Molecular and serological diagnosis of multiple bacterial zoonoses in febrile outpatients in Garissa County, north-eastern Kenya

Abstract

Bacterial zoonoses are diseases caused by bacterial pathogens that can be naturally transmitted between humans and vertebrate animals. They are important causes of non-malarial fevers in Kenya, yet their epidemiology remains unclear. We investigated brucellosis, Q-fever and leptospirosis in the venous blood of 216 malaria-negative febrile patients recruited in two health centres (98 from Ijara and 118 from Sangailu health centres) in Garissa County in north-eastern Kenya. We determined exposure to the three zoonoses using serological (Rose Bengal test for Brucella spp., ELISA for C. burnetti and microscopic agglutination test for Leptospira spp.) and real-time PCR testing and identified risk factors for exposure. We also used non-targeted metagenomic sequencing on nine selected patients to assess the presence of other possible bacterial causes of non-malarial fevers. Considerable PCR positivity was found for Brucella (19.4%, 95% confidence intervals [CI] 14.2-25.5) and Leptospira spp. (1.7%, 95% CI 0.4-4.9), and high endpoint titres were observed against leptospiral serovar Grippotyphosa from the serological testing. Patients aged 5-17 years old had 4.02 (95% CI 1.18-13.70, p-value = 0.03) and 2.42 (95% CI 1.09-5.34, p-value = 0.03) times higher odds of infection with Brucella spp. and Coxiella burnetii than those of ages 35-80. Additionally, patients who sourced water from dams/springs, and other sources (protected wells, boreholes, bottled water, and water pans) had 2.39 (95% CI 1.22-4.68, p-value = 0.01) and 2.24 (1.15-4.35, p-value = 0.02) times higher odds of exposure to C. burnetii than those who used unprotected wells. Streptococcus and Moraxella spp. were determined using metagenomic sequencing. Brucellosis, leptospirosis, Streptococcus and Moraxella infections are potentially important causes of non-malarial fevers in Garissa. This knowledge can guide routine diagnosis, thus helping lower the disease burden and ensure better health outcomes, especially in younger populations.

Figure 1

Comparison of bacterial community compositions in patient samples based on metagenomic sequencing. The relative abundance of each bacterial phylum is shown, as identified using kraken2 and further re-estimated using a Bayesian approach with bracken. The data are presented as percentages of the total bacterial community in each sample and “Other” comprised bacteria that had less than 5% relative abundance.

Figure 2

Alpha diversity scores of the metagenomic profiles, calculated using the observed, Chao1, and Shannon indices. These indices represent a measure of species richness and evenness in the datasets, which were obtained through shotgun metagenomic sequencing of patient sera.

Figure 3

Classified genus- and species-specific reads per million total reads in the metagenomic analyses. To enhance the visualisation and avoid computing log₁₀ of zero, a pseudo count of 1 was added to the original data and the transformed values were expressed as log₁₀. Therefore, 4 on the heat map scale should be interpreted as 3 in the original data for instance (or 1000 reads per 1,000,000 total reads).

Figure 4

A scatter plot summarizing the number of BLASTn-confirmed reads for the top twenty fever-causing agents in the metagenomic datasets. Each dot represents one agent and its coordinates are the number of reads confirmed against the number of total reads BLAST’ed. The x-axis is capped at 1000 reads to show only the agents that reached the cap, while the y-axis shows the number of confirmed reads. The plot suggests that many agents were confirmed by a low number of reads, and highlights the agents that were confirmed by the most reads.