Phylogenetic analyses and antimicrobial resistance profiles of Campylobacter spp. from diarrhoeal patients and chickens in Botswana

Campylobacter spp. are a leading cause of bacterial enteritis worldwide, including countries in Africa, and have been identified by the World Health Organisation (WHO) as one of the high priority antimicrobial resistant pathogens. However, at present there is little knowledge on the prevalence, molecular epidemiology or antimicrobial susceptibility of Campylobacter spp. isolates in Botswana, both in patients and in the zoonotic context. Some data indicate that ~14% of diarrhoeal disease cases in a paediatric setting can be ascribed to Campylobacter spp., urging the need for the magnitude of Campylobacter-associated diarrhoea to be established. In this survey, we have characterised the genomic diversity of Campylobacter spp. circulating in Botswana isolated from cases of diarrhoeal disease in humans (n = 20) and from those that colonised commercial broiler (n = 35) and free-range (n = 35) chickens. Phylogeny showed that the Campylobacter spp. isolated from the different poultry and human sources were highly related, suggesting that zoonotic transmission has likely occurred. We found that for Campylobacter spp. isolated from humans, broilers and free-range chickens, 52% was positive for tetO, 47% for gyrA-T86I, 72% for blaOXA-61, with 27% carrying all three resistance determinants. No 23S mutations conferring macrolide resistance were detected in this survey. In summary, our study provides insight into Campylobacter spp. in poultry reservoirs and in diarrhoeal patients, and the relevance for treatment regimens in Botswana.

Introduction has been associated with increases in resistant Campylobacter spp. causing disease in humans [16]. However, little is understood regarding the potential contribution of this source of antimicrobial resistance in the African context.
Genotypic information about Campylobacter spp. in developing countries is lacking [17]. Currently there are no available data on the prevalence, molecular epidemiology or antimicrobial susceptibility of Campylobacter spp. isolates in Botswana, neither in patients nor in poultry reservoirs. As such, this is the first survey conducted in Botswana with the aim to characterise Campylobacter spp. isolated from diarrhoeal patients, and from free-range and broiler chickens to determine the genetic diversity and the presence of antimicrobial resistance genes.

