with 18:2 compared to same-cage controls gavaged with saline, signifying that other OTUs had been depleted. There was no correlation between the ability of these strains to grow in vitro in 18:2 and their abundance in mice gavaged with 18:2 for mice on either diet (Figs. S6B and S6C). Note, however, that we cannot exclude the possibility that the isolation procedure (growth in MRS, which contains a very small amount, 0.1 μg/ml, of 18:2) favored susceptible strains over resistant strains, such that the in vitro resistance patterns of the isolates is not representative of the whole population in vivo. With this caveat in mind, these results indicate that while chronic exposure to 18:2 can result in L. reuteri strains with higher 18:2 resistance, the mouse gut environment also buffers its toxicity, allowing susceptible strains to remain in the population.
We partially replicated these findings with L. johnsonii: as with L. reuteri, we isolated and tested the resistance of L. johnsonii strains from the ileum, where it was quite abundant (∼10% relative abundance from 16S rRNA amplicon data in all mice). Again, all isolates were sensitive to 18:2, but L. johnsonii from the HF diet-fed mice were more strongly inhibited by 18:2 than those isolated from the LF-fed mice (159 isolates from 22 mice in 12 cages; t-test, p<0.01, Fig. 4G). The results for L. johnsonii are similar to those of L. reuteri, with a lack of congruence between the response of the population in vivo and the resistance of isolates in vitro.
Putative fatty acid responsive genes are mutated in HF diet isolated L. reuteri
We sequenced to 50x coverage, an isolate of L. reuteri resistant to 18:2 and derived from a HF diet mouse (strain LRHF, Table S2, Fig. S7). Although we cannot be certain that L. reuteri from the HF diet mice shared a common ancestor with those present in the LF diet mice, we compared LRHF to LR0, the 18:2-susceptible isolate derived from a LF diet mouse and used in the in vitro evolution assay. The comparison revealed 71 mutations in 60 genes present in LRHF and absent in LR0 with functions in DNA metabolism, energy metabolism, and stress response (Table 3, Table S3). None of the genes mutated in the in vitro evolution assay were recovered as mutated in LRHF. However, relative to LR0, LRHF exhibited mutations in a sodium-hydrogen antiporter gene and a peroxide stress (PerF) gene, both of which may represent adaptation to an acidic environment caused by exposure to FAs. Of potential relevance to FA exposure, we observed mutations in a membrane-bound lytic murein transglycosylase D precursor involved in the production of the peptidoglycan layer (Vollmer et al., 2008), and the fructosyltransferase Ftf involved in the production of exopolysaccharide (Sims et al., 2011). These results suggest that, evolution in response to 18:2 exposure in vivo involves genes for acid stress and the cell wall/membrane.
Discussion
A drastic change in dietary macronutrient composition has the capacity to restructure the microbiome within a day (David et al., 2014b; Faith et al., 2011; Turnbaugh et al., 2009) and is one of the most influential contributors to microbiome composition (Carmody et al., 2015). Here, we present the first study to consider how the gut microbiome is influenced by diet from theperspective of a single FA known to be toxic to gut microbes. In this work, we investigated the interaction between lactobacilli and linoleic acid (18:2). As previously observed, we found 18:2 to inhibit the growth of most natural-derived lactobacilli in vitro. However, in the murine gut, L. reuteri and L. johnsonii persisted through both chronic and acute exposures to 18:2. L. reuteri isolates derived from mice on a diet high in 18:2 included some that were more resistant to 18:2. This observation suggests that 18:2 resistance has the potential to be selected in a host. In vitro, we showed that L. reuteri and L. johnsonii both evolve 18:2 resistance through mutations in the cell wall/membrane and fat metabolism genes. Collectively, these data suggest that the host-specific environment protects gut microbes from the inhibitory effects of FAs, but that these microbes can also evolve resistance, providing additional resilience.
In addition to being anionic surfactants, previous research demonstrated that long chain unsaturated FAs are both bacteriostatic and bactericidal (Kabara et al., 1972; Kankaanpaa et al., 2001; Kodicek, 1945; Nieman, 1954), by increasing membrane fluidity and permeability (Greenway and Dyke, 1979) potentially leading to cell lysis or leakage (Galbraith and Miller, 1973b; Parsons et al., 2012), blocking absorption of essential nutrients (Nieman, 1954), and by inhibiting FA synthesis (Zheng et al., 2005) and oxidative phosphorylation (Galbraith and Miller, 1973a). We showed through directed evolution that 18:2 susceptible lactobacilli could evolve resistance. The mutations we recovered from our 18:2 in vitro adapted lactobacilli strains are consistent with these killing mechanisms: over 60% of each of our 18:2 adapted populations had mutations affecting FA metabolism and/or the bacterial membrane/wall structure. Additionally, genes contributing to the bacterial membrane/wall structure were mutated in an 18:2 resistant L. reuteri isolate recovered from a mouse on the HF SBO diet.
