Impact of long-term dietary habits on the human gut resistome in the Dutch population

The human gut microbiome plays a central role in health and disease. Environmental factors, such as lifestyle and diet, are known to shape the gut microbiome as well as the reservoir of resistance genes that these microbes harbour; the resistome. In this study we assessed whether long-term dietary habits within a single geographical region (the Netherlands) impact the human gut resistome. Faecal samples from Dutch omnivores, pescatarians, vegetarians and vegans were analysed by metagenomic shotgun sequencing (MSS) (n = 149) and resistome capture sequencing approach (ResCap) (n = 64). Among all diet groups, 119 and 145 unique antibiotic resistance genes (ARGs) were detected by MSS or ResCap, respectively. Five or fifteen ARGs were shared between all diet groups, based on MSS and ResCap, respectively. The total number of detected ARGs by MSS or ResCap was not significantly different between the groups. MSS also revealed that vegans have a distinct microbiome composition, compared to other diet groups. Vegans had a lower abundance of Streptococcus thermophilus and Lactococcus lactis compared to pescatarians and a lower abundance of S. thermophilus when compared to omnivores. In summary, our study showed that long-term dietary habits are not associated with a specific resistome signature.

www.nature.com/scientificreports/ The human gut microbiome is also an important reservoir of antibiotic resistance genes (ARGs) [21][22][23][24] . It is therefore important to understand how long-term dietary habits not only impact the microbiome composition, but also the composition of the total of ARGs, the resistome, in the human gut. Advances in high-throughput sequencing have allowed in depth studies of the human gut resistome. The gut resistome of healthy humans can typically contain over 100 unique ARGs, with the most abundant ARGs being those that encode for tetracycline resistance, followed by macrolide and beta lactam resistance genes 25,26 . Just as with the microbiome, several factors are known to alter the resistome composition. Orally administered antibiotics are known to select for bacteria that are resistant to these antibiotics, therefore increasing the abundance of these ARGs in the gut 27,28 . This antibiotic-driven enrichment can take place in either the general population or more specific, in clinical settings, where effects on the resistome were observed in a matter of days [29][30][31] . Additionally, the human gut resistome can be affected by environmental factors such as international travel and living conditions. This has been shown in studies comparing urbanized with agricultural populations, in which the use of antibiotics plays an important role 32,33 . Finally, meat contaminated with bacteria carrying ARGs as a result of antibiotic usage in livestock, has been highlighted as a possible transmission route for resistant bacteria and could therefore influence the gut resistome as well [34][35][36][37][38] . More specifically, zoonotic pathogens such as species of Salmonella and Campylobacter and certain types of Escherichia coli (e.g. Shiga-toxin-producing E. coli (STEC)) are known for causing foodborne infections and are frequent carriers of ARGs [39][40][41][42] .
In this study we assessed whether long-term dietary habits within a single geographical region impact the human gut resistome in the general Dutch population. Using metagenomic shotgun sequencing (MSS), we were able to detect 877 unique bacterial species and an extensive resistome composed of 119 unique ARGs, in the gut microbiome of healthy Dutch residents. resistome capture sequencing approach (ResCap) was applied for a subset of samples and revealed 145 unique ARGs, thereby surpassing the detection limits of MSS. Despite the high resolution of the sequencing data, the total number of ARGs detected by MSS or ResCap per diet group was not significantly different in the general Dutch population.