Results and discussion
Diversity of Campylobacter spp.
Campylobacter spp. were isolated from diarrhoeal paediatric patients admitted to the Princess Marina Hospital in Gaborone, Botswana, and from caecal samples collected from commercial broiler and free-range chickens. The chickens were purchased from various poultry farms within the Gaborone catchment area, which is located close to the border with South Africa. Chickens were obtained from five sites with commercial broilers and eight sites with freerange chickens.
Genomes were sequenced using Illumina HiSeq or MiSeq technologies for 20 Campylobacter spp. isolates from human diarrhoeal patients and 70 isolates from chickens, of which 35 were from commercial broilers and 35 were from free-range chickens. To assess which Campylobacter species each isolate belonged to, an initial comparative genome analysis was performed in by pairwise average nucleotide identity (ANI) BLAST analysis in JSpeciesWS [18] using the generally accepted ANI species cut-off of 94% [19] (S1 Table). Human and freerange chicken isolates were predominantly C. jejuni, with 19 human isolates being C. jejuni (95%) and one being C. coli (5%), while 29 (82.8%) free-range chicken isolates were C. jejuni and 6 (17.1%) were C. coli. Out of the 35 commercial broiler isolates, 14 (40.0%) were C. jejuni and 21 (60.0%) were C. coli (S1 Table).
Mutilocus sequence typing (MLST) identified 19 sequence-types (STs) for C. jejuni and 7 for C. coli (S1 Table). Two novel STs were identified for C. jejuni; ST9024 and ST9027, which both belong to clonal complex ST354. Three novel STs were found for C. coli; ST9025 (clonal complex ST1150), and ST9026 and ST9028 that both belong to clonal complex ST828.
Noteworthy, 24 out of 28 C. coli isolates belonged to clonal complex ST2828 (S1 Table), indicating that, based on MLST, the C. coli isolates are genetically more conserved. In recent Catalonian (north-east Spain) and Italian studies, the predominant clonal complex of C. coli isolates was also ST828 [20,21] Phylogenetic analyses of Campylobacter spp.
Phylogeny was built separately for C. jejuni (Fig 1) and C. coli (Fig 2) based on core genome alignments, which include 1107 or 1197 genes respectively. In addition, phylogeny was also assessed using extracted single nucleotide polymorphisms (SNPs) derived from a referencebased alignment; C. jejuni NCTC11168 [22] was used as a reference (S1 Fig and S2 Fig). Clustering of isolates in both core genome-and SNP-based phylogenies was identical, indicating that both methods successfully extracted the existing phylogenetic signal. It is noteworthy that MLST seems to be an effective indication of related lineages as isolates belonging to the same ST clustered together. Each clade, as defined by ST, appears to contain very little genetic diversity within the clade compared to the differences between clades. These results suggest that the employed methods provide a clear indication of the relatedness of the isolates. However, due to the genetic promiscuity of Campylobacter spp., the trees are unlikely to represent the true phylogenies.
Our data show that isolates from across the phylogenies of each Campylobacter spp. are present in broilers, free-range chickens and humans. As anticipated, strains from the same sample set/location were closely related. However, there is also evidence of (closely) related strains in different locations, suggesting that geographic spread may have occurred. Regarding the possibility of geographic spread of C. jejuni, we observed that isolate S26 from a free-range chicken in Lentsweletau was present in a cluster with isolates from free-range chickens from Boatle that were isolated six months previously. S3 from a free-range chicken in Mochudi was present in a cluster with isolates from free-range chickens from Ramapatle isolated fourteen months later. S28 and S29 from free-range chickens in Lentsweletau are present in a cluster with isolates from free-range chickens from Kopong isolated one month later. S50, S51 and Phylogeny and AMR profile of C. jejuni isolates from this study. Core genome maximum likelihood phylogeny of C. jejuni isolates visualised in the interactive Tree of life tool (iTol) [23]. The tree was rooted on isolate 2445 to facilitate comparison with the SNP-based phylogeny shown in S1 Fig. Clustering of isolates was found to be in accordance between core genome and SNP-based phylogenies (S1 Fig). Clustering of isolates belonging to the same ST was consistent. Shown for each isolate are: isolate identifier, the geographic location of isolation, presence of AMR determinants and ST.  S55 from free-range chickens in Gabane are present in a cluster with isolates from free-range chickens from Gaborone-block 9 (S64) isolated~2 weeks later and Ramapatle (S47) isolated 1 week previous. S4, S10 and S27 from free-range chickens, each from a different geographical location and from a different sampling time (spanning seven months) form a cluster (S1 Table, Fig 1 and S1 Fig).
Regarding the possibility of geographic spread of C. coli, we observed that isolate B39 from a broiler chicken in Gaborone-Old Naledi was present in a cluster with isolates from broiler chickens in Ramotswa (B21) isolated fifteen months previously and Mmatseta (B54) isolated three months later. Isolates B23, B24 and B27 from broiler chickens in Ramotswa were present in a cluster with isolates from broiler chickens in Matebele (B9, B10 and B11) isolated four months later. Isolates B30, B31 and B32 from broiler chickens in Metsimaswane were present in a cluster with isolates from broiler chickens in Ramotswa (B25) and Metsimaswane (B29) isolated on the same day and Gaborone-Old Naledi (B37) isolated fifteen months later. C. coli B15 isolated from a broiler in Gaborone-North, ST9025, a novel ST belonging to clonal complex ST1140, is distant to the other C. coli isolates in this study (Fig 2 and S2 Fig). Pair-wise ANI analyses with the other C. coli isolates in our study confirmed that B15 belongs to the C. coli species, i.e. all were >94% ANI, which is generally considered the species cut-off value [19].
We observed limited mixing of strains between broiler and free-range chickens, as evidenced by most branches of the phylogeny only having strains from a single host, with the exception of C. coli isolates S11 from a free-range chicken in Ramotswa, S66 from a free-range chicken in Gaborone Block 9 isolated nineteen months later and S74 from a free-range chicken in Notwane isolated eighteen months later were present in a cluster with C. coli isolates from a broiler chicken B72 isolated nineteen months later from Kumakwane and a human isolate 2511 from Tlokweng (date of isolation not recorded, but before the isolates from the chickens were collected) (S1 Table, Fig 1 and Fig 2).
For C. jejuni we observed human isolates that were highly related to chicken isolates. This included i) human isolate 2144 from Gaborone and S65 isolated from a free-range chicken in Gaborone Block 9, ii) human isolate 2162 (Gaborone) is related to broiler isolates B13 and B14 from Gaborone North, iii) human isolate 2521 is related to broiler isolates B60, B62, B64, B65, B68 and B70 (Fig 1 and S1 Fig).