Despite the toxicity of 18:2 towards lactobacilli, mouse-associated L. reuteri and L. johnsonii were present at equivalent relative abundances in mice fed diets high or low in 18:2. Moreover, these microbes survived a gavage of 18:2 equal to double what mice normally encounter in their daily diet. Our results are consistent with the findings of Holmes and colleagues, who analyzed the fecal microbiomes of mice on 25 different SBO diets varying in their macronutrient (fat, protein, carbohydrate) composition. Their results demonstrate that fat has only a minor effect on microbiome structure (Holmes et al., 2017). In contrast, in engineered microbial systems for waste processing, concentrations of linoleic acid exceeding its IC50 but within the range predicted to be consumed by animals can cause failure of the desired microbial biodegradation processes (Lalman and Bagley, 2000). Thus, the resilience of the murine small intestinal microbiome to 18:2 points to mechanisms of resistance.
How then do lactobacilli survive 18:2 in the gut? In mice, lactobacilli colonize both the small intestine and forestomach (Walter et al., 2007). While lingual lipases exist in mice (DeNigris et al., 1988), fat digestion occurs primarily the small intestine. As a result, forestomach microbes should not be exposed to a high concentration free FAs, and SBO itself is not toxic. A gavage of 18:2, on the other hand, should expose forestomach microbes to free 18:2. However, in the rodent forestomach, lactobacilli form a dense biofilm on non-mucus secreting stratified epithelial cells (Frese et al., 2013), which may protect them from free FAs.
The decline of L. reuteri in Western populations may never be fully explained. Prior to the emergence of SBO as a major dietary fat source, L. reuteri was readily isolated from human stool. Today, however, it is found in less than 10% of humans in the USA and Europe (Walter et al., 2011). Moreover, L. reuteri appears to have undergone a population bottleneck that can be dated to approximately when SBO consumption increased (Walter et al., 2011). These observations raise the question of whether a change in dietary habits drove the decline in the prevalence of L. reuteri in Western populations. In humans, L. reuteri forms neither high gastric populations nor biofilms (Frese et al., 2011; Walter, 2008), thus human L. reuteri strains may have survived increased exposure to 18:2 by developing resistance. Indeed, we did observe that some human L. reuteri strains are resistant to 18:2, but not all. While the increase in SBO consumption may have conspired with other facets of modernization to reduce the prevalence of L. reuteri in Western populations, it did not appear to have resulted in a selective sweep of 18:2 resistant L. reuteri.
The mechanistic underpinnings of how dietary components shape the composition of the gut microbiome need to be further elucidated if manipulation of the microbiome for therapeutic applications is to succeed. Dietary components have the potential to inhibit microbes directly through their toxicity, or indirectly by promoting the growth of other, more fit, microbes. While FAs are generally toxic to many lactobacilli, this work suggests that toxicity is greatly reduced when lactobacilli are host-associated. Future work in this area will elucidate how the host environment protects gut microbes from otherwise toxic dietary components, and the ways specific strains within the microbiome can be resilient to such stresses.
Author Contributions
Conception/design of study (SD, RL); acquisition, analysis, and interpretation of data (SD, RL, JJ, EK, MB, JLW, QS, PL); contributed resources (JW, JB); writing and editing manuscript (SD, RL, JW, JB).
Materials and Methods
Strains
Table S1 details the naturally derived L. reuteri strains from various hosts and countries. The L. reuteri strain (LR0) and L. johnsonii strain (LJ0) used in the in vitro 18:2 evolution assay were isolated from the jejunum contents of a mouse originally purchased from Taconic Biosciences (Hudson, NY, USA) and maintained on the low fat soybean oil diet for 6 weeks since weaning, and strain LRHF was isolated from a parallel mouse on the high fat soybean oil diet for 6 weeks since weaning (see Mouse care section for further details).
Media and culturing
Lactobacilli were cultured in MRS liquid medium (Criterion, Hardy Diagnostics, Santa Maria, CA, USA) or on MRS agar plates (Difco, BD, Sparks, MD, USA), pH-adjusted to 5.55 using glacial acetic acid. All liquid cultures and plates were incubated at 37°C in an anoxic chamber (Coy Lab Products, Grass Lake, MI, USA) supplied a gas mix of 5% H2, 20% CO2, and 75% N2.
