A Novel 16S rRNA PCR-Restriction Fragment Length Polymorphism Assay to Accurately Distinguish Zoonotic Capnocytophaga canimorsus and C. cynodegmi

ABSTRACT The zoonotic bacteria Capnocytophaga canimorsus and C. cynodegmi, the predominant Capnocytophaga species in the canine oral biota, can cause human local wound infections or lethal sepsis, usually transmitted through dog bites. Molecular surveying of these Capnocytophaga species using conventional 16S rRNA-based PCR is not always accurate due to their high genetic homogeneity. In this study, we isolated Capnocytophaga spp. from the canine oral cavity and identified them using 16S rRNA and phylogenetic analysis. A novel 16S rRNA PCR-restriction fragment length polymorphism (RFLP) method was designed based on our isolates and validated using published C. canimorsus and C. cynodegmi 16S rRNA sequences. The results showed that 51% of dogs carried Capnocytophaga spp. Among these, C. cynodegmi (47/98, 48%) was the predominant isolated species along with one strain of C. canimorsus (1/98, 1%). Alignment analysis of 16S rRNA sequences revealed specific site nucleotide diversity in 23% (11/47) of the C. cynodegmi isolates, which were misidentified as C. canimorsus using previously reported species-specific PCR. Four RFLP types could be classified from all the isolated Capnocytophaga strains. The proposed method demonstrates superior resolution in distinguishing C. cynodegmi (with site-specific polymorphism) from C. canimorsus and especially in distinguishing C. canimorsus from other Capnocytophaga species. After in silico validation, this method was revealed to have an overall detection accuracy of 84%; notably, accuracy reached 100% in C. canimorsus strains isolated from human patients. Overall, the proposed method is a useful molecular tool for the epidemiological study of Capnocytophaga in small animals and for the rapid diagnosis of human C. canimorsus infections. IMPORTANCE With the increased number of small animal breeding populations, zoonotic infections associated with small animals need to be taken more seriously. Capnocytophaga canimorsus and C. cynodegmi are part of common biota in the mouths of small animals and can cause human infections through bites or scratches. In this study, C. cynodegmi with site-specific 16S rRNA sequence polymorphisms was erroneously identified as C. canimorsus during the investigation of canine Capnocytophaga by conventional PCR. Consequently, the prevalence of C. canimorsus is incorrectly overestimated in epidemiological studies in small animals. We designed a new 16S rRNA PCR-RFLP method to accurately distinguish zoonotic C. canimorsus from C. cynodegmi. After validation against published Capnocytophaga strains, this novel molecular method had high accuracy and could detect 100% of C. canimorsus-strain infections in humans. This novel method can be used for epidemiological studies and the diagnosis of human Capnocytophaga infection following exposure to small animals.

T he genus Capnocytophaga comprises fastidious, Gram-negative, thin or filamentous rods with tapered or spindle-shaped ends, and are facultative anaerobic and capnophilic bacteria (1,2). Four Capnocytophaga species have been isolated from the canine oral cavity; among these, C. canimorsus and C. cynodegmi are more predominant than C. canis and C. stomatis (3)(4)(5)(6). They are most commonly transmitted to humans via dog bites, followed by scratches or close contact (7). Although infected patients exhibit symptoms ranging from mild to fulminant, immunocompromised patients tend to show more severe clinical signs (7,8). Compared to other species, C. canimorsus usually causes systemic infections, including cellulitis, meningitis, and sepsis (9). C. cynodegmi is considered less pathogenic than C. canimorsus, as most C. cynodegmi-infected cases develop local wound infection (cellulitis) which rarely advances to systemic infection (10,11).
In some epidemiological studies, culture-and/or direct PCR-based techniques are often used to evaluate the potential hazards of zoonotic Capnocytophaga spp. transmission from the canine oral cavity to humans. Among the species investigated, C. canimorsus has been the most extensively surveyed (3,5,(12)(13)(14)(15). PCR-based methods usually reveal a higher prevalence of Capnocytophaga than culture-based methods because of the fastidious property of this genus and the high sensitivity of PCR. However, without the culture process, the phenotypic and other important genotypic characteristics of Capnocytophaga cannot be determined.
