1. A. B. Proteobacteria Epsilon Proteobacteria Epsilon Proteobacteria Supplemental Figure 1. Consensus sequences for (A) forward and (B) reverse primers based on alignment of 49 proteobacteria (upper panels) or a subset of 19 Epsilon proteobactiera (lower panels). Sequences of the Epsilon enrichment primers JH0108 and JH0102 are shown below the Epsilon proteobacteria consensus sequences. Figures were generated by Weblogo (http://weblogo.berkeley.edu/). 5'-ATGAANTTTCARCCWYTWGG-3’ JH0108 5'-ARCATHKCTTTTCTTCTRTC-3’ JH0102

2. Campylobacter jejuni (ε) Campylobacter fetus (ε) Helicobacter sp. (ε) Salmonella (γ) Pasteurella (γ) Pseudomonas (γ) Bartonella (α) Supplemental Figure 2. Temperature gradient PCR with Epsilon-proteobacteria enrichment primers on pure proteobacteria cultures. PCR primers JH0108 and JH0102 were tested with genomic DNA extracts in a gradient PCR with annealing temperatures from 49.3°C to 60.1°C (indicated by increasing triangle), plus no template control (-). Arrows indicate the 1200 bp marker. The cpn10- cpn60 target of ~1200 bp was clearly seen with epsilon-proteobacteria genomic DNA (C. jejuni, C. fetus and Helicobacter sp.), while only non-specific bands were obtained from gamma- and alpha- proteobacteria genomic DNA (Salmonella, Pasteurella, Pseudomonas and Bartonella samples).

3. Supplemental Figure 3. Rarefaction curves of dog fecal microbiota libraries generated in this study. Both rarefaction analysis and the Good’s coverage estimate (given in parentheses after each library) were calculated using mothur software.

4. Supplemental figure 3 (continued). Rarefaction curves of dog fecal microbiota libraries generated in this study. Both rarefaction analysis and the Good’s coverage estimate (given in parentheses after each library) were calculated using mothur software.

5. Supplemental Figure 4A. A jackknifed clustering of the samples by Unifrac distances (100 permutations) of non-enriched libraries. The numbers at the nodes indicate the number of times that particular node was observed (out of 100) in a random sampling of the dataset. Brackets highlight the distinction between the cluster that is predominantly healthy dog libraries and the cluster that is predominantly diarrheic dog libraries. 0.1 HDS1fall HDS2fall HDS2spring DDS18 HDS8 HDS1spring HDS18D HDS30 DDS2 DDS64 DDS34 HDS19 DDS19 DDS32 DDS30 HDS9A DDS11 DDS51 43 70 87 100 61 70 87 84 100 49 100

6. 0.1 HDS19.enrich HDS1fall.enrich.rep HDS1fall.enrich DDS19.enrich DDS30.enrich HDS2fall.enrich HDS2spring.enrich HDS18D.enrich HDS1spring.enrich HDS8.enrich HDS9A.enrich DDS51.enrich HDS30.enrich DDS34.enrich DDS2.enrich DDS32.enrich DDS64.enrich DDS18.enrich HDS9A.universal HDS1fall.universal HDS2fall.universal HDS2spring.universal DDS18.universal HDS8.universal HDS1spring.universal HDS18D.universal HDS30.universal DDS19.universal DDS32.universal DDS2.universal DDS64.universal DDS34.universal HDS19.universal DDS11.universal DDS51.universal DDS11.enrich DDS30.universal 100 98 100 97 100 97 85 100 72 62 65 67 36 34 42 67 48 71 97 100 99 100 98 100 Supplemental Figure 4B. A jackknifed clustering of the samples by Unifrac distances (100 permutations) of all libraries generated in the study. The numbers at the nodes indicate the number of times that particular node was observed (out of 100) in a random sampling of the dataset. Brackets highlight the distinction between the cluster that is predominantly universal (non-enriched) cpn60 libraries and the cluster that is predominantly epsilon- proteobacteria enriched libraries.

7. Phylum HDS1fall _enrich1 HDS1fall _enrich2 Actinobacteria<1 a 22 Bacteroidetes42 Firmicutes531183 Proteobacteria24662796 a Scaled number of reads in library Genus HDS1fall _enrich1 HDS1fall _enrich2 Aeromonas0a0a 10 Atopobium<17 Bacteroides42 Bifidobacterium05 Blautia10 Butyrivibrio27051 Campylobacter340 Clostridium10 Coprococcus25229 Cupriavidus05 Enterobacter<10 Gardnerella010 Hafnia02 Helicobacter24292659 Lactobacillus149 Peptostreptococcus<139 Ruminococcus40 Shewanella2120 Streptococcus<115 a Scaled number of reads in library Supplemental Figure 5. Comparison of technical replicates of HDS1fall at the (A) phylum and (B) genus level. A B

8. 126 isotigs (possible novel Helicobacter sp.) isotig 232 12 isotigs H. cinaedi isotig 536 22 isotigs H. canis H. hepaticus isotig 164 isotig 646 H. mustelae H. nemestrinae H. acinonychis H. fennelliae isotig 887 (Flexispira rappini OTU) isotig 864 H. felis H. salomonis H. bizzozeronii H. pylori H. cholecystus H. winghamensis H. pametensis H. muridarum H. pullorum H. bilis H. trogontum isotig 027 isotig 731 A. skirrowii H. canadensis A. nitrofigilis C. sputorum isotig 905 isotig 579 isotig 883 A. mytili C. hyointestinalis C. upsaliensis C. helveticus C. coli C. fetus venerealis C. jejuni C. fetus fetus C. gracilis C. retus C. lari isotig 585 isotig 632 (possible novel Campylobacter species) isotig 594 C. showae C. concisus C. mucosalis C. curvus E. coli 0.1 Supplemental Figure 6. Phylogenetic tree of epsilon-proteobacteria OTU. A 100 bp region (from nucleotide 200-300 within the cpn60 UT region) was aligned and used to generate a neighbour-joined tree. Alignments were manually inspected to identify any anomolous sequences that incorrectly passed quality control (none were identified). Species are highlighted by coloured backgrounds. Isotigs (equivelant to OTU) that represent possible novel species have been highlighted with the label “(possible novel X species)”.