Disc diffusions
We plated 100 μl of a dense, overnight culture of L. reuteri strain ATCC 53608 on an agar plate and applied sterile Whatman paper (Buckinghamshire, UK) discs to the surface of the culture plate. To each disc, we added 10 μl of each test compound or control. Compounds tested were alpha-linolenic acid (18:3) (≥ 99%, L2376, Sigma Aldrich, St. Louis, MO, USA), linoleic acid (18:2) (≥ 99%, L1376, Sigma Aldrich), oleic acid (18:1) (≥ 99%, O1008, Sigma Aldrich), stearic acid (18:0) (≥ 98.5%, S4751, Sigma Aldrich), palmitic acid (16:0) (≥ 99%, P0500, Sigma Aldrich), 0.85% NaCl (saline), DMSO, glycerol, all afore mentioned FAs mixed (FA mix), the FA mix with glycerol, and soybean oil (Wegmans, NY, USA). FAs were dissolved in DMSO to a concentration of 50 mg/ml, except for stearic acid, which was dissolved to a concentration of 5 mg/mL due to its lower solubility. For the FA mix, the five FAs were mixed in the ratio that these FAs are present in soybean oil: 14% 16:0, 4% 18:0, 23% 18:1, 52% 18:2, 6% 18:3. For the FA mix with glycerol, glycerol was mixed with the FA mix to a molar mass ratio of 0.1 (e.g., the molar mass ratio of glycerol in the total molar mass of soybean oil). For testing glycerol alone, the same amount of glycerol used in the FA mix with glycerol was used, and the total volume was brought up to 10 μl with DMSO. Plates were dried for 20 min at 37°C before being turned agar side up and incubated overnight.
Live/dead assay
First, we centrifuged 5 mL of an overnight culture of L. reuteri ATCC 53608 at 10,000 rcf for 10 min and resuspended the pellets in 30 mL of 0.85% NaCl solution. Then we centrifuged 1 mL aliquots of the resuspended culture at 15,000 rcf for 5 min. The resulting pellets were resuspended in 0.85% NaCl solution to a total volume of 1 mL in the presence of 18:2, 18:3, 0.85% NaCl, or ethanol. We diluted FAs in 100% ethanol in a ten-fold dilution series ranging from 0.01 to 1000 μg/ml. We incubated samples at room temperature for 90 min on a rocking platform (setting 6; VWR, Radnor, PA, USA) and inverted the samples by hand every 20 min to ensure adequate mixing. After exposure to the FA, we washed the cells by centrifuging at 15,000 rcf for 5 min, and resuspending the pellets in 1 mL 0.85% NaCl; we repeated this wash a second time. To measure the permeability of the cells, we stained samples using the Live/Dead BacLight Bacterial Viability Kit (L7007, Invitrogen, Life Technologies, Grand Island, NY, USA) according to the manufacturer’s instructions. We measured fluorescence from propidium iodide and SYTO9 on a BioTek Synergy H1 Hybrid Reader (BioTek Instruments, Inc., VT, USA). We used the drc package (Ritz et al., 2015) in R (Team, 2016) for dose-response modeling and statistical analyses.
Linoleic acid liquid growth assay
We inoculated a Lactobacillus reuteri or L. johnsonii colony grown 1 to 2 days on an MRS agar plate into a well containing 300 μl MRS liquid medium on a sterile 2 ml 96 well polypropylene plate (PlateOne, USA Scientific, FL, USA). We covered the plate with Breathe-Easy polyurethane film (USA Scientific, FL, USA) and incubated the plate overnight at 37°C in an anoxic chamber (Coy Lab Products, Grass Lake, MI, USA) supplied a gas mix of 5% H2, 20% CO2, and 75% N2. Following overnight growth, we split the cultures 100-fold into a new 96 well plate, whereby each overnight culture was diluted into a well containing MRS medium and to a well containing MRS medium plus 1 mg/ml linoleic acid. To emulsify the FA in solution, prior to and following inoculation, we vortexed the 2 ml plate on a Multi-Tube Vortexer (VWR, PA, USA) for 30 seconds at setting 3.5. We then transferred the entire plate to a 300 μl Microtest Flat Bottom non-tissue treated culture plate (Falcon, Corning, NY, USA). We measured the OD600 of the plate on a BioTek Synergy H1 Hybrid Reader (BioTek Instruments, Inc., VT, USA) at approximately 0, 2, 4, 6, and 8 hours.
We quantified how well the strain grew in 18:2 compared to the without 18:2 control by analyzing the last three time points of the growth assay. We used this approach over fitting a doubling time because, in the first few points of the growth curve, the OD values in wells with cells and 18:2 were lower than those in inoculation-control wells (e.g., with 18:2, but lacking cells). Hence, for the first few time points when subtracting the OD600 of medium with 18:2, without cells from the OD600 of medium with 18:2, with cells, we obtained negative OD600 values. As well, the time spent in log phase varied among the strains and proper modeling of log to late-log phase could not be achieved without significant trimming and manipulation of the data. At these final three time points, we determined the ratio of the “blanked OD600s” for the strain growing in MRS medium with linoleic acid to the strain growing in MRS medium alone:
We excluded time points in which the OD600 in MRS medium alone was less than 0.1 (i.e. strain did not grow). We determined the mean of the above ratios for the last three time points and refer to this mean as the “18:2 growth ratio”. All negative 18:2 growth ratios were confirmed to result from negative values in the OD600 of cells growing in 18:2.