Results
Diet-associated differences in the gut microbiome. Before determining diet-associated resistome differences, we first assessed the effect of long-term dietary habits on the gut microbiome, as this ecological niche is an important reservoir of antibiotic resistance genes. Faecal samples from 149 Dutch individuals were selected based on their dietary habits and categorized in four matched diet groups: (1) omnivores, (2) pescatarians, (3) vegetarians and (4) vegans (Table 1). Faecal samples from these four groups were used for metagenomic shotgun sequencing (MSS) in order to study the effect of dietary habits on the gut microbiome. The mOTUs2based taxonomic binning method revealed that the top 10 most abundant genera did not differ between the diet groups ( Fig. 1a). In all our study groups, the most abundant taxa belong to the order of Clostridiales and genera Faecalibacterium, Bacteroides, Clostridium and Prevotella, which matches with the gut microbiome composition observed in previous studies 43,44 . In addition, the gut microbiome diversity, expressed as Shannon index, was not significantly different between diet groups, indicating that the total species diversity is highly similar among diet groups (Fig. 1b). Statistical analysis of the inter-diet group beta diversity based principal component analysis (PCA) using Aitchison distance, Bray-Curtis distance, or Jaccard distance further revealed that diet was not a main driver of the observed variance in microbiome composition between the samples (figure s1).
In addition to mOTUs2 we also applied MetaPhlAn3 to profile the microbiome composition. In concordance with the results of the mOTUs2, using MetaPhlAn3 we did not observe differences in the top 10 most abundant genera between the diet groups (figure s2). Using MetaPhlAn3 for differential abundance analysis, we also found a lower relative abundance of R. torques, S. thermophilus and C. saccharolyticum in vegans compared to omnivores and higher abundance of E. eligens in pescatarians compared to omnivores (figure s3). Furthermore, the MetaPhlAn3 approach detected a higher relative abundance of Lactobacillus delbrueckii, Coprococcus comes, www.nature.com/scientificreports/ ANCOM-BC analysis for differential relative abundance of bacterial species between diet groups. Abundance was plotted on the relative abundance scale from 0.00 to 1.00. Adjusted p values are indicated by *< 0.05, **< 0.01 and ***< 0.001. Composition of the gut resistome across diet groups based on metagenomic shotgun sequencing. We next set out to investigate whether long-term dietary habits impacted the gut resistome composition.
Using MSS we were able to identify 119 unique antibiotic resistance genes (ARGs) among all diet groups. The total number of detected ARGs was not significantly different between the groups, with an average of 17 ± 4 genes found in omnivores, 16 ± 5 in pescatarians, 17 ± 4 in vegetarians and 17 ± 5 in vegans (Fig. 2a). Among these, five were consistently detected in all diet groups (detected in 95% of the samples), namely the aminoglycoside resistance gene ant(6)-Ia, the macrolide-lincosamide-streptogramin B resistance gene erm(B), and the tetracycline resistance genes tet (40), tet(Q) and tet(W) (Fig. 2b). The most abundant genes detected encode for resistance to the classes of tetracyclines, macrolides, betalactams, aminoglycosides and phenicols. There were no differences when comparing the overall top 10 most abundant ARGs between the diet groups (Fig. 3a). Also, the resistome diversity, expressed by Shannon index, was not significantly different between the diet groups (Fig. 3b). In addition, diet does not seem to be the main driver of the observed variance in resistome composition between the samples when analysed by beta diversity based principal component analysis (PCA, figure s4). Supervised analysis using ANCOM-BC only revealed a significant difference of the abundance of tet(X), which is present in low abundance in pescatarians in comparison to omnivores (p adj = 4.4E−02) and vegetarians (p adj = 1.0E−02) (Fig. 3c). Overall, based on MSS, the resistome was highly similar among the studied diet groups.
Higher antibiotic resistance gene detection sensitivity of ResCap compared to metagenomic shotgun sequencing. Although MSS was able to detect a large variety of ARGs, these only represented 0.06 ± 0.03% of the total number of reads. This indicates that the proportion of ARGs is relatively low when compared to the total gene pool present in the samples. In order to improve the sensitivity to detect ARGs, we applied the probe based resistome capture sequencing approach (ResCap), on a subset (64/149; 43%) of samples ( Table 2) 46 . ResCap was able to greatly enrich the number ARGs specific reads, with 40.5 ± 15.2% of the total number of reads sequenced mapping to ARGs.
