Introduction
Zoonoses account for 61% of infectious diseases of humans and 75% of emerging infectious diseases (Taylor et al., Reference Taylor, Latham and Woolhouse2001; Woolhouse and Gowtage-Sequeria, Reference Woolhouse and Gowtage-Sequeria2005). Endemic zoonoses, characterised by a widespread distribution and frequent transmission between animals and humans, include some of the most important diseases of poverty (Maudlin et al., Reference Maudlin, Eisler and Welburn2009; ILRI, 2012). However, in the poorest communities in the developing world, a lack of surveillance and control (Halliday et al., Reference Halliday, Daborn, Auty, Mtema, Lembo, Bronsvoort, Handel, Knobel, Hampson and Cleaveland2012) and misdiagnosis (Maudlin et al., Reference Maudlin, Eisler and Welburn2009) result in these diseases being amongst the most neglected in the world (ILRI, 2012). It has been suggested that, in terms of human health impacts, livestock productivity loss and amenity to agricultural intervention, leptospirosis, a neglected endemic zoonosis causing an estimated 1 million cases of human disease annually (Costa et al., Reference Costa, Hagan, Calcagno, Kane, Torgerson, Martinez-silveira, Stein, Abela-ridder and Ko2015), is the second most important zoonosis globally (ILRI, 2012). Although rodents are frequently implicated as the key reservoir hosts of Leptospira (Picardeau, Reference Picardeau2017), recent evidence suggests that, in rural Africa, livestock may also be significant reservoir hosts (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018).
Leptospira are phylogenetically delineated into 22 species, 10 of which are pathogenic, and further divided into more than 300 serovars which may demonstrate specific, but not absolute, host preferences (Picardeau, Reference Picardeau2017). In Madagascar, four Leptospira species have been identified in terrestrial small mammals (Rahelinirina et al., Reference Rahelinirina, Léon, Harstskeerl, Sertour, Ahmed, Raharimanana, Ferquel, Garnier, Chartier, Duplantier, Rahalison, Cornet and Harstkeerl2010; Dietrich et al., Reference Dietrich, Wilkinson, Soarimalala, Goodman, Dellagi and Tortosa2014; Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) and a molecular link demonstrated between a recent acute, severe case of human leptospirosis (Pagès et al., Reference Pagès, Kuli, Moiton, Goarant and Jaffar-Bandjee2015) and small mammals (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). However, a recent household cross-sectional serosurvey in Madagascar identified contact with cattle as the only significant risk factor [OR = 3, 95% CI (1.03–10.03)] for human exposure despite serological typing based on Microscopic Agglutination Testing (MAT) identifying Icterohaemorrhagiae, a serogroup traditionally associated with rodents, as the predominant serogroup (Ratsitorahina et al., Reference Ratsitorahina, Rahelinirina, Michault, Rajerison, Rajatonirina and Richard2015).
Therefore, understanding the role of livestock in the epidemiology of leptospirosis is critical to understanding public health risks and informing control strategies. Vaccination of livestock is used to both improve productivity and to reduce zoonotic risk from livestock (Marshall and Chereshsky, Reference Marshall and Chereshsky1996). However, recent studies have demonstrated that commercial leptospirosis vaccines may lack efficacy when challenged by autochthonous livestock strains in Brazil (Sonada et al., Reference Sonada, de Azevedo, Soto, da Costa, de Morais, de Souza, Gonçales, Miraglia and Vasconcellos2018) and will not reduce human infections from serovars associated with wildlife (Thornley et al., Reference Thornley, Baker, Weinstein and Maas2002). Therefore, to inform control measures, we used abattoir sampling and molecular methods to determine the prevalence and diversity of Leptospira in livestock in Madagascar and their phylogenetic relationship to Leptospira identified in livestock from similar studies in other developing countries.
Materials and methods
Abattoir sampling
In 2015, we estimated Leptospira prevalence by sampling a total of 205 livestock: 25 cattle and 25 pigs sampled at each of three abattoirs around the capital, Antananarivo, and 30 cattle and 25 pigs sampled at the abattoir in the town of Moramanga. Livestock presented at abattoirs were sourced from a wide geographic area (Supplementary Fig. S1). Whole kidneys and urine aspirates were collected and cooled for transportation to a laboratory.