For the naturally derived L. reuteri strains, we tested strains in triplicate to sextuplet and averaged replicate 18:2 growth ratios. For each L. reuteri and L. johnsonii strain isolated from mice on the SBO diet, we tested eight isolates from two mice per cage. Sixty-two isolates were tested between 2 and 5 times and 18:2 growth ratios were averaged across replicates. Statistical analyses were completed using the aov and t.test functions in the R stats package (Team, 2016).
In vitro evolution of 18:2 resistant lactobacilli
For L. reuteri strain LR0 and L. johnsonii strain LJ0, both originating from a mouse on the LF SBO diet for 6 weeks, we inoculated a single colony into 5 ml MRS and grew the cultures overnight. The following day, we diluted the overnight cultures for LR0 and LJ0 100-fold, separately, into five 5 ml MRS medium supplemented with 5 mg/ml 18:2. These five cultures became the five populations evolved for L. reuteri or L. johnsonii and we refer to them as LR1-5 and LJ1-5, respectively. We passaged these cultures twice daily using a 100-fold dilution. We omitted an emulsifier (DMSO or ethanol) from this assay to avoid the possibility of the lactobacilli adapting to the emulsifiers rather than to 18:2. As a result we needed to use a relatively high concentration of 18:2. To promote and maintain emulsification of the FA, we rigorously vortexed the tubes every few hours throughout the day. After seven days, we increased the concentration of 18:2 to 6 mg/ml. Each subsequent week, we increased the concentration by 1 mg/ml until reaching a final concentration of 10 mg/ml. Each week, we froze a 20% glycerol stock of each population at -80°C. We excluded L. johnsonii population #1, LJ1, from further study due to contamination.
Whole genome sequencing of Lactobacillus populations and isolates
We isolated genomic DNA from approximately 30 μl cell pellets frozen at -20°C using the Gentra Puregene Yeast/Bact. Kit (Qiagen, MD, USA). For isolates, we grew a single 50 ml log to late-log phase culture from a single colony. For populations, we inoculated five 10 ml cultures directly from glycerol stock, grew the cultures to log to late-log phase, and thoroughly mixed the replicate cultures together before pelleting to aid in representing the diversity of original population structure. We grew 18:2-adapted isolates and populations in MRS medium with 10 mg/ml 18:2, and non-adapted isolates in MRS medium. We used the Gentra Puregene Yeast/Bact. kit following the optional protocol adjustments: a 5 min incubation at 80°C following addition of the Cell Lysis Solution, a 45 min to 60 min incubation at 37°C following RNase A Solution addition, and a 60 min incubation on ice following addition of Protein Precipitation Solution. DNA was resuspended in Tris-EDTA and further purified using the Genomic DNA Clean & Concentrator™-25 (Zymo Research, CA, USA). We quantified isolated DNA using the Quant-it PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific MA, USA). Lastly, to ensure we had obtained large molecular weight DNA, we ran the DNA on a 1% sodium borate agarose gel (Agarose I, Amresco, OH, USA).
We prepared barcoded, 350 bp insert libraries using the TruSeq DNA PCR-Free Library Preparation Kit (Illumina, CA, USA). We fragmented starting genomic DNA (1.4 μg) using the recommended settings on a Covaris model S2 (Covaris, MA, USA). The barcodes used for each library are indicated in Table S2. We submitted these barcoded libraries to the Cornell University Institute of Biotechnology Resource Center Genomics Facility where they were quantified by digital PCR using a QX100 Droplet Reader (Bio-Rad Laboratories, CA, USA), pooled (Table S2), and pair-end sequenced on an Illumina MiSeq 2x300 bp platform using reagent kit V3 (Illumina, CA, USA). Resulting reads from libraries sequenced on multiple MiSeq runs were merged for further analyses.
Genome assembly of Lactobacillus populations and isolates
To generate reference genomes for the ancestor strains used in the in vitro evolution assay, we assembled paired-end sequences for L. reuteri LR0 and L. johnsonii LJ0 using SPAdes v3.7.1 (Nurk et al., 2013; Prjibelski et al., 2014) with k-mers 21, 33, 55, 77, 99, and 127 using the “careful” option to reduce mismatches and indels. To select and order contigs, we aligned the assembled genomes against the closest complete genome available: NCC 533 for L. johnsonii and TD1 for L. reuteri as determined by a whole genome alignment using nucmer in MUMmer (Kurtz et al., 2004). The assembled genomes we aligned against the NCC 533 or TD1 genome using ABACAS.1.3.1 (Assefa et al., 2009) with the “nucmer” program. Next, we aligned previously unaligned contigs using promer. We merged these sets of aligned contigs into one file and contigs with low coverage, less than 20, were removed. Finally, we ordered these filtered contigs using promer without the maxmatch option (-d) to prevent multiple reference-subject hits. For the LR0 genome, we identified a contig representing a plasmid from the assembly and included it in the set of assembled contigs. We uploaded these assembled genomes to RAST (Aziz et al., 2008; Brettin et al., 2015; Overbeek et al., 2014) for annotation (see Table S2 for details on the assembled genomes).