To evaluate the sensitivity of ResCap compared to MSS, we compared the observed number of ARGs identified per sequencing depth, by using rarefaction curves in the same 16 samples per diet group, subjected to both MSS and ResCap. Overall, ResCap was able to detect a higher number of ARGs than MSS. Even at 70 M reads per sample, MSS did not reach the level of sensitivity that ResCap was able to achieve (figure s5a). Where MSS detected 16 ± 5 to 18 ± 4 ARGs per diet group with a sequencing depth of 70 M reads, ResCap resulted in the detection of 33 ± 8 to 39 ± 7 ARGs per diet group with a sequencing depth of 20 M reads (figure s5b).
The total number of ARGs detected by ResCap per diet group was not significantly different, with an average of 36 ± 5 genes found in omnivores, 32 ± 9 in pescatarians, 40 ± 10 in vegetarians and 40 ± 10 in vegans (figure s5b). The most abundant ARGs identified encode for tetracycline, macrolide, beta-lactam and aminoglycoside resistance (Fig. 5a). No differences were observed when comparing the overall top 10 most abundant ARG. No significant differences were observed when comparing the resistome diversity as calculated by Shannon index (Fig. 5b). Beta-diversity based PCA further confirmed that diet is not a main driver of the observed variance in resistome composition between the samples (figure s6). Supervised analysis using ANCOM-BC revealed that the abundance of two ARGs, lsa(C) and tet(L), were significantly different between the diet groups pescatarians and omnivores (Fig. 5c). However, these genes are present in very low abundance (median below 0.05%) and in low prevalence (< 15%). Based on these results we can conclude that differences in dietary habits did not result in significant differences in resistome composition.

Discussion
By using the combined power of metagenomic shotgun sequencing (MSS) and ResCap we were able to detect over 850 different bacterial species and more than 100 unique antibiotic resistance genes (ARGs) in the gut microbiome of participants from the general Dutch population with distinct dietary habits. Our results show that long-term dietary habits are not associated with specific resistome signatures.
To investigate the gut resistome in humans, we used two techniques, MSS and ResCap. Using genomic DNA from human faecal samples, Lanza   www.nature.com/scientificreports/ www.nature.com/scientificreports/ in the probe library, including nitroimidazole resistance genes nimA, nimB, nimC, nimD, nimE, nimF, nimH and nimJ (table s1). We currently have no explanation why we were not able to detect the 14 genes for which probes were present in the ResCap library and that were detected by MSS but not by ResCap. We observed limited differences in the resistome composition between the diet groups, independently of the detection approach. A previous MSS based comparison of the gut resistome in the Chinese, Danish and Spanish general population, revealed the overall high abundance of ARGs that confer resistance to tetracycline, followed www.nature.com/scientificreports/ by macrolide, β-lactam and aminoglycoside resistance 26 . This is also observed in our study which included samples from only Dutch population. When using ResCap as high-resolution method to profile the gut resistome, only two ARGs (lsa(C) and tet(L)) were differentially abundant between omnivores and pescatarians. Since these genes are observed in low abundance and prevalence, the biological relevance of this finding is unclear. MSS revealed that tet(X) was more abundant in the resistome of omnivores compared to pescatarians. It is most likely that the increased sample size (n = 149 vs n = 64 samples included for MSS and ResCap respectively) resulted in the detection of differentially abundant tet(X) by MSS, but not by ResCap. This suggests that the small sample size for ResCap is a potential limitation of the current study. Two methods, mOTUs2 and MetaPhlAn3 were used to study the impact of dietary habits on the gut microbiome 49,50 . This showed that long-term dietary habits did not result in significant differences in the top 10 most abundant genera, nor in differences in gut microbiome diversity and that diet was not a main driver of the observed variance in microbiome composition between the samples. Both methods were largely in concordance regarding which genera were differentially abundant using a supervised analysis where only a few compositional differences between the microbiome of omnivores, pescatarians and vegetarians were detected. Variation between methods include the detection of differentially abundant Clostridium species by the mOTUs2, compared to differentially abundant species from the family of Lachnospiraceae, by the MetaPhlAn3. This variation could be attributed to the distinct ways in which the tools correct for genomic sequences from yet unknown species. While mOTUs2 makes use of marker genes, using its clusters of orthologues groups of proteins approach (COGs), MetaPhlAn3 instead estimates the 'unknown' portion by using the average gene length and genome length.