Sample preparation
As Leptospira may not be homogeneously distributed within kidney tissue and urinary shedding is intermittent, to advise future studies/surveillance we compared results from three sample preparations: a single kidney excision, a pool of four kidney excisions from different lobes (cattle) or anatomical locations (pigs) and urine. Kidney samples were stored in 95% ethanol and urine samples at −80 °C.
(i). Kidney samples
Extraction of the kidney samples was performed on 40 mg of tissue using the DNeasy Blood and Tissue Kit (Qiagen) as per the manufacturer's instructions with the volumes of buffers ATL and AL increased to account for the increased amount of tissue extracted and the volume of elution buffer reduced to 100 µL to concentrate the DNA.
For preparations from a single lobe/location, 40 mg of tissue from the corticomedullary junction was extracted. For pooled samples of four kidney lobes, 40 mg of tissue from each lobe was placed in a sterile 1.5 mL Eppendorf tube containing 400 µL of the ATL lysis buffer and ground with a glass pestle to obtain a homogenate. One hundred microliters of the homogenate were then placed in a new sterile 1.5 mL Eppendorf and extracted as outlined above. To monitor inhibition, 1.5 µL of DNA extraction control 560 (Bioline) was added to each extraction.
(ii). Urine samples
After slaughter urine samples were obtained by aspiration of 2 mL of urine from the bladder using sterile needles and 2 mL syringes. If samples arrived at the laboratory within 24 hours they were transferred to a clean cryotube and stored at −80 °C. For the Moramanga abattoir, this was not possible. Consequently, samples were centrifuged at 7000 rpm for 20 min before discarding the supernatant and adding 200 µL of TE buffer (Sigma-Aldrich). The samples were then kept at +4 °C, before transferring to the laboratory in a cool box with ice packs.
Extraction of the urine samples was performed using the same protocol as for the kidney samples with the exception of the following pre-extraction step. Samples frozen at −80 °C were incubated at 67 °C for one hour before 1 mL was transferred to a 2 mL Eppendorf. Samples were then centrifuged at 7000 rpm for 10 min. After discarding the supernatant, the pellet was suspended in 200 µL of TE buffer. The re-suspended pellet was centrifuged again at 7000 rpm for 10 min and the supernatant discarded. The pellet was then re-suspended again with 100 µL of TE buffer, 50 µL of lysozyme (10 µg mL−1), 50 µL of mutanolysin (4 KU mL−1), 4 µL of lysostaphin and incubated at 37 °C for one hour. After addition of 20 µL of proteinase K and 180 µL of AL buffer (Qiagen) and vortexing, samples were incubated at 56 °C for 10 min. The final step involved the addition of 200 µL of 95% ethanol and vortexing before following the extraction protocol used for kidney samples.
Leptospira detection and DNA sequencing
Prevalence estimates were obtained using a 16s (rrs) qPCR (Smythe et al., Reference Smythe, Smith, Smith, Dohnt, Symonds, Barnett and McKay2002) as described previously (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). Each sample preparation (single lobe/location, four lobes/location and urine sample) was tested separately. An individual animal was identified as infected if any of the sample preparations tested positive. χ2 tests were used to test for differences in prevalence estimates between cattle and pigs and between prevalence estimates from different sample preparations within the same host. Mann–Whitney–Wilcoxon tests were used to determine whether parasite load, as measured by 16s qPCR amplification threshold (Ct), affected typing success. Statistical analyses were performed using R version 3.4.1 software (R Core Team).
Amplification of 200–300 bp of the lfb1 gene (Merien et al., Reference Merien, Portnoi, Bourhy, Charavay, Berlioz-Arthaud and Baranton2005; Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) was then performed on positive kidney and urine sample from each animal and on positive livestock samples previously identified in Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018). Initially, samples were tested using lfb1 primers (Merien et al., Reference Merien, Portnoi, Bourhy, Charavay, Berlioz-Arthaud and Baranton2005) designed to detect all pathogenic Leptospira. Previous molecular studies in Africa have identified L. borgpetersenii, L. kirschneri and a L. kirschneri-like species in livestock (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018) and suggest that existing lfb1 primers fail to amplify L. borgpetersenii in some cases (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). Therefore, redesigned lfb1 forward primers targeting L. kirschneri and L. borgpetersenii (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) were subsequently used on samples that tested negative using the standard lfb1 qPCR. In addition, to test for mixed infections, samples infected with L. borgpetersenii were tested using L. kirschneri targeting primers and vice versa. All lfb1 assays were performed using reaction conditions as previously described (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) and all amplicons were sequenced.