Genetic identification of Capnocytophaga spp. is commonly performed by 16S rRNA sequencing (3,15). However, considering the high homogeneity of the 16S rRNA gene between C. canimorsus and C. cynodegmi, other techniques, including pulsed-field gel electrophoresis (PFGE), multilocus sequence typing (MLST), and PCR-restriction fragment length polymorphism (RFLP), have also been applied to discriminate between canine Capnocytophaga spp. (14)(15)(16)(17). Suzuki et al. developed genus-and species-specific PCR targeting the 16S rRNA gene to allow rapid identification and distinction between C. canimorsus and C. cynodegmi strains isolated from dogs and cats (5). These primer sets have also been widely adopted in other studies to investigate the prevalence of Capnocytophaga spp. in small animals (4,14,18).
In this study, we discovered that the 16S rRNA sequence polymorphism at the specific location in C. cynodegmi strains showed ambiguous results when using the previously established species-specific PCR for culture-based epidemiological investigation of Capnocytophaga in dogs. Therefore, we utilized the isolated strains to develop a novel 16S rRNA PCR-RFLP method which accurately distinguished between C. canimorsus and C. cynodegmi, and demonstrated its potential application in future epidemiological studies and disease diagnosis.

RESULTS
Prevalence investigated using conventional methods: species-specific PCR and 16S rRNA sequences. Oral swabs were collected from 98 dogs (50 males and 48 females), including 82 owned and 16 sheltered dogs. The median age of the dogs was 8 years (range: 0.25 to 17 years). Relevant individual parameters were summarized; none of the basic individual parameters was significantly correlated with the presence of Capnocytophaga (Table S1). The prevalence of Capnocytophaga sp. in the canine oral cavity was 51% (49/98). Fifty Capnocytophaga strains were isolated from 49 dogs. After sequencing the amplified 16S rRNA genes of these 50 strains and trimming the poor signal sequences from both ends, sequences of approximately 1,331 bp were used for the BLAST and subsequent phylogenetic analysis. C. cynodegmi showed the same high prevalence (47.96%, 47/98) among all the sampled dogs in both the species-specific PCR and 16S rRNA sequence BLAST results. The differences in the samples surveyed by species-specific PCR and 16S rRNA sequencing in C. canimorsus were 12.24% (12/98) and 1.02% (1/98), respectively. Eleven strains identified as C. cynodegmi based on the 16S rRNA gene, with identities ranging from 98.5% to 99.7%, were positive for both C. canimorsus and C. cynodegmi by species-specific PCR. The other two strains (62067N and 91462) were identified as C. cynodegmi based on the 16S rRNA gene, with 97.4% and 95.7% identity, respectively. These are likely to be new species and remained unidentified in the genus-and species-specific PCR. Two different Capnocytophaga species were present in a single specimen: C. cynodegmi and a suspected new species (strain 62067N).
Effect of site-specific sequence polymorphism on genus-and species-specific PCR and the phylogenetic tree. We constructed a phylogenetic tree based on the 16S rRNA sequences of the 50 isolates with characteristic annotation and 9 reference strains obtained from GenBank (Fig. 1). The only C. canimorsus isolate was well separated from the C. cynodegmi clade, with 99% bootstrap support. In contrast, the 47 C. cynodegmi isolates showed high similarity. Strain 91462 belonged to the same clade as C. canis, with 99% bootstrap support; however, the 16S rRNA BLAST results showed 97.4% identity with C. cynodegmi, suggesting the need for further species identification. Another unidentified Capnocytophaga strain (62067N) was shown to be entirely independent of other canine-related Capnocytophaga strains (Fig. 1A). The reference strain C. stomatis H2177 (GenBank accession no. NZ_CP022387.1) was not well distinguished from C. cynodegmi according to the 16S rRNA-based phylogenetic tree.