Variant allele detection in 18:2 resistant lactobacilli
First, we manually identified variant alleles in an isolate from L. reuteri population LR2, LR2-1, and an isolate from L. johnsonii population LJ4, LJ41072, using the Integrative Genomics Viewer (Robinson et al., 2011; Thorvaldsdottir et al., 2013). We used the variants in these isolates to calibrate the allele detection methods applied to the whole populations. Next, we identified variant alleles in the populations by aligning the paired-end sequence reads to the ancestor genome (LR0 or LJ0) using BWA-MEM (Li and Durbin, 2009). We marked duplicate sequences using Picard 2.1.1 (http://broadinstitute.github.io/picard) and utilized Genome Analysis Toolkit (GATK) (McKenna et al., 2010), and the GATK Best Practices recommendations (DePristo et al., 2011; Van der Auwera et al., 2013) to accurately select true variants. This pipeline realigns indels and recalibrates and filters base calls using the known alleles identified in the isolates using a BQSR BAQ gap open penalty of 30. We used the GATK HaplotypeCaller to call alleles with the maxReadsInRegionPerSample option set utilizing the observed coverage binned across the genome by the GATK DepthOfCoverage script. We applied the following options for populations and isolates: pcr_indel_model was set to “NONE”, stand_call_conf was set at “10”, stand_emit_conf at “4”. For populations only, we set sample_ploidy at “10” and for isolates, “1”. After we had separately processed all populations and isolates, we jointly called alleles across the entire set of populations and isolates using GenotypeGVCFs with sample_ploidy at “10”, stand_call_conf at “10”, and stand_emit_conf at “4”.
We filtered these results to remove alleles with frequencies less than 10% and to remove alleles in genes annotated with “mobile element protein”, “transposase”, “phage”, or “RNA”. In addition, the ancestor genomic reads were mapped onto the ancestor genome to aid in the removal of poorly mapping reads. We removed alleles discovered in the evolved isolates and populations that were also present at frequencies greater than 0.5 in the aligned ancestor reads against the reference. The remaining alleles we manually checked using IGV to remove any alleles in regions of the genome with abnormally high coverage, compared to the directly adjacent regions, likely representing genomic repeat regions. Filtered and unfiltered reads are presented in Tables S3 and S4.
Analysis of mutated genes
We used PredictProtein (Yachdav et al., 2014) to predict the cellular location and structure of hypothetical and putative proteins and SignalP 4.0 (Petersen et al., 2011) to predict signal peptides sequences.
Mouse experiments
All animal experimental procedures were reviewed and approved by the Institutional Animal Care and Usage Committee of Cornell University protocol 2010-0065.
Mouse soybean oil diets
The 16% and 44% SBO diets were custom designed by and purchased pelleted, irradiated, and vacuum packed from Envigo (formerly Harlan Laboratories, Inc., Madison, WI, USA, www.envigo.com). We stored open, in-use diet bags at 4°C and unopened, bags at -20°C. See Table S5 for the diet compositions.
Determination of transit time of fatty acids to the bloodstream
We gavaged nine mice with 6 mg per gram mouse weight 18:2. Every half hour following gavage, we euthanized a mouse by CO2 asphyxiation and collected blood by cardiac puncture. Blood was collected into EDTA coated tubes and stored on ice. Tubes were spun at 900 rcf at 4°C for 10 min, plasma was collected and stored at -80°C. We extracted lipids using the Bligh and Dyer method (Bligh and Dyer, 1959) and quantified FA methyl esters on a Hewlett-Packard 5890 series II gas chromatograph with a flame ionization detector (GC-FID) using H2 as the carrier. See Su et al. (1999) for further details.
Mouse care
In this study, we used three sets of male C57BL/6 mice bred in three different facilities: Jackson Laboratories (Bar Harbor, ME, USA), Taconic (Hudson, NY, USA), and an F2 generation of mice originally purchased from Jackson Laboratories. At weaning (3 weeks of age), we split littermates into cages housing up to four mice and provided the mice either the LF (16% kcal SBO) or HF (44% kcal SBO) diet (Table S5). Littermates were split so to balance mouse weights within a cage and between the two diets. All mice were housed in the Accepted Pathogen Facility for Mice at Cornell University.