To facilitate comparisons with other studies, amplification and sequencing of the secY gene, a widely used target for phylogenetic analysis of Leptospira (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008), was undertaken. Initial amplification of a ~450 bp fragment was performed using reaction conditions previously described (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018). Where initial amplification failed to amplify sufficient material for sequencing, a second round of amplification of a ~350 bp fragment was performed in a 25 µL reaction volume using a nested forward primer [5′-AATCCATTYTCYCARATYTGGTA-3′] and the first round reverse primer at concentrations of 0.5 µ m, 12.5 µL of MyTaq Red mix 2x (Bioline Reagents Ltd), 9.5 µL of molecular grade water and 1 µL of first round product. The thermal profile comprised initial denaturation at 95 °C for 3 min, followed by 20 cycles of denaturation at 95 °C for 30 s, annealing at 50 °C for 30 s and extension at 72 °C for 1 min, with a final extension at 72 °C for 7 min.
DNA sequence analysis
Phylogenetic analysis of lfb1 sequences was supplemented with lfb1 sequences obtained from small mammals in Madagascar (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018), a goat opportunistically sampled during village-based rodent surveys and four cattle sampled during a pilot study in Antananarivo. In addition, to provide additional lfb1 sequences from the region and to test for mixed infections, the same lfb1 assays were used to obtain sequences from 28 livestock sampled in a similar abattoir study in northern Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018) and typed using secY sequencing. To identify the serovars to which these strains are most closely related, lfb1 sequences were queried against the NCBI refseq_genomes and nr/nt database using the blastn algorithm and identical or closely related (>99% identity) records with associated serovar information reported.
To place the strains identified in this study in a global context, secY sequences were analysed alongside sequences from similar livestock studies from Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018), Reunion (Guernier et al., Reference Guernier, Lagadec, Cordonin, Le Minter, Gomard, Pagès, Jaffar-Bandjee, Michault, Tortosa and Dellagi2016) and Brazil (Hamond et al., Reference Hamond, Pestana, Medeiros and Lilenbaum2015; Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019) (Supplementary Table S1). Reference secY sequences (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008) with 100% identity to any of the secY genotypes were included in the phylogenetic analysis. Sequence alignment and phylogenetic analysis were performed using MEGA7 (Kumar et al., Reference Kumar, Stecher and Tamura2016).
Results
Prevalence estimates and effect of sample preparation
Cattle had a significantly higher overall prevalence than pigs (19 vs 5%, χ2 = 8.17, P < 0.01) (Table 1). Cattle had 13 infections detected in kidney samples and 11 in urine samples, with only four individuals testing positive for both sample types. Pooled kidney excisions detected more infections than single excisions (12 vs 6%) (Table 1), although the difference was not significant (χ2 = 1.11, P = 0.29), and only one individual tested positive by the single excision but negative by the pooled excision. Of the positive pigs, only one infection was detected in the kidney (single excision), compared to four infections detected in urine.
Table 1. Prevalence of Leptospira infection using different sample preparations from cattle and pigs
Cattle kidneys were sampled at four lobes and pig kidneys were sampled at four locations (pig kidneys are not lobed). Infection prevalence and 95% confidence intervals (logit method) were determined using the binom package (https://cran.r-project.org/ package = binom). Samples with evidence of inhibition were excluded from prevalence estimate calculations for each sample preparation but each individual animal had at least one sample preparation without evidence of inhibition so every animal was represented in overall prevalence estimates
Typing success
In Madagascar, interpretable lfb1 and secY sequences were obtained from 50% (10/20) and 40% (8/20) of 16s qPCR positive cattle samples respectively. All secY sequences were obtained from samples in which an lfb1 sequence had previously been amplified. In kidney samples, parasite load, as measured by 16s qPCR Ct, had an effect on sequencing success with samples from which sequencing data was obtained having a higher parasite load (median 16s Ct = 33.31, range 32.38–33.99) than samples in which sequencing was unsuccessful (median 16s Ct = 37.85, range 33.95–39.51) (W = 2, P = 0.002). No such effect was noted in urine samples (median 16s Ct = 35, range 34–37 vs median 16s Ct = 37.4, range 33–39) (W = 13.5, P = 0.18). None of the 16s qPCR positive pig (n = 5, 16s Ct range 37–39) samples yielded interpretable lfb1 or secY sequences. Samples from Tanzania identified to be infected with an unknown L. kirschneri-like Leptospira species based on secY sequences (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018), failed to satisfactorily amplify using any of the lfb1 primer combinations used.