Comparative analyses of Campylobacter spp. from Botswana with other African isolates
The distribution of Campylobacter spp. isolates from Botswana was analysed with other African isolates using multi locus sequence type (MLST) allelic profiles available from the PubMLST database [24] (accessed in January 2017). Minimum spanning trees generated and visualised in GrapeTree [25] were used to assess the phylogenetic relatedness amongst African isolates; trees were built separately for C. jejuni (Fig 3) and C. coli (Fig 4) isolates. STs represented by Botswanan isolates predominantly comprised isolates from Botswana only, with the exception of i) C. jejuni ST1932 that comprised 6 Nigerian and 1 Botswanan isolate, ii) C. jejuni ST52 comprised 4 isolates from Senegal and 3 isolates from Botswana, iii) C. coli ST3241 with 3 isolates from Botswana and 1 from Egypt. In line with the dispersed distribution of African C. jejuni and C. coli isolates across the MLST-based tree, Botswanan Campylobacter spp. isolates were also widely distributed across the tree (Fig 3 and Fig 4). Several C. jejuni STs are located at the outside of the tree, e.g. ST137 (human isolate 2556) and ST583 (human isolate 2162, and broiler isolates B13 and B14) that belong to clonal complex ST45, and the clonal complex ST21 isolates that belong to ST19, ST883 and ST5649 that comprises human, free-range and broiler isolates (S1 Table and Fig 3). Some of the commercial broiler C. coli isolates appear to be more distantly related to other African C. coli isolates, i.e. B15 (ST9025) and Nigeria [35] South Africa [26] Burkina Faso [23] Egypt [23] Malawi [18] The Gambia [16] Country Fig 3. MLST of C. jejuni isolates from this study and other African C. jejuni isolates. Minimum spanning tree generated from C. jejuni allelic profiles visualised in GrapeTree [25], which included African C. jejuni isolates that were available in the PubMLST database [24] (accessed in January 2017). Isolates are colour-coded according to country and labelled with STs. The number of isolates per node (ST) are indicated through pie-charts. The number of nucleotide differences between STs is depicted at the branches. If STs differ by more than 2 nucleotides, branches are truncated (dashed lines). More details can be found in S1 Table and  Botswana [28] Nigeria [24] Egypt [16] Burkina Faso [13] Senegal [8] Tanzania [7] Malawi [2] Country Fig 4. MLST of C. coli isolates from this study and other African C. coli isolates. Minimum spanning tree generated from C. coli allelic profiles visualised in GrapeTree [25], which included African C. coli isolates that were available in the PubMLST database [24] (accessed in January 2017). Isolates are colour-coded according to country and labelled with STs. The number of isolates per node (ST) are indicated through pie-charts. The number of isolates belonging to ST9028 (B25, B29, B30, B31 and B27) (S1 Table and Fig 4). The data highlight the diversity of Campylobacter spp. in Botswana and indicate both local and global strains, confirming previous MLST studies with C. jejuni and C. coli isolates from poultry, cattle and humans in Nigeria [17].