The aforementioned 13 Capnocytophaga isolates showed inconsistent results between the 16S rRNA gene BLAST and genus-and species-specific PCR. The negative PCR results (strains 91462 and 62067N) are likely due to marked differences in the recognition sequences of the forward primer CaL2 shared by genus-and species-specific PCR ( Fig. 2A and B). Upon further analysis of C. cynodegmi-specific PCR reverse primer recognition sequences in our isolates, the 50 Capnocytophaga strains were classified into five different types, A to E ( Fig. 1B and 2C). Strains belonging to types A, C, D, and E were consistent with the speciesspecific PCR results and 16S rRNA gene sequence BLAST analysis. The type B strains, mostly aggregated in two clades, were all identified as C. cynodegmi-the 11 strains described white squares correspond to strains that were positive and negative, respectively, in the catalase test, oxidase test, or C. canimorsus-specific (CaR) PCR test. Molecular typing according to the nucleotide polymorphism of species-specific PCR reverse primer binding sites is presented using a color code, with the key shown in the left upper corner of the diagram. T, type strain; H, human isolate; D, canine isolate; F, feline isolate. above which exhibited both C. canimorsusand C. cynodegmi-specific PCR positivity (Fig. 1B). These strains showed a substitution of nucleotide G for A at 16S rRNA gene position 482, which resulted in only two nucleotides (positions 479 and 483) differing from the C. canimorsus-specific reverse primer (Fig. 2C). This explains the ambiguous species-specific PCR results of the type B strains.
Biochemical identification. From the biochemical profiling, only three strains (3/ 47) had species IDs matching their 16S rRNA results. No ID was shown for the other 44 isolates, which showed low similarity indices. Among these isolates, the results for 82% (36/44) of the strains were consistent with the 16S rRNA sequencing results. In contrast,  Specific carbohydrate utilization (previously used to differentiate C. canimorsus from C. cynodegmi) was applied to our isolates, and the results are shown in Fig. 1B. All 47 strains of C. cynodegmi were catalase positive, and 87.2% (41/47) were oxidase positive (Fig. 1B). The single C. canimorsus strain showed positive results for both the catalase and oxidase tests. The unidentified Capnocytophaga strain (62067N) was the only strain negative for both the catalase and oxidase tests.
Differentiation of C. canimorsus and C. cynodegmi using the novel 16S rRNA PCR-RFLP method. The amplified 16S rRNA gene products of 50 Capnocytophaga isolates were treated with REs, KpnI, and PpuMI and analyzed using gel electrophoresis (Fig. 3). Only the PCR product from C. canimorsus was cleaved by PpuMI. In contrast, the PCR products of C. cynodegmi and the two unidentified strains (91462 and 62067N) were cleaved only by KpnI at different sites. Using the 16S rRNA gene of the reference strain, four patterns were predicted in silico by combining KpnI and PpuMI (Table 1 and Fig. 1B). The type 1 pattern was represented by fragment patterns of 14, 51, 286, and 1,114 bp in the C. canimorsus reference strains; our isolates showed consistent results (Fig. 3). However, the small molecular weight fragments (,100 bp) are barely visible because of poor stainability. The C. cynodegmi isolates showed two different restriction fragment patterns after treatment with KpnI: the type 2 pattern (478 and 987 bp) was found in 6 isolates, whereas the type 3 pattern (51, 478, and 936 bp) was found in 41 isolates and three reference strains. As predicted in silico (Table 1), the reference strain C. stomatis showed the same pattern as C. cynodegmi (type 3). The last pattern, type 4 (51 and 1,414 bp), was observed in the reference strains of C. felis and C. canis, and in our unidentified strains (Table 1). In summary, C. canimorsus (fragments 286 and 1,114 bp) and C. cynodegmi (fragments 478 and 936 bp or 987 bp) could be clearly and easily differentiated by gel electrophoresis of the 16S rRNA PCR amplicons digested by KpnI and PpuMI (Fig. 3).