In total, 24 mice were purchased directly from Jackson Laboratories and maintained in 6 cages on the LF diet and 24 mice in 6 cages on the HF diet; from Taconic, 12 mice in 3 cages on the LF diet and 12 mice in 3 cages on the HF diet; and the F2 mice from Jackson Laboratories were comprised of 11 mice in 5 cages on the LF diet and 15 mice in 5 cages on the HF diet. Up to four mice were co-caged. We stocked cages with Pure-o-cel (The Andersons, Maumee, Ohio, USA), cotton nestlets, and plastic igloos so to avoid the introduction of exogenous fat. Food was placed in the cages and not on the wire racks to minimize loss and crumb buildup of the diets as the HF SBO diet does not maintain pelleted form. Twice weekly, we completely replaced cages and food. We weighed the amount of new food provided. To obtain mouse weights, we weighed mice in plastic beakers at the same approximate time of day twice weekly. We collected fresh fecal samples once weekly from the beakers into tubes on dry ice, which were later stored at - 80°C. Mice were handled exclusively inside of a biosafety cabinet. We changed personal protective equipment and wiped all surfaces with a sterilant between cages to prevent cross-contamination. To measure food consumption, we filtered food crumbs out of the used bedding using a large hole colander followed by a fine mesh sieve, weighed the recovered food, and subtracted this amount from the known amount of food provided.
After 10 weeks on the SBO diets, we gavaged the Jackson Laboratory mice with saline (0.85% NaCl) or 18:2. The Taconic mice were gavaged with phosphate buffered saline (PBS) or 18:2, and the F2 mice from Jackson Laboratories with PBS, 18:2, or 18:3. The volume gavaged was 6 mg per gram mouse weight. The amount of FA gavaged is roughly double the amount of 18:2 consumed by mice on the LF diet each day, and more than half of the 18:2 consumed per day by mice on the HF diet. Within a cage, we gavaged half of the mice with a FA and the other half with saline/PBS, selecting which mouse received which gavage so to balance mouse weights between gavage groups. Following gavage, we moved mice to a fresh cage supplied with water, but lacking food. After 1.5 hours, we euthanized mice by decapitation and harvested small intestines contents (see below).
Processing of small intestine contents
To harvest the jejunal contents, we divided mouse small intestines into three equivalent pieces. For Jackson Laboratory mice, we flushed the middle segment, the jejunum, with 10 ml anoxic 0.85% NaCl using a blunt, 18G, 1.5 inch needle into a 15 ml conical tube that we immediately placed on ice. After flushing, we quickly shook the tube and split its contents roughly equally into a second 15 ml conical tube. One of the tubes we covered with foil to which we added 12.5 μl of propidium monoazide (PMA) (Biotium, Fremont, CA, USA; f.c. 50 μM from a 2 mM stock dissolved in DMSO). Which tube received PMA, the original or the second, we alternated between mice. To the other tube, we added 12.5 μl DMSO. To allow the PMA time to enter permeabilized cells, we placed all tubes on ice on a rocking platform for 5 min. To activate the azido group in PMA and cause DNA damage, we removed the foil from the tubes, placed the tubes horizontally on ice, and exposed the tubes for 5 min to a 650W halogen bulb (Osram 64553 C318, Danvers, MA, USA) positioned 20 cm from the samples. We frequently rotated the tubes during these 5 min to ensure equal light exposure across the whole sample. We immediately spun these tubes at 4,500 rcf for 5 min at 4°C. After we discarded the supernatant, we flash froze the tubes on liquid N2, placed them on dry ice, and later stored the tubes at -80°C. We also flushed the upper half of the last segment of the small intestine, the ileum, with MRS medium and 20% glycerol, immediately placed the glycerol stock on dry ice, which we later stored at -80°C. For the other mice, we flushed the jejunum with 10 ml anoxic PBS (pH 7.4) and did not use a PMA treatment. The small intestine contents for these mice were pelleted as described above.
DNA isolation and 16S rRNA gene sequencing
We used the PowerSoil DNA isolation kit (Mo Bio Laboratories, Carlsbad, CA, USA) to extract DNA from these jejunum pellets frozen in 2 ml tubes containing 0.1 mm glass beads (Mo Bio Laboratories, Carlsbad, CA, USA). We eluted the DNA on the spin filter using 50 μl Solution C6 and stored the DNA at -20°C. We conducted blank extractions in parallel. We processed mouse fecal pellets in a similar manner.
We quantified DNA samples and blank extractions using the Quant-it PicoGreen dsDNA Assay Kit. For each sample, we performed two 50 μl PCRs to amplify the V4 region of the 16S rRNA gene using primers 515F (f.c. 100 nM), Golay barcoded 806R (f.c. 100 nM) (Caporaso et al., 2012), 5 Prime Mix (Quanta Biosciences, CA, USA) or Classic++™ Taq DNA Polymerase Master Mix (TONBO biosciences, CA, USA), and 25 ng of DNA. PCR conditions were 94°C for 3 min, 30 cycles of 94°C for 45 seconds, 50°C for 1 min, and 72°C for 1.5 min, followed by a final extension at 72°C for 10 min. Reactions were held at 4°C and stored at -20°C.
