Gut microbiota and age shape susceptibility to clostridial enteritis in lorikeets under human care

Background Enteritis is a common cause of morbidity and mortality in lorikeets that can be challenging to diagnose and treat. In this study, we examine gut microbiota in two lorikeet flocks with enteritis (Columbus Zoo and Aquarium—CZA; Denver Zoo—DZ). Since 2012, the CZA flock has experienced repeated outbreaks of enteritis despite extensive diet, husbandry, and clinical modifications. In 2018, both CZA and DZ observed a spike in enteritis. Recent research has revealed that the gut microbiota can influence susceptibility to enteropathogens. We hypothesized that a dysbiosis, or alteration in the gut microbial community, was making some lorikeets more susceptible to enteritis, and our goal was to characterize this dysbiosis and determine the features that predicted susceptibility. Results We employed 16S rRNA sequencing to characterize the cloacal microbiota in lorikeets (CZA n = 67, DZ n = 24) over time. We compared the microbiota of healthy lorikeets, to lorikeets with enteritis, and lorikeets susceptible to enteritis, with “susceptible” being defined as healthy birds that subsequently developed enteritis. Based on sequencing data, culture, and toxin gene detection in intestinal contents, we identified Clostridium perfringens type A (CZA and DZ) and C. colinum (CZA only) at increased relative abundances in birds with enteritis. Histopathology and immunohistochemistry further identified the presence of gram-positive bacilli and C. perfringens, respectively, in the necrotizing intestinal lesions. Finally, using Random Forests and LASSO models, we identified several features (young age and the presence of Rhodococcus fascians and Pseudomonas umsongensis) associated with susceptibility to clostridial enteritis. Conclusions We identified C. perfringens type A and C. colinum associated with lorikeet necrohemorrhagic enteritis at CZA and DZ. Susceptibility testing of isolates lead to an updated clinical treatment plan which ultimately resolved the outbreaks at both institutions. This work provides a foundation for understanding gut microbiota features that are permissive to clostridial colonization and host factors (e.g. age, prior infection) that shape responses to infection. Supplementary Information The online version contains supplementary material available at 10.1186/s42523-021-00148-7.

The majority of lorikeets had a clinical history of sudden weight loss, diarrhea, or sudden death. 225 Most often, gross findings consisted of marked muscle wasting with a prominent keel (Fig. 3b), 226 and multiple severely dilated loops of intestines with thin walls and scant or watery contents 227 and/or gas (Fig. 3c). Other gross findings included tan opaque viscous coelomic fluid, intestinal 228 segments impacted with friable dark red contents, and thickened tan mesentery with multiple 229 small, white, firm to soft nodules within the intestines as well as on the serosa (Fig. 3d). 230 231 By histopathology, the most common finding (96 % of cases) was an intraluminal coagulum 232 comprised of red blood cells, bacterial colonies and sloughed necrotic mucosa (Fig 3e). 233 Approximately two thirds (64 %) of the cases had full thickness heterophilic and/or 234 granulomatous enteritis with ulceration (Fig 3f). Approximately two thirds of the cases (61 %) 235 had marked intestinal loop dilation, with over one third (39 %) having villous fusion and/or 236 blunting. Thirty six percent had fibrosis within the intestinal wall, most often at sites of 237 transmural inflammation and necrosis. The most commonly associated lesion was granulomatous 238 and/or heterophilic coelomitis (86 %) with half of those cases having intracoelomic bacteria, 239 most frequently bacilli or mixed bacteria. Other common lesions associated with necrotizing 240 enteritis included mild to moderate renal tubular necrosis and/or mineralization most likely due 241 to dehydration or septicemia (50 %) and marked extramedullary hematopoiesis within the liver 242  Eighty-nine percent of cases had large gram-positive bacilli present within the necrohemorrhagic 245 coagulum and these gram-positive bacilli were also the most abundant bacteria present, followed 246 by gram-positive cocci (71%), and gram-negative bacilli (54%) (Fig. 3g), whereas 57 % of the 247 transmural enteritis lesions contained large gram-positive bacilli, followed by 25 % with gram-248 positive cocci and 21 % with gram-negative bacilli. intestinal samples were positive for any of the following toxin genes: cpb (beta), etx (epsilon), itx 265 (iota), cpe (enterotoxin), or netB (necrotic B-like). Additionally, although the gene encoding 266 cpb2 (beta-2 toxin) was identified in intestinal isolates, we did not find cpb2 in the FFPE 267 enteritis-positive samples ( Table 4) To determine whether C. perfringens could also be found in lorikeets with enteritis at other 272 institutions, we collected cloacal swabs and intestinal content from lorikeets at the DZ between 273 November 2018 and May 2019. We identified 12 birds that died or were euthanized due to 274 enteritis, and sampled these birds at necropsy ("post-mortem"). These birds were then age-, sex-, 275 and species-matched as closely as possible to 12 healthy lorikeets that were sampled during a 276 flock survey in May 2019 when all birds were reported to be healthy (Additional File 10). 