The RFLP type could be predicted in combination with site-specific nucleotide polymorphisms. Based on the KpnI cleavage site, sequences with a "C" nucleotide at position 479 can be cleaved by KpnI at position 480 (Fig. 2C). Therefore, the reverse primer types A, B, and C correspond to RFLP type 2 or 3, while types D and E correspond to RFLP type 1 or 4 ( Fig. 1B).
Verification of expected RE sites in the published C. canimorsus and C. cynodegmi strains. After phylogenetic analysis, 172 C. canimorsus and 35 C. cynodegmi 16S rRNA gene sequences were collected. The sequence of the expected restriction enzyme (RE) site of each strain is presented in the supplemental material ( Fig. S1 and S2). The results showed that 83% (142/172) of C. canimorsus and 91% (32/35) of C. cynodegmi strains were consistent with the designed 16S rRNA PCR-RFLP results. The overall accuracy for RE detection was 84% (174/207). Importantly, the accuracy for detecting C. canimorsus strains isolated from human patients was 100%. All the strains (30 C. canimorsus and 3 C. cynodegmi strains) that showed inconsistent results (no detection of the predicted restriction sites) were isolated from dogs and cats. Failure to identify the expected RE in C. canimorsus strains mainly (28/30, 93.3%) resulted from a C/T substitution at position 289 and a G/T substitution at position 300. The other two C. canimorsus strains (accession no. GQ167582.1 and GQ167602.1) did not show the expected RE due to a single deletion at the expected RE position (position 288). A similar finding was obtained in three C. cynodegmi strains, where the deletion occurred at position 475 ( Fig. S1 and S2).
The phylogenetic tree indicated that 28 of the 30 C. canimorsus strains lacking the expected RE site were allocated to a separate clade, with 89% bootstrap support (Fig. 4A). In contrast, the two C. canimorsus strains with a single deletion mentioned above (accession no. GQ167582.1 and GQ167602.1) were phylogenetically close to other C. canimorsus strains (Fig. 4A). Similarly, three strains of C. cynodegmi with a single deletion at the expected RE site were in proximity to the neighboring clades of other C. cynodegmi strains (Fig. 4B).

DISCUSSION
In this study, we investigated the presence of Capnocytophaga in the canine oral cavity in Taiwan using a novel combination of culture-based and molecular methods. Molecular screening of Capnocytophaga using DNA extracted directly from specimens allows rapid identification in large-scale studies; however, the absence of a culture method reduces the specificity and accuracy of molecular methods. A combination of culture-based and molecular methods facilitates the realistic characterization of a specimen. In this study, we discovered some C. cynodegmi isolates with nucleotide polymorphisms that were poorly detected using the existing species-specific PCR methods; however, these strains were correctly identified using our newly developed 16S rRNA PCR-RFLP.
The molecular techniques used to identify and differentiate C. canimorsusand C. cynodegmi-related species are summarized in Table 2. Similar PCR-RFLP methods have been reported to identify the presence of Capnocytophaga spp. and distinguish between C. canimorsus and C. cynodegmi (15,16). Ciantar et al. (16) used CfoI digestion of PCR-amplified targeted 16S rRNA gene fragments to successfully distinguish between human isolatederived Capnocytophaga species. When we applied this method to our isolates, C. canimorsus and C. cynodegmi only differed by a fragment of 100 bp, which is not easily identified because of its low molecular weight. Another study utilized two REs (StyI and MseI) to cleave rpoB for species distinction (15); however, these RFLP patterns are complicated and may result in multiple patterns from a single species. The novel restrictive enzymes system allowed the RFLP patterns in our method to show clear and easy identification, with distinct patterns for C. canimorsus and C. cynodegmi. The discovery of new nucleotide polymorphisms in the 16S rRNA gene of some C. cynodegmi