We combined the two 50 μl PCRs and purified DNA using Mag-Bind® E-Z Pure (OMEGA Bio-tek, GA, USA) following the manufacturer’s instructions and eluting with 35 μl TE. We measured DNA concentrations using PicoGreen. We pooled 100 ng of amplicon DNA from each sample together and sequenced the pool using the Illumina MiSeq 2x250 bp platform at the Cornell Biotechnology Resource Center Genomics Facility.
16S rRNA gene amplicon analysis
We processed, filtered, and analyzed the 16S rRNA gene amplicon data from all studies using QIIME 1.9.0 (Caporaso et al., 2010). Paired-end reads were joined using join_paired_ends.py running the fastq-join method and requiring at least 200 bp of sequence overlap. Joined reads were demultiplexed using split_libraries_fastq.py requiring a Phred quality cutoff of 25 to remove ambiguous barcodes and low quality reads. Reads were clustered into operational taxonomic units (OTUs) using open-reference OTU picking at 97% sequence identity to the Greengenes database version 13.8 (DeSantis et al., 2006). We focused our analyses on the two most abundant lactobacilli OTUs: OTU 692154 identified as L. reuteri and OTU 592160 as L. johnsonii as denoted by the Greengenes assignment. We confirmed these assignments by sequencing the full 16S rRNA gene of lactobacilli isolates (see below).
Except where noted, for all subsequent analyses, we rarified data to 40,000 sequences per sample. We calculated beta-diversity using the weighted UniFrac metric implemented in QIIME 1.9.0. We performed adonis (PERMANOVA) with 10,000 iterations in QIIME using the compare_categories.py script. Beta-diversity plots were made using the phyloseq package (McMurdie and Holmes, 2013) with the ordplot function using a t-distribution. We identified OTUs differentiating samples by first filtering OTU tables to only include those OTUs present in at least 25% of samples and with at least one sample having at least 100 counts of that OTU. To the filtered OTU tables, we applied a Kruskal-Wallis test with an FDR cutoff of 10% using the group_significance.py script in QIIME. We created heatmaps of OTUs passing with FDR < 0.1 using the make_otu_heatmap.py script in QIIME. To detect L. reuteri in the fecal pellets of F2 mice from Jackson Laboratories, samples with at least 10,000 sequences were used (sequencing depth was lower for the fecal pellets), and data were not rarefied so to maximize detection of L. reuteri.
qPCR analysis of lactobacilli copy number altered by PMA treatment
We determined the copy numbers of the lactobacilli 16S rRNA gene and total Eubacterial 16S rRNA gene in the PMA and non-PMA treated jejunal aliquots by quantitative real-time PCR (qPCR) using the LightCycler 480 platform and the SYBR Green I Master kit (Roche Diagnostics Corporation, Indianapolis, IN, USA). We utilized the lactobacilli and Eubacterial primers described by Oh et al. (2012). PMA treatment reduces the total amount of DNA extracted by removing DNA from any dead cells. Thus, using the same mass of DNA for the PMA and non-PMA aliquots would result in quantifying copy numbers relative to the total amount of DNA assayed, similar to the relative abundances determined from the 16S rRNA gene sequencing. Therefore, we fixed the amount of DNA used for all non-PMA samples to 10 ng. Thus 10 μl qPCRs consisted of 10 ng of DNA for the non-PMA aliquots and equal volume for the PMA aliquot, each qPCR primer at 500 nM, and 5 μl of SYBR Green I Master mix. Cycling conditions were 5 min at 95°C followed by 45 cycles consisting of 10 s at 95°C, 20 s at 56°C for the Eubacterial primers and 61°C for the lactobacilli primers, and 30 s at 72°C after which fluorescence from SYBR Green was read. Melting curve analysis was used to determine whether each qPCR reaction generated a unique product. Cycle threshold (Ct) values were calculated using the absolute quantification/2nd derivative max function available on the LightCycler 480 software. All reactions were run in triplicate, and the mean Ct values were used in subsequent calculations.
To determine if the Lactobacillus population decreased due to the 18:2 gavage, we calculated the difference in lactobacilli copy number between the PMA (live-only biomass) and non-PMA (total biomass) aliquots relative to that for Eubacteria. That is,
If the Lactobacillus population is not affected by the 18:2 gavage, no difference should be observed between the saline and 18:2 gavage samples. Moreover, this ratio is expected to be close to one lactobacilli were not specifically killed by the 18:2 gavage.