Genetic diversity and mixed infections
Based on lfb1 sequences, we identified L. borgpetersenii and L. kirschneri in cattle and L. interrogans in a goat sample in Madagascar (Fig. 1). Including lfb1 sequences from both Madagascan and Tanzanian samples, five L. borgpetersenii lfb1 clades (A–E) were identified in cattle, two of which (clade A and clade D) were shared between both countries. L. borgpetersenii clade A sequences were identical to serovar Hardjo-bovis strain L550, a livestock associated strain (Bulach et al., Reference Bulach, Zuerner, Wilson, Seemann, McGrath, Cullen, Davis, Johnson, Kuczek, Alt, Peterson-Burch, Coppel, Rood, Davies and Adler2006). However, L. borgpetersenii clade D had no lfb1 homologue. Of the remaining three L. borgpetersenii clades, two (clade C and clade E) were present in Madagascar, Clade E sequences were identical to lfb1 sequences previously obtained from small mammals (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) and the clade C sequence was identical to a strain previously identified as serovar Tarassovi. The remaining L. borgpetersenii clade, detected only in Tanzania, was identical to serovar Hardjo-bovis strain JB197 (Bulach et al., Reference Bulach, Zuerner, Wilson, Seemann, McGrath, Cullen, Davis, Johnson, Kuczek, Alt, Peterson-Burch, Coppel, Rood, Davies and Adler2006).
Fig. 1. Maximum likelihood phylogenetic tree of 171 bp lfb1 sequences (accession numbers MK244296-MK244341) using the Kimura 2-Parameter model with a uniform distribution. Samples from Madagascar and Tanzania are highlighted in bold and the number of animals from which sequences were obtained is indicated. Reference sequences are labelled by Leptospira species, serovar and accession number. Nodes are labelled with bootstrap support.
L. kirschneri lfb1 sequences, obtained from the kidney and urine samples of a single animal from the Moramanga region of Madagascar, demonstrated 99% identity to sequences previously obtained from endemic small mammals (Hemicentetes semispinosus) (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) in the same region and the L. interrogans sequence obtained from a goat was identical to a genotype previously identified predominantly in black rats (Rattus rattus) (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) and to that obtained from an acute human case of leptospirosis (Pagès et al., Reference Pagès, Kuli, Moiton, Goarant and Jaffar-Bandjee2015).
Using the lfb1 primers targeting different Leptospira species, mixed infections were detected in two cattle in Madagascar and one in Tanzania. In Madagascar, L. kirschneri and L. borgpetersenii clade E were detected in a urine sample from one animal and L. borgpetersenii clade A was detected in the urine sample and L. borgpetersenii clade E in the kidney sample from another animal. In Tanzania, L. kirschneri and L. borgpetersenii clade A were detected in a single kidney sample.