strains hinders species differentiation primer design for conventional PCR. After retrieving the published 16S rRNA sequences of C. cynodegmi, we found nucleotide polymorphisms at the same position in the previously published C. cynodegmi strains Ccy19 (accession no. GQ167551.1), 11019B-1 (AB851839.1), and LUMC-HA2 (EU124419.1) with occurrence frequencies of 33% (1/3), 14% (1/7), and 17% (1/6), respectively, among the C. cynodegmi isolates in the respective studies (13)(14)(15). This frequency is similar to our finding of 23% (11/47). These strains were identified using culture-based methods which did not include species-specific PCR identification; therefore, the identification error could not be determined. Nucleotide polymorphism C. cynodegmi strains (NP-Ccy) have been observed in Taiwan and other countries, including Switzerland, Japan, and the Netherlands (13)(14)(15). This worldwide distribution of NP-Ccy strains affects the accuracy of species-specific PCR surveys, resulting in an overestimated frequency of C. canimorsus from the canine oral cavity. If the same primer sets were used in this study, the prevalence of C. canimorsus would have been overestimated by 12.2%, instead of 1%, as reported here. The proposed PCR-RFLP method overcomes the misidentification of C. cynodegmi as C. canimorsus and can be applied in the future for PCR-or culture-based epidemiological surveys of Capnocytophaga in small animals. In short, this novel PCR-RFLP method has superior accuracy, efficiency, and ease of application compared to FIG 3 Legend (Continued) pattern type 2 (478 and 987 bp); Ccy2, C. cynodegmi belonging to RFLP pattern type 3 (51, 478, and 936 bp); Cca, C. canimorsus belonging to RFLP pattern type 1 (14, 51, 286, and 1,114 bp); C. sp.: strains 62067N and 91462, the suspected new Capnocytophaga species, belonging to RFLP pattern type 4 (51 and 1,414 bp); P, undigested PCR-amplified 16S rRNA product. other methods. However, this technique can possibly misidentify non-Capnocytophaga bacteria with the same/similar RFLP patterns. Genus-specific PCR could thus be used for confirmation, which will be discussed later. Applying this novel PCR-RFLP analysis to culture-based studies can significantly improve the diversity of Capnocytophaga species detected in oral specimens. Moreover, this method is more efficient than direct 16S rRNA gene sequencing for detecting the presence of C. canimorsus (PpuMI-specific), C. cynodegmi (KpnI-specific), and other species (mismatched patterns of both REs). Considering that non-Capnocytophaga bacteria exhibit DNA extracted from previous restriction fragments (containing primer pair recognition sites) as genus-specific PCR templates would rule out possible misidentification. Moreover, the method can also be used in certain resource-limited research or clinical units to identify Capnocytophaga, especially to detect C. canimorsus. For example, genomic DNA was extracted from oral or blood specimens as a template for C. canimorsusand C. cynodegmi-specific PCR, followed by the cleavage of positive PCR products with PpuMI. The results showed that C. canimorsus-specific PCR cleaved products with a fragment pattern of 15,198, and 216 bp were identified as C. canimorsus (confirmed by our C. canimorsus isolate, data not shown), whereas cleaved products 427 bp in size were identified as C.
cynodegmi. This method facilitates the rapid diagnosis of patients with a history of small animal contact suspected of C. canimorsus infection. Differentiating between Capnocytophaga species is challenging because C. stomatis cannot be distinguished from other C. cynodegmi strains. This finding was consistent with a previous study reporting that C. stomatis (isolated from a human patient) and C. cynodegmi, which show high genetic similarity, can only be distinguished by whole-genome sequencing (6). This implies that specimens identified as C. cynodegmi might also include some strains of C. stomatis (even at a low frequency) because these two species cannot be distinguished by 16S rRNA gene sequencing.