Gavage ratio calculations
For each cage, we split the mice according to which gavage they received (18:2 or saline) and we took the mean of the rarefied sequence counts for OTU 692154 (L. reuteri). Then we calculated the log10 of the ratio of the 18:2 mean rarefied sequence counts to the mean saline relative abundance sequence counts. We refer to this ratio as the 18:2 gavage ratio:
L. reuteri and L. johnsonii isolation from small intestine contents
We streaked the glycerol stocks of mouse ileum contents onto MRS agar plates. One or two colony morphologies were present on nearly all plates: lowly abundant bright cream, round colonies present on most plates, and abundant flatter, dull white colonies present on all plates. We determined the species identity of these colony morphologies by full length 16S rRNA gene sequencing using primers 27F (f.c. 1 nM) and 1391R (f.c. 1 nM) (Turner et al., 1999), 10 μl of Classic++™ Hot Start Taq DNA Polymerase Master Mix (Tonbo Biosciences, CA, USA), and a small amount of a single bacterial colony in a 25 μl reaction. PCR conditions were 94°C for 3 min, 38 cycles of 94°C for 45 s, 50°C for 1 min, and 72°C for 1.5 min, followed by a final extension at 72°C for 10 min. We purified PCRs using Zymo DNA Clean & Concentrator™-5 (Zymo Research, CA, USA) and submitted samples to Cornell University Institute of Biotechnology Sanger sequencing facility. Returned sequences were assembled using Sequencher® version 5.4.6 (DNA sequence analysis software, Gene Codes Corporation, Ann Arbor, MI, USA, http://www.genecodes.com) and aligned against National Center for Biotechnology Institute’s nr database.
Statistical analyses
We conducted all other statistical analyses in R (Team, 2016). All t-tests are two-tailed two-sample t-tests.
Data deposition
The lactobacilli raw sequencing reads and the assembled genomes for strains LR0 and LJ0 are available under BioProject accession PRJNA376205 at National Center for Biotechnology Institute. The RAST genome annotations for these genomes are available in Supplement Files S1 and S2. The 16S rRNA gene amplicon data are available under the study accession PRJEB19690 at European Nucleotide Archive.
Table S1. L. reuteri strains isolated from various hosts. (Related to Figure 2). Host, strain name, country of origin, clade, and site of isolation on the human body (if applicable) are given as determined by Oh et al. (2010) and Zheng et al. (2015).
Table S2. Lactobacillus in vitro population sequencing. (Related to Figure 3). Tab “SequencingCoverage” gives information on the sequencing run, barcode, number of sequences obtained, and estimated genomic coverage. Tab “AncestorGenomes” gives information on the assembled LJ0 and LR0 genomes.
Tables S3. Filtered and unfiltered mutations in the L. reuteri in vitro population and HF diet isolate sequencing data. (Related to Figure 3, Table 1, and Table 3). Tab “Key” describes the information in the subsequent tabs. Tab “All” shows all variants passing filtering by GATK. Tab “Filtered” shows variants filtered to exclude alleles in genes annotated with “mobile element protein”, “transposase”, “phage”, or “RNA”, alleles at less than 10% frequency, and alleles at frequency greater than 0.5 in the aligned ancestor reads against the reference genome. Tab “Handchecked” shows variants passing previous filtering and confirmed manually in IGV. For L. reuteri two additional tabs are included: “LRHF only” shows variants only found in the L. reuteri isolate from a mouse on the HF SBO diet. “Populations only” tab shows variants only found in the in vitro evolution assay. Genomic details are taken from the RAST annotation of the ancestor genome. Other columns are taken from the GATK vcf file. For allele variants falling in intergenic regions, the surrounding genes are listed in HitGene and HitChrom, HitStrand, HitStart, HitEnd, HitDNA, and HitProtein are listed as “NA”.
Table S4. Filtered and unfiltered mutations in the L. johnsonii in vitro population sequencing data. (Related to Figure 3 and Table 2). Tabs and details are the same as for Table S3.
File S1. RAST annotation for the assembled LR0 genome. (Related to Figure 3, Table 1, and Table 3).
File S2. RAST annotation for the assembled LJ0 genome. (Related to Figure 3 and Table 2).
Acknowledgements
We thank members of the Ley lab, as well as Jiyao Zhang, Donghao Wang, Andrew Clark, the staff of the Cornell Animal Facility, Jennifer Mosher, Sylvie Allen, and Romano Miojevic for their assistance, helpful discussions, and insight. SD is an Eli & Edythe Broad Fellow of the Life Sciences Research Foundation. RL was supported by a David and Lucile Packard Fellowship in Science and Engineering, an Arnold and Mabel Beckman Foundation Fellowship, and the Max Planck Society. This work was funded by NIH grant DP2OD007444 (RL).
Abbreviations
- LF
- low fat
- HF
- high fat
- SBO
- soybean oil
- FA
- fatty acid