Antimicrobial resistance profiles derived from whole genome sequence data
Fluoroquinolones, tetracyclines, beta-lactams, aminoglycosides and macrolides are the most frequently used antimicrobials for the treatment of campylobacteriosis [6,7,9,[11][12][13][14]. In Campylobacter spp., as in other Gram-negative bacteria, the main mechanism of fluoroquinolone resistance is chromosomally mediated through mutation of gyrA by alteration of codon 86 from threonine to isoleucine [6]. Campylobacter spp. carrying the Thr-86-Ile change in the GyrA subunit of DNA gyrase can persist in the absence of antimicrobial selection pressure [11]. In Campylobacter spp. resistance to tetracycline is primarily mediated by a ribosomal protection protein (tetO) that is transferred as a plasmid-encoded gene or on the chromosome where it is not self-mobile [6,9,14]. Resistance to ampicillin and other beta-lactams is widely reported among Campylobacter spp. isolated from humans and poultry [26,27]. A significant knowledge gap still exists concerning the molecular basis of beta-lactam resistance in Campylobacter spp. [28]. Resistance to beta-lactams in Campylobacter spp. can be mediated by OXA-61 beta-lactamase, which is encoded by the bla OXA-61 gene [29], but other beta-lactamases also exist [26]. Aminoglycoside (gentamicin) resistance occurs through the presence of genes including aph(2")-IF, aac/aphD, aph(2")-Ig and aacA4 [30]. High-level macrolide resistance is mediated through mutation of the 23S rRNA [6]. We performed an in silico assessment for mutations conferring antimicrobial resistance and for the presence of antimicrobial resistance genes. For both C. jejuni and C. coli isolates there was a noticeable presence of tetO (52%), gyrA-T86I (47%) and bla OXA-61 (72%) (Fig 1, Fig  2, Table 1 and S1 Table). We found no evidence of other antimicrobial resistance genes (as described in Materials and Methods). In comparison, a recent WGS-based study of 16 Catalonian (north-east Spain) broiler Campylobacter spp. isolates (12 C. jejuni and 4 C. coli) revealed higher prevalence of tetO (81%) and gyrA-T86I (100%) [20].
In this study, mutation of gyrA was limited to T86I, confirming that this is the most common gyrA mutation. This is also in line with studies conducted in Portugal, Catalonia and Poland in which the only detected gyrA mutation was T86I [20,31,32].
Mutations of the 23S rRNA gene that confer macrolide resistance were not detected in the Botswanan Campylobacter spp. isolates reported in this study, indicating that prevalence of macrolide resistance is low in Botswana. This is in line with a Polish study, which detected the A2075G 23S rRNA mutation only in 8 out of 802 poultry isolates [32]. However, a Catalonian study found the A2075G 23S rRNA mutation in 3 out of 4 tested C. coli isolates but not in the C. jejuni isolates [20], whereas an Italian study observed the A2075G 23S rRNA mutation in 49 out of 197 tested Campylobacter spp. isolates [21].
We observed that the gyrA-T86I mutation was more prevalent in Campylobacter spp. isolated from broilers than from free-range chickens (P = 0.0082) ( Table 1). Also, we found that bla OXA-61 was widely distributed in C. jejuni 43/62 (69.4%) and C. coli 22/28 (78.6%) ( Table 1). This is in line with a survey conducted in the United Kingdom, in which 52% of poultry associated Campylobacter spp. isolates were ampicillin-resistant, of which 92% carried bla OXA-61 [26].
C. coli isolates had higher numbers of individual antimicrobial resistance genes, hence they also had a higher proportion of combinations of resistance genes. All three resistance genes i.e. tetO, gyrA-T86I and bla OXA-61 were present in 5/62 (8.1%) of C. jejuni isolates and 19/28 (67.9%) of C. coli isolates (Table 1). This finding supports previous studies that have suggested that C. coli has increased resistance due to exposure to a larger number of antimicrobial agents compared to C. jejuni [12,14]. None of the C. coli isolates were negative for any of the antimicrobial resistance genes 0/28 (0%), while 8/62 (12.9%) of the C. jejuni isolates were negative (Table 1). Interestingly, all six C. coli isolates from the free-range birds contained all three resistance genes.

Conclusion
Here we report on the genomic diversity of circulating Campylobacter spp. causing diarrhoeal disease in humans and of colonised chickens in Botswana. Phylogeny showed that the Campylobacter spp. isolated from the different poultry and human sources were highly related, suggesting that zoonotic transmission is likely.
In a large proportion of Campylobacter spp. isolated from humans, broilers and free-range chickens in this study we detected tetO (52%), gyrA-T86I (47%) and bla OXA-61 (72%) genes. Our study has identified isolates carrying multiple drug resistance genes, i.e. 27% of the isolates carried all three above mentioned resistance determinants. The C. coli isolates analysed in our study had a higher prevalence of the individual antimicrobial resistance genes when compared to C. jejuni isolates, which is in line with other studies [20,21,31,32]. The whole genome sequence data indicates that Campylobacter spp. isolated from humans and chickens in Botswana have the potential to be resistant or multiply resistant to a number of antimicrobial classes, however we did not test specifically for phenotypic antimicrobial resistance. The presence of antimicrobial resistance genes to first line treatment drugs is a concern. Collectively, the data suggest that the efficacy of using current treatments for treating human campylobacteriosis and for the treatment or use as growth promoters in chickens should be evaluated. Our analyses did not identify any Campylobacter spp. carrying 23S rRNA gene mutations that confer macrolide resistance, suggesting that erythromycin or azithromycin may still be an effective treatment. In addition, aminoglycosides such as gentamycin may be an alternative option as we found no evidence of aphA-3 and aacA4 genes.
Importantly, this survey provides detailed insight into the phylogeny of and antimicrobial resistance profiles of Campylobacter spp. in Botswana. Further studies that cover a larger geographic study area are necessary to detect emerging resistance patterns and to assess the impact of strategies designed to mitigate antimicrobial resistance.