Verification of our designed method revealed a high prediction accuracy for the detection of currently published C. canimorsus and C. cynodegmi strains. There are three possible reasons for profiling failure to identify C. canimorsus and C. cynodegmi strains: (i) some C. canimorsus strains were probably other Capnocytophaga species, (ii) incomplete signals were obtained during sequencing, and (iii) sequences between 16S rRNA gene copies were divergent. In the phylogenetic trees of C. canimorsus isolated in the respective studies, strains of C. canimorsus without the expected RE locus formed a separate clade from that of the other strains (3,13,14). Umeda et al. (14) proposed that the C. canimorsus strains in a separate clade (21 isolates in that study) were likely to be other Capnocytophaga species or subspecies of C. canimorsus. After our phylogenetic analysis, we obtained consistent results and considered that these strains were very likely to be other Capnocytophaga species. In addition, those strains were highly homogeneous with the others, with only a single deletion at the expected RE position; thus, we speculated that there might have been signal recognition errors in the sequencing process of these strains. Finally, multiple copies of the 16S rRNA gene of C. canimorsus strain 24231 have been identified from sequence accession numbers AY643078.1 and AY603477.1, which originated from different clones that are highly similar to the type strains of C. cynodegmi and C. canimorsus, respectively. The final identification may depend on other genes or whole-genome sequencing. If these reasons for suspicion are all correct, our method can serve as an efficient molecular tool for epidemiological studies in small animals and for diagnosing C. canimorsus and C. cynodegmi infection in humans.
Identifying Capnocytophaga through biochemical testing is difficult due to its fastidious characteristics. The Vitek 2 automatic biochemical identification system (bioMérieux, Marcyl' Etoile, France) has been used to identify C. canimorsus and C. cynodegmi in blood and  (18), with a 50% rate of correct identification at the genus level (10/20) (18).
The Biolog system had low species identification power but a high reference value. Compared to Vitek 2, Biolog is a semiautomatic identification system and requires more manual manipulation to enable a favorable biochemical response by bacteria. Carbohydrate utilization (fructose, sucrose, melibiose, inulin, and raffinose) have been previously used to differentiate between C. canimorsus and C. cynodegmi (6). In this study, C. cynodegmi isolates demonstrated variable carbohydrate utilization. Thus, carbohydrate utilization may not be sufficient for differentiating between C. canimorsus and C. cynodegmi. Positive oxidase and catalase tests were previously considered to be indicative of C. canimorsus and C. cynodegmi. However, this study revealed variable oxidase test results in C. cynodegmi isolates. Suzuki et al. (17) conducted an MLST analysis and found that the virulence of C. canis strains in humans is associated with oxidase activity. Thus, whether oxidase activity affects the virulence of C. cynodegmi requires further research for validation. The prevalence of Capnocytophaga differs across countries and is dictated by the detection methods used, such as culture-or PCR-based methods. In general, the frequency of Capnocytophaga spp. detected in the canine oral cavity using culture-based methods is lower than that obtained using PCR-based methods (3,5,13,15,19,20). For example, the occurrence rate of C. canimorsus determined using PCR-based methods ranges from 41% to 74% whereas that with culture-based methods ranges from 5% to 58%. At the species level, the prevalence rate of C. canimorsus in Taiwan was lower than that reported in the United States (21.7%) (3), Britain (24%) (20), and Switzerland (58%) (13). All of these countries are at relatively higher latitudes than Taiwan. Thus, climatic factors may also affect the distribution of Capnocytophaga species. Cold temperatures are likely to be more suitable for adapting C. canimorsus to the canine oral cavity (21). Further, C. canimorsus has been shown to adapt to the canine oral cavity by foraging glycans from salivary mucin and Nlinked glycoproteins. Increased mucin production can be induced under cold conditions through activation of the transient receptor potential channel melastatin 8 in oral epithelial cells and the tongue (22). Taiwan is located in a subtropical region, and the average annual temperature during specimen sampling was 23.7°C. This relatively high temperature may not be suitable for the presence of C. canimorsus, as implied in a study conducted in Brazil wherein the detection rates of C. canimorsus and C. cynodegmi in dogs with periodontal disease using a PCR-based method were 19% and 66.9%, respectively (23). Additionally, the prevalence of C. canimorsus and C. cynodegmi in dogs in Shiraz (Iran) was reported as 16% and 28.8%, respectively (24). The average annual temperature in both Brazil and Shiraz is approximately 25°C, similar to that in Taiwan. Thus, factors affecting the adaptation of Capnocytophaga spp. to the canine oral cavity need to be investigated further to better understand the epidemiology of Capnocytophaga.