277 Similar to the CZA lorikeets, we observed decreased microbial diversity and altered microbial 278 composition in the post-mortem lorikeets with enteritis as compared to healthy birds (Observed 279 Features Kruskal-Wallis p < 0.0005, Weighted UniFrac PERMANOVA p = 0.001, Fig. 4; 280 Additional File 11). C. perfringens was also significantly increased in relative abundance in the 281 post-mortem birds (Fig. 4c, included in these models along with demographic variables including lorikeet age and species. 312 Sex was not included as it was unknown for 31% of the birds. Seventy-five percent of the 313 samples were used as a training set and 25% of the samples were used as a test set. The RF 314 model identified the relative importance of variables as predictors (Additional File 12,c) while 315 the LASSO model identified whether a variable was associated with susceptibility or true health 316 (Additional File 12,d). We then collated variables that were identified in both the RF and 317 LASSO models (Fig. 5c). The top 26 variables included 23 taxa associated with Susceptible 318 birds and 1 taxon (family Peptostreptococcaceae) associated with True Healthy birds. Rainbow 319 lorikeets (as opposed to coconut lorikeets) were also associated with health while the 320 "WasQuarantined TRUE" variable was associated with susceptibility. This variable represented 321 young lorikeets (< 1 year old) that were transferred from another institution; these birds 322 underwent an initial quarantine prior to integration with the CZA flock. Some of the taxa 323 predictive of susceptibility included: Rhodococcus fascians, Kocuria spp., Pseudomonas 324 umsongensis, two taxa in the family Enterobacteriacea and an Aeromonas spp. 325 326

Dietary analysis for trypsin inhibitors 327
Finally, we examined lorikeet diets in relation to C. perfringens susceptibility. Several C. 328 perfringens toxins, including cpa, cpb, pfo, and cpb2 (the toxin observed in 5 CZA C. 329 perfringens isolates) are sensitive to the host-produced protease trypsin [39]. However, trypsin 330 inhibitors present in the diet can block the activity of trypsin and thereby increase the risk of C. 331 perfringens toxin-mediated enteritis. As lorikeets are nectivores and their main diet under human 332 care consists of reconstituted powdered nectar, we opted to test trypsin inhibitor levels in six 333 commercial nectar brands including brands used at CZA and DZ. The range of trypsin inhibition 334 for the nectars was 0-1.79 trypsin inhibitor units (TIU)/mg dry nectar, denoting relatively low 335 inhibition (Fig. 6). For reference, raw soybeans, which have been linked to C. perfringens toxin-336 mediated enteritis in poultry, contain approximately 46 TIU/mg, and soy protein concentrate 337 contains 9.45 TIU/mg [40]. As such, the low levels of trypsin inhibition detected in commercial 338 nectars are unlikely to be playing a major role in susceptibility to C. perfringens enteritis in 339 lorikeets; although we cannot rule out the possibility that other supplementary food items (e.g. 340 sweet potatoes or legumes) may have contributed to toxin-mediated clostridial enteritis. Notably, 341 C. colinum toxins have yet to be characterized; therefore, the role of trypsin and dietary trypsin 342 inhibitors on C. colinum pathogenesis is unknown. 343 344

Discussion 345
Our initial goal in this study was to characterize the gut microbiota of sick and healthy lorikeets 346 with the hypothesis that a dysbiosis was driving susceptibility to enteritis. While we did identify 347 gut microbial alterations associated with susceptibility, we also ended up identifying the 348 probable etiologic agents of enteritis in both the CZA and the DZ lorikeet flocks. Specifically, 349 we observed increased relative abundances of C. perfringens and C. colinum (CZA lorikeets 350 only) in the 16S rRNA sequencing data. We then cultured lorikeet intestinal contents and 351 identified, genotyped, and susceptibility-tested multiple C. perfringens isolates. A 352 histopathologic examination of intestinal tissues further revealed inflammation, necrosis, and 353 ulcerative lesions that also contained gram-positive bacilli consistent with clostridial enteritis and 354 specifically C. colinum or C. perfringens. IHC and toxinotyping of intestinal tissues confirmed 355 the presence of C. perfringens in lorikeets with enteritis. PCR testing also confirmed the 356 presence of C. colinum in CZA lorikeets with enteritis. We then compared the gut microbiota of 357 healthy CZA lorikeets that developed enteritis to healthy CZA lorikeets that never developed 358 enteritis during our 10-month sampling period, and we identified several features associated with 359 susceptibility to enteritis including: age (younger birds are more susceptible), and increased 360 relative abundances of Rhodococcus fascians, Pseudomonas umsongensis, two taxa in the family 361 Enterobacteriacea, and an Aeromonas spp., among others. This work allowed us to identify the 362 probable causative agents of lorikeet enteritis at two zoos, develop an optimal treatment plan 363 based on genotyping and susceptibility testing, and profile healthy birds at high risk of clostridial 364 enteritis. 