Phylogenetic relationship to Leptospira from other livestock studies in the developing world
Analysis of secY sequences from this study and other studies in northern Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018), Reunion island (Guernier et al., Reference Guernier, Lagadec, Cordonin, Le Minter, Gomard, Pagès, Jaffar-Bandjee, Michault, Tortosa and Dellagi2016) and Brazil (Hamond et al., Reference Hamond, Pestana, Medeiros and Lilenbaum2015; Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019) that used similar molecular approaches identified five recognized pathogenic Leptospira species (L. borgpetersenii, L. santarosai, L. interrogans, L. noguchii, L. kirschneri) and one L. kirschneri-like species infecting cattle (Fig. 2). In Madagascar, four L. borgpetersenii secY clades (A–D) were identified. With the exception of clade B, all remaining clades contained secY sequences from cattle from other regions of the world. Both clades A and D were detected in cattle from Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018), Reunion (Guernier et al., Reference Guernier, Lagadec, Cordonin, Le Minter, Gomard, Pagès, Jaffar-Bandjee, Michault, Tortosa and Dellagi2016) and Brazil (Hamond et al., Reference Hamond, Pestana, Medeiros and Lilenbaum2015; Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019). Clade A sequences were identical to L. borgpetersenii serovar Hardjo-bovis strains JB197 and L550, strains associated with cattle (Bulach et al., Reference Bulach, Zuerner, Wilson, Seemann, McGrath, Cullen, Davis, Johnson, Kuczek, Alt, Peterson-Burch, Coppel, Rood, Davies and Adler2006), which lfb1 sequencing had been able to differentiate in Tanzanian samples (Fig. 1: L. borgpetersenii clade A and B). However, clade D sequences matched no reference strains, with sequences in clade B providing the closest match (98% identity). Sequences in clade C were detected in a single animal in this study and a single animal in Reunion (Guernier et al., Reference Guernier, Lagadec, Cordonin, Le Minter, Gomard, Pagès, Jaffar-Bandjee, Michault, Tortosa and Dellagi2016) and were identical to reference sequences from strains previously associated with small mammals (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008). Moreover, the corresponding lfb1 sequence from Madagascan cattle (Fig. 1, L. borgpetersenii clade E) was identical to numerous lfb1 sequences obtained from Madagascan small mammals (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). Although the single L. kirschneri sequence from this study was novel, it was closely related to sequences obtained from cattle in the Brazilian Amazon (Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019) and to L. kirschneri serovar Galtoni, which was obtained from a cattle kidney sample in Argentina (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008), differing by a single synonymous polymorphism. L. santarosai and L. noguchii, which have been identified in Brazil (Hamond et al., Reference Hamond, Pestana, Medeiros and Lilenbaum2015; Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019) and a L. kirschneri-like species recently identified in both Brazil (Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019) and Tanzania (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018) were not identified in Madagascan cattle.
Fig. 2. Maximum likelihood phylogenetic tree of 329 bp secY sequences (accession numbers MK244342-MK244344 and MK610269-MKMK610274) using the Tamura 3-Parameter model with a discrete Gamma distribution. Samples from Madagascar are highlighted in bold with the number of animals from which sequences were obtained and a representative accession number indicated. Sequences from similar livestock studies are labelled by country of origin, host, number of sequences and accession number of a representative sequence. Reference sequences (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008) are labelled by Leptospira species, serovar and accession number. Nodes are labelled with bootstrap support.
Discussion
We show that cattle have a higher prevalence of Leptospira infection than pigs in Madagascar and are potential reservoir hosts for a diversity of Leptospira with genetic similarities to Leptospira strains identified in cattle in other tropical regions of the world, some of which are genetically distinct from reference strains. Increased probability of detection of Leptospira infections in sample preparations that included multiple kidney lobes or urine rather than single kidney lobes, suggest a localized distribution for Leptospira infections in cattle kidneys or low Leptospira loads in single samples. These findings support previous studies (Guedes et al., Reference Guedes, de Araújo, de Souza, de Souza Silva, Taniwaki, Cortez, Brandão and Heinemann2019), where prevalence estimates from urine samples (14.9%, 31/208) were higher than those obtained from kidney samples (5.8%, 12/208), and highlight the importance of considering sample collection methods when evaluating prevalence estimates. Leptospira sequencing success was dependent on parasite load and the inability to obtain Leptospira sequence data from pigs was likely due to the low parasite load in these samples.
The identification of strains in Madagascar, Tanzania, Reunion and Brazil, such as L. borgpetersenii clade D (Fig. 2), genetically distinct from any reference strains, has implications for the control and diagnosis of leptospirosis in the developing world. In Brazil, it has been demonstrated that the efficacy of commercial vaccines is poor in the face of autochthonous strain challenge (Sonada et al., Reference Sonada, de Azevedo, Soto, da Costa, de Morais, de Souza, Gonçales, Miraglia and Vasconcellos2018) and it is likely that commercial vaccines are likely to perform as poorly in Madagascar and other developing countries. Moreover, the inclusion of local isolates in antigen panels for serological assays, such as MAT, can dramatically increase the serological detection of infections (Mgode et al., Reference Mgode, Machang'u, Mhamphi, Katakweba, Mulungu, Durnez, Leirs, Hartskeerl and Belmain2015). Therefore, the identification of strains with no associated reference strains suggests that existing antigen panels might underestimate the prevalence of human infections.