Overall, nucleotide polymorphisms in some C. cynodegmi strains are reported to impact the accuracy of species-specific PCR methods worldwide. Our results indicate that the newly developed 16S rRNA PCR-RFLP method is a more accurate approach for distinguishing between C. canimorsus and C. cynodegmi. After verification, our method demonstrated high accuracy, especially for human C. canimorsus isolates. Combining the developed genus-and species-specific PCR with our RFLP system provides an improved molecular strategy for the epidemiological investigation of Capnocytophaga in small animals and serves as a diagnostic tool for human infections caused by animal bites.

MATERIALS AND METHODS
Specimen collection, culture, and isolation of Capnocytophaga. Specimens were collected from the oral cavity of dogs brought to the Veterinary Medicine Teaching Hospital (VMTH), National Chung Hsing University (NCHU), or to a regional animal shelter from January to September 2020. Oral specimens were obtained with sterile cotton swabs by firmly rubbing the gingival margin and tongue of each dog in a conscious state. Basic information, including age, sex, breed, recent history of antibiotic use, main diet, and stage of periodontal disease was recorded. This study was approved by the Institutional Animal Care and Use Committee (approval no. 109-064). All owners provided informed consent.
Each swab was immediately plated onto a selective medium, prepared using Columbia agar (Becton, Dickinson, and Company, Franklin Lakes, NJ, USA) with 10% sheep's blood (Taiwan Prepared Medium, Taipei, Taiwan), a polyvitaminic supplement (Supplement VX; Becton, Dickinson and Company, Franklin Lakes, NJ, USA), and an antibiotic mixture containing 2.5 mg/mL trimethoprim (Sigma-Aldrich, St. Louis, MO, USA), 5 mg/ Identification and phylogenetic analysis of Capnocytophaga isolates. Isolated and purified colonies (see previous section) subsequently underwent genus-specific and/or species-specific PCR to confirm that the selected colonies were Capnocytophaga sp. The primers and PCR conditions used were previously described by Suzuki et al. (5). Colonies with positive PCR results were propagated for further DNA extraction and the enriched bacterial cells from the subcultured plates were stocked in cryogenic vials with preservation beads (Taiwan Prepared Media, Taipei, Taiwan) and stored at 280°C. DNA extraction was performed using an AxyPrep bacterial genomic DNA miniprep kit (Axygen Scientific Inc., Union City, CA, USA) according to the manufacturer's instructions. The 16S rRNA gene was subsequently amplified from extracted DNA and sequenced for species identification. To validate isolates not recognized by genus-specific or species-specific PCR and for colonies that exhibited the morphological features of Capnocytophaga but had negative PCR results, confirmation was performed using 16S rRNA sequences.
Amplification of the 16S rRNA gene from previously extracted DNA was performed using primers 27F (59-AGAGTTTGATCCTGGCTCAG-39) and 1492R (59-GGTTACCTTGTTACGACTT-39) (25). The thermal protocol was as follows: initial denaturation at 94°C for 2 min; 30 cycles of denaturation at 94°C for 15 s, annealing at 55°C for 15 s, and extension at 72°C for 90 s; and final extension at 72°C for 10 min in a MiniAmp Plus Thermal Cycler (Thermo Fisher Scientific, Waltham, MA, USA). Approximately 1,465 bp of the amplified products was analyzed by electrophoresis on a 1% agarose gel. The PCR products were pretreated using ExoSAP-IT PCR Product Cleanup Reagent (Thermo Fisher Scientific) and sequenced using an ABI 3730 automated DNA sequencer (Applied Biosystems, Rockville, MD, USA). Sequences were compiled using SeqMan software (version 7.0; DNASTAR, Inc., Madison, WI, USA) and submitted to the GenBank database. The accession numbers of the 16S rRNA sequences of the cultured Capnocytophaga species isolated from the canine oral cavity are MZ314747 to MZ314793, MZ314818, and MZ314826 to MZ314827. The reference strains retrieved from GenBank included three strains of C. canimorsus (ATCC 35979, CP022382, and NC015846), three strains of C. cynodegmi (ATCC 49044, AY643076, and CP022378), one strain of C. felis (accession no. LC411961), one strain of C. canis (accession no. LC100036.1), and one strain of C. stomatis (CP022387).