365 366 Demographics of lorikeet enteritis 367 C. perfringens has been linked to enteritis in multiple mammal and bird species [34,39,[41][42][43][44], 368 including in lorikeets and other psittacines [5,32,33,[45][46][47]. In this study, young lorikeets (< 2 369 years old) were more likely to develop clostridial enteritis. We also observed some differences in 370 microbial composition by age (Additional File 5), and age emerged as a predictor of 371 susceptibility in the Random Forests model (Additional File 12,c). Previous reports in other 372 avian species note that the immunological naivete of young birds may increase susceptibility to 373 C. perfringens while adult birds are more resistant [48,49]. We also found that coconut lorikeets 374 were more likely to develop enteritis as compared to other lorikeet species. It is less clear what 375 may be driving species differences in clostridial enteritis; although, there were more coconut 376 lorikeets than any other species at CZA, and C. perfringens has been associated with necrotic 377 enteritis in coconut lorikeets at other institutions [47]. While type A was the dominant C. 378 perfringens toxinotype reported in the previous study on lorikeets, toxinotype C was also 379 common. Toxinotypes B, D, E, F, and G were also observed but less common [47]. Sex has also 380 been reported as a factor that influences susceptibility to necrotic enteritis in birds [41]; although, 381 we did not observe differences by sex in this study. Taken together, our results suggest that both 382 microbial and immunological factors may contribute to clostridial enteritis in young lorikeets. Lorikeets with enteritis demonstrated microbial community shifts and histopathological changes 398 as compared to healthy lorikeets. Based on 16S rRNA sequencing, culture, and genotyping, we 399 confirmed the presence and increased relative abundance of C. perfringens type A in CZA and 400 DZ lorikeets with enteritis. In the CZA lorikeets, we further determined that C. perfringens was 401 directly associated with necroulcerative intestinal enteritis via IHC and multiple intestinal 402 isolates of C. perfringens contained the cpb2 toxin. C. colinum was also found at increased 403 relative abundances in CZA but not DZ birds based on 16S sequences and PCR. In both CZA 404 and DZ birds, we observed decreased microbial diversity and altered microbial composition in 405 lorikeets with enteritis as compared to healthy lorikeets. C. perfringens and C. colinum have 406 been implicated previously in necroulcerative enteritis in birds including lories, lorikeets, and 407 poultry [47,[59][60][61][62][63]. Microbial community alterations have also been reported in chickens 408 infected with C. perfringens [64][65][66][67]. While several clostridial species are considered normal 409 flora in some avian species including poultry, in psittacines, clostridial species are rarely found 410 in the intestines of healthy birds, and taxa such as C. colinum  Susceptible birds displayed minor but significant differences in microbial community diversity 453 (Observed Features), composition, and differentially abundant taxa. These microbial community 454 differences could be linked to age as young birds were also the most susceptible. Although we 455 found no significant difference in microbial diversity or composition in healthy lorikeets by age, 456 birds in the youngest age group (< 2 years old) had the greatest microbial diversity (Additional 457 File 4, c,f) which was also true in the Susceptible birds (Fig. 5a). Increased microbial diversity 458 has also been observed in young chickens susceptible to C. perfringens infection as opposed to 459 those that were more resistant [92]. Moreover, differences in microbial composition by age 460 (Unweighted UniFrac, but not Weighted UniFrac) were significant in healthy birds (p = 0.01 461 Additional File 5). This suggests that age may influence lorikeet microbial community structure, 462 and with a larger sample size, this may have been more apparent. We identified several microbes that were associated with susceptibility including Rhodococcus 476 fascians, Pseudomonas umsongensis, an Aeromonas spp., and two taxa in the family 477 Enterobacteriacea. Rhodococcus fascians has been found at increased abundances in juvenile 478 birds (sparrows, < 1 year old) as compared to