In addition to strains associated with livestock, we also identified the infection of cattle with strains previously detected in small mammals. For example, in Madagascar, a L. interrogans strain common in small mammals (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018) and identified in an acute human case of leptospirosis (Pagès et al., Reference Pagès, Kuli, Moiton, Goarant and Jaffar-Bandjee2015) was also identified in a goat. Moreover, secY sequencing identified a L. borgpetersenii strain previously detected in small mammals (Victoria et al., Reference Victoria, Ahmed, Zuerner, Ahmed, Bulach, Quinteiro, Harstkeerl and Hartskeerl2008) in cattle in this study and in Reunion (Guernier et al., Reference Guernier, Lagadec, Cordonin, Le Minter, Gomard, Pagès, Jaffar-Bandjee, Michault, Tortosa and Dellagi2016) and lfb1 sequencing confirmed this strain as identical to L. borgpetersenii previously identified in small mammals in Madagascar (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). Thus, our results suggest transmission of at least some Leptospira strains between small mammals and livestock, and raise the potential that livestock may play a role in the transmission of these strains. For example, due to their large size, livestock could act as amplification hosts for small mammal associated strains by increasing environmental contamination through urinary shedding after incidental infection or could act as reservoir hosts in their own right, possibly explaining why contact with cattle has been identified as a risk factor for human Leptospira infection in Madagascar despite serological typing identifying serogroup Icterohaemorrhagiae, a serogroup associated with rodents, as the predominant serogroup (Ratsitorahina et al., Reference Ratsitorahina, Rahelinirina, Michault, Rajerison, Rajatonirina and Richard2015). In contrast, in Tanzania, where small mammals sampled tested negative for Leptospira infection (Allan et al., Reference Allan, Halliday, Moseley, Carter, Ahmed, Goris, Hartskeerl, Keyyu, Kibona, Maro, Maze, Mmbaga, Tarimo, Crump and Cleaveland2018), no evidence was found for infection of livestock with strains previously associated with small mammals.
We also confirm the presence of mixed infections in livestock in both Madagascar and Tanzania, supporting previous evidence for mixed infections in small mammals (Moseley et al., Reference Moseley, Rahelinirina, Rajerison, Garin, Piertney and Telfer2018). Horizontal genetic transfer plays an important role in the evolution and serological classification of Leptospira (Llanes et al., Reference Llanes, Restrepo and Rajeev2016), and mixed infections within the same host provide the ideal environment for this to take place. Moreover, for Borrelia, it has been proposed that mixed infections may facilitate the maintenance of infection in reservoir hosts (Andersson et al., Reference Andersson, Scherman and Råberg2013). Where serological diagnostic assays, such as MAT, rely on evaluating serological response to specific antigens, mixed infections with strains which may represent different serovars could complicate interpretation. Further research is needed to clarify the role of mixed infections in the evolution and epidemiology of Leptospira.
Our results emphasize that Leptospira epidemiology in tropical, developing country contexts, where close human contact with livestock is more likely and farming systems may promote contact between small mammals and livestock, may be very different to developed country settings. In addition, considering sampling methodology is important when comparing studies and planning surveillance and further work is needed to optimize abattoir sampling strategies. In Madagascar and other developing countries, evidence that livestock is infected with potentially novel Leptospira strains highlights the need for understanding the diversity of Leptospira circulating in livestock to inform diagnostic antigen panels and vaccine development in these regions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0031182019001252.
Acknowledgements
We thank the staff of the Plague Central Laboratory, Institut Pasteur de Madagascar, for technical assistance, especially Fehivola Andriamiarimanana, as well as staff from the abattoirs for their assistance during sampling.
Financial support
This work was supported the Wellcome Trust (Senior Fellowship no. 095171 to S.T., Veterinary Training Fellowship no. 096400/Z/11/Z to K.A., Institutional Strategic Support Fund to the University of Aberdeen no. 2014815/Z/16/Z), the Biotechnology and Biological Sciences Research Council (no. BB/M010996/1), the University of Aberdeen Environment and Food Security theme and Institut Pasteur Madagascar.
Conflict of interest
None.
Ethical standards
Not applicable.