As part of a short-term diet intervention study, David et al. associated the consumption of animal derived products with an increased relative abundance of Alistipes, Bilophila and Bacteroides and a decrease in the relative abundance of Roseburia, Eubacterium rectale and Ruminococcus bromii 14 . Although our study included similar diet groups with omnivores consuming animal derived products, and vegans that do not, we did not observe similar findings. When compared to pescatarians, we detected a lower relative abundance of Eubacterium in omnivores, namely Eubacterium eligens. Chung et al. described E. eligens to be a pectin degrading specialist, which is a major component of plant cell walls and increased plant consumption might therefore explain the lower abundance of E. eligens in omnivores 51 . The non-concordance between the studies by David et al. and our study might be explained by differences in resolution of 16S rRNA gene sequencing used in David et al., as compared to MSS in our study, but maybe more importantly this illustrates the difference between an interventionbased study compared to our study design, where long-term diet habits were studied.
More recently, De Angelis et al. compared the gut microbiome, proteome and metabolome of Italian omnivores, vegetarians and vegans based on 16S rRNA gene profiling. Similar to our study, they showed only slight differences in the abundance of bacterial families between the diet groups. Ruminococcaceae were found to be most abundant in omnivores, while Lachnospira were associated with vegans and vegetarians 52 . Similarly, Losasso et al. observed differences in the relative abundance of Ruminococcaceae when comparing the microbiome of vegans, vegetarians, and omnivores 20 , using 16S rRNA profiling. Although the observations are limited to the family level, these studies align with our results, where omnivores were found to have a higher relative abundance of R. torques when compared to vegans. In contrast to these dietary studies, we observed differences in the relative abundance of lactic acid bacteria when comparing the microbiome between diet groups. This difference could be explained by the relatively high consumption of dairy products in the Netherlands 53 . Omnivores and pescatarians showed higher relative abundance of Streptococcus thermophilus when compared to vegans. Similarly, Lactococcus lactis was present in higher relative abundance in the microbiome of pescatarians compared to vegans. Both S. thermophilus as L. lactis have been associated with the consumption of dairy products 54 . Zhernakova et al. also detected a specific association between the consumption of buttermilk and the abundance of Leuconostoc mesenteroides and L. lactis, when assessing factors contributing to variation of the gut microbiome composition in the Dutch and Belgium population 12,55 . Since dairy products are excluded from the vegan diet, it could explain the low abundance of these lactic acid bacteria in this group.
To conclude, using ResCap we were able to increasingly detect resistance genes in complex samples like human faeces in our study, compared to MSS, and found that the gut microbiome of humans included in this study represented a large reservoir of 145 different antibiotic resistance genes in the studied population. However, differences in long-term dietary habits in the Dutch population did not result in significant differences in resistome composition between the four diet groups. When comparing the microbiome composition of diet groups in the Dutch population, mainly the vegan diet was associated with a distinct taxonomic composition. Since geographic location likely has an impact on the microbiome and resistome composition in the general population, future studies involving human populations from other geographic regions, are needed to determine the generalizability of our findings in the Dutch population.