older birds and could be an age-related taxa [21]. A 479 single human case report also highlights a co-infection between R. fascians and a clostridial 480 species (C. difficile) [95], which leads to the intriguing question as to whether these 2 species 481 interact in ways that may support each other's growth. However, this co-infection was in an 482 immunocompromised individual, so the relevance is unclear. Both Pseudomonas and Aeromonas 483 species have been independently associated with enteritis in birds [86,96,97]. In a previous 484 study that employed a subcutaneous abscess model, the addition of Pseudomonas aeruginosa or 485 various Enterobacteriaceae species enhanced the growth of C. perfringens [98] suggesting that 486 interactions between these taxa may indeed facilitate clostridial infections. Our RF and LASSO 487 models also identified several other microbial taxa associated with susceptibility; although, the 488 potential role these taxa may be playing in clostridial infections or enteritis is undetermined and 489 requires additional study. 490

491
This study had several limitations: We identified both C. colinum and/or Type A C. perfringens 492 in lorikeets with enteritis; however, the mechanisms by which these bacteria caused disease 493 remain unclear. For example, while C. colinum has been empirically and experimentally linked 494 to ulcerative enteritis in birds, its virulence factors have yet to be elucidated [60]. Second, 495 although both C. perfringens alpha toxin (cpa) and beta-2 toxin (cpb2) have been associated with 496 enteritis in multiple host species including birds, the role of these toxins in enteritis pathogenesis 497 is ambiguous, and both of these toxin genes have been found in the intestines of healthy animals 498 [39,60,99]. It is possible that neither cpa nor cpb2 are key virulence factors in Type A. C. 499 perfringens and that other unidentified virulence factors played a role in lorikeet enteritis. 500 Additionally, toxin gene presence (e.g. PCR, used in this study) does not necessarily equate to 501 toxin gene expression. However, clostridia and its respective toxin genes are considered aberrant 502 in healthy psittacines [4,31,59,68,69], suggesting that they are playing a role in enteritis even 503 if their virulence factors are not fully defined. The surrounding gut microbiota and metabolites 504 could also be mediating C. perfringens pathogenesis including colonization and toxin expression 505 as has been demonstrated in Clostridiodes difficile [35,36]  In this study, we systematically examined gut microbiota and susceptibility to clostridial enteritis 509 in two lorikeet flocks under human care. A few of our key take-aways: 1) Clostridia can be 510 challenging to detect via culture in lorikeet cloacal swabs, but anaerobic culture of intestinal 511 contents yielded C. perfringens in 6 out of 13 isolates from CZA, and 16S sequencing allowed 512 ready identification of C. perfringens and C. colinum in birds with enteritis. As clostridia are not 513 normal inhabitants in psittacines, this was a significant finding. 2) Clostridial isolates then 514 underwent genotyping and susceptibility testing, which allowed us to update the lorikeets' 515 clinical treatment plans to more targeted therapies, aligned with antimicrobial stewardship 516 practices (DZmetronidazole, CZA -florfenicol and clindamycin in clinically affected birds 517 and prophylactic flock-wide application of bacitracin). Since June of 2019, and as of this 518 writing, there have been no new cases of enteritis in lorikeets at either CZA or DZ. 3) Young age 519 (potentially linked with immunological naivete, limited exposures, or lower trypsin activity 520 [100]), prior enteritis, and specific microbes including R. fascians, P. umsongensis, and 521 Enterobacteriacea taxa are linked with susceptibility to enteritis, and these microbes could be 522 promoting clostridial infections by establishing a niche conducive to colonization in a yet-to-be 523 determined manner. 4) Dietincluding trypsin inhibitorscan also influence susceptibility to 524 clostridial enteritis. Although, commercial nectars were low in trypsin inhibitors, we cannot rule 525 out the possibility that other supplementary food items (e.g. sweet potatoes or legumes) could 526 have contributed to toxin-mediated enteritis. Clostridial enteritis, and C. perfringens in 527 particular, not only affects lorikeets, but can also cause devastating losses in the poultry industry 528 (commonly Type G C. perfringens with NetB toxin), and lead to gastrointestinal disease in 529 humans and other mammalsdepending on the toxinotype. This work provides a foundation for 530 understanding gut microbiota features that are permissive to clostridial colonization and host 531 factors (e.g. age, prior infection) that shape responses to infection. The 16S rRNA sequences were processed, filtered, and analyzed using QIIME 2 version 2020.11 581 [101] and DADA2 [102]. Taxonomic assignment of amplicon sequence variants (ASVs) was 582 performed using the Greenegenes 13_8 database with 99% sequence identity cutoff. (Note: We 583 also performed taxonomic assignment with SILVA 132 and found that, in this case, Greengenes 584 provided more specific taxonomic assignments, particularly in the Clostridia taxa.) A total of 246 585 samples from CZA and 30 samples from DZ were submitted for sequencing. Samples with fewer 586 than 1000 reads were removed from analyses including 23 CZA samples and 6 DZ samples. 