Isolation of Campylobacter spp. from commercial and free-range broilers
The chickens were purchased from various poultry farms within the Gaborone catchment area, which is located close to the border with South Africa. Chickens were obtained from five sites with commercial broilers and eight sites with free-range chickens. The dataset included fully grown ready-for-slaughter chickens from poultry farms and free-range chickens obtained from small enterprises; only one farm was sampled in each sample represented in S1 Table. 'Free-range' chickens refer to the indigenous breed of chickens kept in Botswana. This type of chicken is slow growing, reaching maturation and ready-for-slaughter at 5-6 months (opposed to the broilers that are slaughtered at 4-5 weeks). These chickens are not housed but are allowed to roam and forage in the farm area during the day, they are only sheltered to protect them from predators at night.
A loop full of caecal sample was streaked on Campylobacter Blood-free selective agar; Charcoal-Cefoperazone-Deoxycholate Agar (CCDA) [33], mixed with a selective supplement SR0155 containing cefoperazone 32 mg/l and amphotericin B 10 mg/l of medium (Oxoid).
Isolates were incubated at 37˚C for 48 h under microaerobic conditions (Oxygen 5%, Carbon dioxide 5-8%) achieved using GasPak TM EZ Campylobacter sachets and incubation chambers (Becton Dickinson and Company, USA). A plate inoculated with Campylobacter spp. strain, C. jejuni ATCC 1 33560 TM (Microbiologics St. Cloud, USA), was included in every incubation chamber as a control. Typical watery, spreading and convex colonies were picked and screened for Campylobacter spp. using Gram stain, catalase and oxidase tests. Gram-negative, curved or seagull wing-shaped, catalase and oxidase positive, bacterial isolates were purified and suspended in cryovials containing plastic beads and 40% glycerol, followed by freezing and storage at -80˚C.

Isolation of Campylobacter spp. from human infection cases
Specimens were obtained from diarrhoeal patients from the largest referral hospital in Botswana (Princess Marina Hospital, Gaborone, Botswana) and surrounding clinics. Inclusion criteria include all stools from patients with acute gastroenteritis under the age of 13 years and loose, watery and bloodstained stools from patients over the age of 13 years. Acute gastroenteritis was defined as having >2 loose bowel movements in a 24 h period or vomiting with any loose stools with a maximum duration of 13 days. Exclusion criteria included, discharge from hospital within 7 days of admission and diarrhoea of 14 days or longer. Samples were only collected from patients that had given approval.
Stools collected in Carry Blair transport media (Becton Dickinson and Company, USA) were cultured on the same day on Campylobacter selective agar; Columbia agar base with Skirrow's selective supplement with added vancomycin 10 mg/l, polymixyn B 5 mg/l and trimethoprim 2500 IU/l and 10% sheep blood. Samples were incubated under microaerobic conditions (CampyGen, Oxoid). The weight and volume of the inoculum was not measured for each sample; a loop-full piece of stool, picked from the mucous part of the specimen, was streaked on the selective agar media. Gram-negative curved or seagull wing-shaped, catalase and oxidase positive, bacterial isolates were purified by subculture on new blood agar plates. The isolates were suspended in cryovials containing plastic beads and 40% glycerol, followed by freezing and storage at -80˚C.

Isolation of Campylobacter spp. genomic DNA for sequencing
Pure cultures of isolates were individually suspended in saline to match 0.5 MacFarland standard. DNA was extracted from the suspensions using MagNA Pure Compact Nucleic Acid Isolation Kit I (Roche Diagnostics GmbH, Germany) on MagNA Pure Compact automated extraction and purification platform, according to the manufacturer's instructions (Roche Diagnostics GmbH, Germany).