Methods
Participant inclusion. In this study, faecal samples from 149 Dutch individuals from a previous study in the general population (the "NLD-VEGA-study") were selected based on their dietary habits (Table 1) and categorized in four different diet groups: (1) omnivores, (2) pescatarians, (3) vegetarians and (4) vegans 56 . Groups were matched for sex, age, education level, medication usage and keeping animals ( Table 1). Differences between groups were compared by the Kruskal-Wallis or X 2 test where appropriate. All participants had a Dutch nationality, were born in the Netherlands and lived in urban areas. None of them had used antibiotics, insulin, proton pump inhibitors or drugs related to cancer treatment and chemotherapy in the past 3 months. www.nature.com/scientificreports/ Ethical approval. The original "NLD-VEGA-study" protocol was approved by the medical ethics committee of the University Medical Center Utrecht, the Netherlands (no. 15-561/C). All experiments were performed in accordance with relevant guidelines and regulations.
Classification of diet groups. Diet groups were defined based on the following criteria: omnivores consumed meat at least three times per week, for the past 6 months or more. Pescatarians consumed fish and animal derived products but did not consume meat in the past 6 months or more. Vegetarians consumed animal derived products but did not consume meat or fish in the past 6 months or more. Vegans did not consume meat, fish, or animal derived products in the past 6 months or more. Furthermore, pescatarians, vegetarians and vegans did not prepare the restricted products described for house members or pets or have house members that consumed those.
Sample collection, storage and DNA extraction. Faecal samples were sent by regular mail and transported for a maximum of 24 h before storage at − 80 °C. Samples were divided into aliquots of 0.2 g, thereby introducing one freeze-thaw cycle. Samples were thawed a second time and used for DNA extraction, using a modified protocol of the QIAamp fast DNA stool mini kit (Qiagen, Venlo, the Netherlands) as described by Knudsen et al. 57 Table 2). Differences between groups were tested by the Kruskal-Wallis or X 2 test where appropriate. ResCap was performed according to the supplied protocol. In brief, 0.8-1.0 µg DNA was used for fragmentation using the KAPA HyperPlus Kit v4.17 (Roche, Woerden, The Netherlands) to generate 400 bp fragments. End repair, A-tailing and adapter ligation were performed as described by the SeqCap EZ HyperCap User's Guide v2.3. Pools of 12 samples were used for hybridization and capture using an extended version of the ResCap probe collection as described in the original publication 46 . Sample pools were sequenced on a NextSeq500 (Illumina), using high output and paired-ends of 2 × 150 bp.
Resistome data processing. ResCap and MSS data were trimmed using Trim Galore version 0.6.4 with standard settings 59 . KMA version 1.3.4 was used to align sequences to the Resfinder database version of 2020-06-02 60 . For KMA, paired-end reads were used as input by using -ipe, together with the options: -tmp, -1t1, -cge, -apm p, -ef. The resulting list of detected genes and their abundance was trimmed by applying a cut-off of 90% identity (called query Identity) and of 80% coverage (called template coverage). Finally, the output value depth was used for subsequent analysis, which represents the amount of aligned base pairs, while correcting for gene length. Analysis of the MSS negative control for antibiotic resistance genes, containing less than 300 k reads, revealed the low abundance of antibiotic resistance genes tet(L) and aadD.
Data analysis. Analysis of sequencing data was performed in R version 4.0 and the functions of the packages phyloseq and ggplot2 [61][62][63] . The top 10 abundances of the microbiome and resistome were plotted using aggregate top taxa and plotting functions of the microbiome package 64 . Shannon index was calculated using the alpha diversity functions of the microbiome package and plotting functions of the microbiomeutilities package 65 . Aitchison, Bray-Curtis and Jaccard distance PCA were generated using the transform function of the microbiome package and ordinate (RDA) and plot ordination functions of the phyloseq package. Correlations of sample dissimilarity and diets were tested using PERMANOVA with the adonis function at 999 permutations of the vegan package 66 . Basic accessor functions of phyloseq were used to generate heatmaps. Differential abundance analysis was performed using ANCOM-BC version 1.0.2, with Bonferroni correction for false discovery rate and an alpha of 0.05 as a threshold for significance 45 . www.nature.com/scientificreports/