587 This resulted in a total of 223 CZA samples and 24 DZ samples that were used in our analyses. 588 After filtering, we obtained a total of 3,236,674 reads from the CZA samples (average: 13,911 589 reads per sample; range: 1003 to 68,537 reads) and 264,769 reads from the DZ samples (average: 590 9,026 reads per sample; range: 1922 to 26,766 reads). Sequences identified as mitochondria, 591 chloroplasts, or eukaryotic reads were removed. Based on an examination of negative controls, 592 we also identified the following taxa as contaminants and removed them from analyses: a taxa in 593 the order RF39 (Mollicutes phyla); a taxa in the genus Allobaculum, a taxa in the genus Massilia; 594 Haemophilus parainfluenzae; Prevotella copri; a taxa in the genus Sphingomonas; a taxa in the 595 genus Bradyrhizobium; Pseudomonas viridiflava; and a taxa in the genus Thermicanus. 596 597

Nectar preparation 676
To eliminate assay interference due to free fatty acids, all nectars were first defatted through a 677 hexane (Thermo Fisher Scientific, Waltham, MA) extraction. Nectars were combined with three 678 times their volume of pure hexane and mixed for one minute. The samples were then allowed to 679 sit for 10 minutes to allow for a separation of layers, and the top hexane-fat layer was removed. 680 This process was repeated a total of three times for each nectar. Defatted nectars were then 681 allowed to dry overnight in a fume hood. Once dry, 1 g of defatted nectar was added to 50 g of 682 0.01M NaOH (Thermo Fisher Scientific, Waltham, MA). The mixture was stirred slowly on a 683 stir plate for 3 hours. Extracts were then centrifuged at 4696 x g, and the supernatant was 684 decanted to produce the final extract. bring it up to temperature, and then the assay reaction was initiated by adding 2 mL pre-warmed 696 trypsin solution (0.02 mg trypsin/mL) to each tube. The tubes were allowed to react for exactly 697 10 minutes at 37 °C, then the reaction was stopped with the addition of 1 mL 30 % acetic acid 698 solution. Samples were allowed to cool to room temperature before measuring absorbance at 410 699 nm with an HP 8453 UV-Vis spectrophotometer (Hewlett Packard, Palo Alto, CA). Absorbance 700 readings were corrected with nectar blanks by mixing all reagents, but adding trypsin solution 701 after the acetic acid to ensure the enzyme was inactive. A positive control sample was also made 702 using 2 mL water in place of nectar and running the assay as delineated above. 703

TIU calculation 705
With the definition that 1 TIU = a decrease in 0.01 absorbance compared to a positive control 706 sample, TIU/mg could be calculated as follows: 707 TIU = (positive control sample absorbancenectar sample absorbance) / 0.01 708 TIU/mg = TIU/ [Nectar concentration(mg/mL) x 10 mL assay solution] 709 To compare trypsin inhibitor concentrations (TIU/mg) between nectars, we applied a one-way 710 ANOVA followed by pairwise Tukey's tests. 711 712

Statistical analyses 713
We compared the number of lorikeets that ever had enteritis versus the number of lorikeets that 714 never developed enteritis by age, sex, and species using a  2 test [104]. In cases where groups 715 had a frequency less than 5, we use the Yates'  2 correction. For cases in which a group 716 contained zero individuals (e.g. 0 females), we used the Freeman-Halton extension of the 717 Fisher's exact test. To compare average age across groups, we used a Kruskal-Wallis test after 718 testing for normality using a Shapiro-Wilk test. For microbial community analyses, alpha 719 diversity was compared between groups using observed features, the Shannon diversity metric, 720 and the Kruskal-Wallis test. Beta diversity was evaluated using permutational multivariate 721 analysis of variances (PERMANOVAs) between groups on Bray-Curtis distance matrices. All 722 alpha and beta diversity p-values were corrected for multiple comparisons using the Benjamini 723 Hochberg false discovery rate (FDR) correction. A p-value < 0.05 was considered significant. 724 Differential abundances of microbes by status (healthy, enteritis, susceptible)    microbial composition that predicts enteritis. Healthy lorikeets that never developed enteritis 851 throughout the sampling period were identified as "True Healthy" while healthy birds that 852 developed enteritis at least once during the sampling period were identified as "Susceptible." 853 "Enteritis" represents birds with enteritis that were sampled while they were clinically ill. No