We investigated bacterial zoonoses (Brucella, C. burnetii, and Leptospira) in 216 febrile patients who visited two health centres in Garissa County using both targeted and untargeted metagenomic detection methods. High PCR positivity was observed for Brucella and Leptospira spp. High endpoint titres were observed against leptospiral serovar Grippotyphosa. Young patients had higher odds of seropositivity to Brucella spp. and C. burnetii than their older counterparts. Lastly, AMR genes and other bacteria such as Streptococcus were found using metagenomic sequencing. Results of this survey reveal multiple causes of non-malarial fevers in Garissa County. This calls for their inclusion in routine diagnosis of non-malarial fevers to lower disease burden.
Brucellosis is an important cause of non-malarial fevers in Kenya and prevalence at the community level is high. The PMr in our study of 13.9% was comparable to other studies of febrile patients in the region 10. The prevalence of animal brucellosis is considerably lower than that of humans in the area 11, and households with one seropositive animal have been shown to have higher odds of having a seropositive human 12. The monitoring of animal brucellosis through One Health surveillance and control strategies is therefore necessary for lowering the human brucellosis prevalence because exposure is usually from livestock and animal products.
A previous study on fever patients in the region showed 19.1% exposure by IgG antibodies against C. burnetii, an estimate much lower than that in our study 13. However, cross-reactivity can occur when patients have been exposed to Bartonella spp., Legionella spp., or Chlamydia spp. 14,15, thereby overestimating the seroprevalence figures.
Higher leptospirosis PMr has been observed in febrile patients in Kenya 16, and 3/12 (25%) patients were Leptospira IgM positive by ELISA in a 2005 outbreak of acute febrile illness in the region 7. The exposure of 3.7% of fever patients therefore demonstrates that leptospirosis should be considered in the differential diagnosis of acute non-malarial fevers in the region.
Our study also determined that younger patients (5–17 years) had higher seroprevalence estimates and were at higher odds of being exposed to Brucella and C. burnetii than their older counterparts between 35 and 80 years. This may be because younger people come into contact with infected animal hosts more regularly than those aged 35–80, through any activities in the production, slaughter, processing, and retail of animal source foods. These professional activities and consumption habits also have gender-specific considerations which may play a role in this risk factor 17. Healthcare-seeking behaviour is also higher in younger people with febrile illnesses in Kenya than in older patients 18, and our observation could therefore have arisen from selection bias. However, community exposure in the country is higher in older patients 12.
Patients that sourced water from unprotected wells had lower odds of exposure to C. burnetii, and tended towards significantly lower odds of exposure to brucellae than those getting water from dams, springs, and other sources. Even though exposure to C. burnetii through contaminated environmental water has been documented, it is an uncommon transmission pathway and in need of further elucidation 19. Contamination of water sources with animal waste can lead to the spread of C. burnetii20. Brucellae can contaminate water sources when animal waste pollutes water sources, and the bacteria can survive for long periods in water, thereby also posing a risk of human exposure 21,22. Therefore, the protection of water sources from possible contamination is vital in reducing human exposure to zoonoses.
Our study also showed a seroprevalence of 3.7% for leptospires in fever patients. This is lower than estimates observed in fever patients in neighbouring Tanzania 23. Serovars Pyrogenes, Sejroe, Bataviae, Tarassovi and Icterohaemorrhagiae were found in the patients, and Grippotyphosa showed particularly high endpoint titres. Even though high titres against serovars other than the infective one (paradoxical reactions) can happen in acute leptospirosis, these are likely important serovars responsible for human leptospirosis in Garissa and MAT panels should include them.
Reads for Streptococcus were confirmed by BLASTn analyses in the metagenomic sequencing and these were also not observed in the negative control used, implying this could be a cause of non-malarial fevers in some patients. The metagenomic datasets generated have the potential to reveal AMR genes which are clinically relevant and cannot be obtained with 16S rRNA amplicon sequencing. Results from shotgun metagenomic experiments can be confounded by external contaminants at various stages of the workflows. These can originate from extraction and library prep reagents 24 and various bacteria have been identified as common contaminants 25,26. The effect of these contaminants is more pronounced in low biomass samples such as serum, as the sample is inundated by contaminating nucleic acid due to the low amount of starting material from the patient, consequently generating misrepresentative results 25. In addition, index hopping and cross-contamination by carry-over of amplicons from previous runs are sources of internal contamination 24 and can bias metagenomics studies. Even though some reads were found in our NEC, they could not be assigned to the top 20 leading fever agents, validating our results as free from internal contamination. We also extracted circulating cell-free DNA which according to the manufacturer may not be detected with spectrophotometric methods such as the Qubit™ assay. There is also growing consensus that a blood biome exists for healthy individuals and may be constituted largely by bacteria speculated to be transient and from sources such as the gastrointestinal tract, skin and oral microbiomes. The putative blood biome is largely constituted by Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes27 which were also the largest bacterial phyla found in this study. Therefore, the identification of these bacteria does not offer definitive proof that they are the causative agents of non-malarial fevers. Lastly, a few important AMR genes were determined in some patients. The gene coverage and coverage depth for most was not high, and these results could have originated from low abundance 28. We therefore cannot establish whether these genes were clinically relevant. However, studies utilising higher sequencing depth could add value in monitoring AMR in fever patients using metagenomic sequencing.
Clinical signs and symptoms for the different pathogens were not distinctive, with a selected few being associated with positive patients using both serology and PCR tests. Future studies making use of the gold standard diagnostic tests can therefore determine the positive predictive values of these clinical signs and symptoms to aid physicians in the region in the diagnosis of these non-malarial fevers. Most of the Brucella isolated from humans in Kenya has been B. melitensis9, as is the case in other countries 29, a likely result of B. melitensis being more prevalent than B. abortus. B. melitensis is often considered more pathogenic than B. abortus due to its higher association with human brucellosis. However, human brucellosis resulting from B. abortus can be equally severe 30, and preventative measures that lower exposure from cattle, sheep and goats should be adopted in the country.
Our study had several limitations. We utilised shallow metagenomic sequencing due to limited resources which may lead to decreased sensitivity in detecting low concentrations of circulating DNA in clinical samples. Our study also had a relatively small sample size of inpatients and this can lead to failure to detect significant effects of putative risk factors in causing disease exposure.
In conclusion, Garissa County is an ASAL area with a high poverty index. As all patients recruited in this study were malaria-negative, investments in the diagnosis and treatment of zoonotic diseases are key in lowering disease burden, improving health outcomes and increasing the productivity of human capital. Prevention of exposure through occupational risks or consumption of contaminated foods, especially to younger individuals, should be considered to lower the high exposures observed. The presence of multiple zoonotic pathogens in febrile patients presenting in local health centres shows the urgent need for surveillance and control programs in the county. Using non-targeted sequencing approaches can add value in detecting uncommon causes of febrile illnesses, especially when recently collected samples are obtained and appropriate contamination controls are included in the sequencing process. Lowering the burden of these diseases in animal hosts is cost-effective. Investments in the local health centres for routine screening of these zoonoses, and sensitisation of medical professionals on the importance of non-malarial fevers in the region should also be carried out to ensure proper use of anti-malarial and antibiotic therapies. Protection of water sources from animal waste should also be encouraged to minimise zoonotic disease exposure.