The 16S rRNA sequences were aligned using ClustalW in MEGA-X software (26). A phylogenetic tree was constructed using the neighbor-joining method, and genetic distances were calculated using the Kimura 2-parameter model with 1,000 bootstrap replications. Phylogenetic analysis and tree construction were conducted using MEGA-X software.
Biochemical tests. Biochemical tests were performed using a Biolog GEN III MicroPlate test panel (Biolog Inc., Hayward, CA, USA) according to the manufacturer's instructions. Briefly, bacteria were enriched on blood agar plates and homogenously mixed with IF-C inoculation fluid until the turbidity reached 65%. The inoculate was added to the microplate and cultured at 37°C in a candle jar. Colorchange analyses and interpretation of the results were conducted using Biolog Microbial Identification Systems software (Biolog Inc.) at 24 and 48 h after incubation. The system compared the test strain results against an established database and used similarity indices to determine the species ID; otherwise, the genus ID or no ID was shown.
Oxidase and catalase tests were performed on each isolated Capnocytophaga strain. The catalase tests were performed using 3% hydrogen peroxide solution (Union Chemical, Hsinchu, Taiwan) and the oxidase tests were performed using oxidase reagent (Becton, Dickinson and Company).
16S rRNA PCR-RFLP. The aforementioned 16S rRNA gene sequence alignment of the isolates and reference strains indicated differences between C. canimorsus and C. cynodegmi. The 16S rRNA gene sequences of C. canimorsus and C. cynodegmi strains with high concordance in their variance sites were identified. We then used SeqMan software (version 7.0; DNASTAR, Inc.) to search for restriction enzymes that could differentiate among the divergent sequences. We selected an RFLP system with the most distinctive restrictive patterns and validated it in silico. The REs KpnI and PpuMI were finally chosen for the RFLP system. The RE reaction mixture was prepared according to the manufacturer's instructions and contained 1 mL amplified PCR product, 2 mL Green Buffer, 1 mL KpnI, and 1 mL PpuMI (Thermo Fisher Scientific). Digestion was performed at 37°C for 10 min followed by inactivation at 80°C for 5 min. Restriction fragments were analyzed by electrophoresis on 1.5% agarose gels.
In silico verification of predicted RE sites in published C. canimorsus and C. cynodegmi strains. We retrieved all published 16S rRNA gene sequences of C. canimorsus and C. cynodegmi from GenBank and the RefSeq data bank of NCBI (27). The sequence search filter used the default settings and sequence length was limited from 200 to 1,600 bp. The results classified by taxon showed 275 matched sequences of C. canimorsus and 84 matched sequences of C. cynodegmi. After phylogenetic analysis, whole-genome shotgun sequences, sequences shorter in length than the RE position, and other sequences that were too diverse from the 16S rRNA gene were removed. A phylogenetic tree was subsequently constructed based on comparable 16S rRNA sequences of the published C. canimorsus and C. cynodegmi strains. The method used for phylogenetic tree construction was the same as that described in the previous section "Identification and phylogenetic analysis of Capnocytophaga isolates." Statistical analysis. The association between individual factors (including sex, age, weight, main diet, tooth-brushing habits, antibiotic history, periodontal disease stage, and whether the dogs were owned or sheltered) and the presence of Capnocytophaga spp. was analyzed using Fisher's exact test. Differences were considered significant at P , 0.05. Data analysis was performed using the Statistical Analysis System version 9.4 (SAS Institute Inc., Cary, NC, USA).
Data availability. The authors confirm that the data supporting the findings of this study are available within the article and its supplemental material.