Whole genome sequencing
Genome sequencing libraries were prepared using the NEBNext Ultra II DNA library prep kit (New England Biolabs) as described in de Vries et al. [34]. The libraries were sequenced using 150 bp paired-end sequencing on the Illumina HiSeq 4000 platform (Genomics core facility at Cancer Research UK) or on the Illumina MiSeq platform using 250-300 bp paired-end sequencing (in-house facility).

Genome assembly and annotation
The Illumina HiSeq 4000 read files were demultiplexed using the demuxFQ tool (Cancer Research UK) and adapter-trimmed in Cutadapt. De novo draft genome assemblies were created using Illumina HiSeq or MiSeq paired-end reads using Spades v3.6.2 [35] and annotated with Prokka [36].

In silico-based MLST phylogeny
MLST was performed using the short-read sequence-typing tool SRST2 [37] with allele sequences derived from the PubMLST database [24] (accessed in January 2017) as reference. Novel allele sequences and STs were submitted to PubMLST database [24]. For comparative analyses with African Campylobacter spp. isolates, MLST minimum spanning trees were generated and visualised in GrapeTree [25], which included African C. jejuni and C. coli isolates available in PubMLST database [24] (accessed in January 2017). Further examination of the assembled genomes of the C. coli revealed some anomalies with the MLST genes. In 17 out of 28 C. coli isolates we observed the presence of an additional MLST gene allele. In all cases, this involved a single gene out of the 7 MLST genes, e.g. in C. coli 2511 both gltA_16 and gltA_30 were present and in C. coli B10 both gltA_10 and gltA_30 were found. The second allele was consistently detected at low coverage and in small contigs. Collectively, this involved 6 out of the 7 MLST genes: gltA, uncA, aspA, glyA, tkt and glnA. For the C. coli MLST analyses presented, the high sequence coverage hits were selected. Importantly, such observations were not made in any of the C. jejuni isolates which were sequenced at the same time, intermingled with the C. coli isolates, suggesting that this is not related to a sequencing error or contamination. This may be the result of horizontal gene transfer with the species C. coli.

Core genome and SNP-based phylogenetic analyses
Core genome phylogenetic analysis was performed in Roary [38] using the Prokka annotation [36]; a 95% nucleotide identity cut-off was applied for inclusion in the core genome. Sequences were aligned using MAFTT [39] and Maximum Likelihood (ML)-phylogeny was reconstructed in RaxML v8.2.9, with GTRCAT approximation and automatic bootstrapping (autoMRE) using rapid bootstrapping mode [40], and visualised in the interactive Tree of life tool (iTol) [23]. For SNP-based phylogenies, sequence reads were aligned against C. jejuni NCTC11168 [22] as a reference genome using SMALT [41], following default settings to identify SNPs. Extracted SNPs were used to construct ML-phylogenies using RaxML v 7.3.6 (rapid bootstrapping and subsequent ML search using 100 bootstrap replicates) [40].

In silico identification of antimicrobial resistance determinants
The gyrA and 23S rRNA gene sequences were extracted from the Prokka genome annotations. The gyrA sequence was translated to the protein sequence and aligned. The 23S rRNA gene sequences were directly aligned. Alignments were conducted in in CLC main workbench v7.3.3. For GyrA we assessed aa position 70 for Ala to Thr (not found); amino acid (aa) position 86 Thr to Ile (found and discussed), aa position 90 Asp to Asn (not found). For the 23S rRNA gene we assessed positions 2074 and 2075 of which mutations confer high-level macrolide resistance [42]; no resistance-associated mutations were found. The intergenic region between cmeR and cmeABC or mutations in the L4 and L22 ribosomal proteins were not assessed as part of this study.
Antimicrobial resistance (AMR) gene detection was conducted through alignment of Illumina sequence reads to a custom AMR gene database, which was built from gene sequences available from the ARG-ANNOT [43] and RESfinder [44] databases. Of note, the compiled AMR gene database also contained ermB that is implicated in macrolide resistance [43].

Statistical analyses
Statistical analyses were performed in R via the rpy2 module in Python. Bonferroni-corrected P-values are reported.

Nucleotide sequence accession numbers
Illumina whole genome sequencing data has been deposited in the European Nucleotide Archive (http://www.ebi.ac.uk/ena) and are available via study accession number PRJEB18561.