Ethics statement

All California sea lion samples were collected under authority of Marine Mammal Protection Act Permits No. 932-1905-00/MA-009526 and No. 932-1489-10 issued by the National Marine Fisheries Service (NMFS), NMFS Permit Numbers 17115–03, 16087–03, and 13430. The sample collection protocol was approved by the Institutional Animal Care and Use Committees (IACUC) of The Marine Mammal Center (Sausalito, CA; protocol # 2008–3) and the University of California Los Angeles (ARC # 2012-035-12. UCLA is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. The Marine Mammal Center and UCLA adhere to the national standards of the U.S. Public Health Service Policy on the Humane Care and Use of Laboratory Animals and the USDA Animal Welfare Act. Isoflurane gas was used to anesthetize all wild-caught, free-ranging sea lions for sampling. All island fox and skunk samples were collected by the National Park Service under USFWS permit TE-08267-2.

Study animals and sample collection

Our dataset comprises samples from California sea lions (Zalophus californianus), island foxes (Urocyon littoralis), and island spotted skunks (Spilogale gracilis) with confirmed infections of L. interrogans serovar Pomona. Samples were collected from 107 sea lions that had stranded along the central California coast between 2004–2017 and were admitted to The Marine Mammal Center (TMMC; Sausalito, California) for rehabilitation. An additional thirty sea lion samples were collected from free-ranging wild sea lions from the central California coast and northern Oregon, between 2010 and 2012, as described in Prager et al., 2020 [33]. The majority of sea lions were diagnosed with acute leptospirosis (97/137) based on clinical signs, serum chemistry results, and necropsy data [34].

Samples from island foxes (n = 59) and island spotted skunks (n = 4) were collected between 2011 and 2016 during annual grid and target trapping conducted by the National Park Service (NPS) as part of a monitoring program on Santa Rosa Island, California. Santa Rosa Island has an area of approximately 214 km² and only three terrestrial mammal species (island foxes, island spotted skunks, and island deer mouse [Peromyscus maniculatus]), and has no known history of Leptospira circulation before our study. Fox and sea lion data include both sexes and all age classes. All four skunks were adult males.

Sample analysis

All animals included in this study had real time polymerase chain reaction (rt-PCR) confirmed Leptospira DNA in urine or kidney tissue as described by Wu et al [35], and the infecting Leptospira serovar was confirmed as L. interrogans serovar Pomona using PFGE (Fig A in S1 Text) as described previously by Galloway & Levett [22] on all cultured isolates (N_CSL = 19, N_fox = 11, N_skunk = 2).

Serum samples were tested by microagglutination test (MAT) against a panel of five Leptospira serovars comprising L. interrogans serovars Pomona (serogroup Pomona), Autumnalis (serogroup Autumnalis), Djasiman (serogroup Djasiman), Bratislava (serogroup Australis), and Icterohaemorrhagiae (serogroup Icterohaemorrhagiae). Most of the samples included in this analysis were tested against more than five serovars (56 CSL samples and 7 fox samples were tested with a 20-serovar panel). We exclude tested serovars that yielded almost entirely negative or very low results for all host species, and serovars for which the overlap between tested samples was low among the host species. All titers used in the host species comparison were analyzed at the Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia using MAT (as described in Prager et al [36]) and run to endpoint dilution. Titer results were log-transformed for ease of interpretation using the following formula: log₂(titer/100) + 1, thus a titer of 1:100 = 1, 1:200 = 2, 1:400 = 3, etc. Titers reported as <1:100 are represented by 0.

In a separate analysis focusing on variability among laboratories, a subset of 46 fox sera were MAT analyzed at three reference laboratories using a 2-serovar panel (Pomona and Autumnalis). The laboratories are referred to as Laboratories A, B, and C. Antibody titers against serovar Pomona were evaluated to endpoint at all three laboratories. Serovar Autumnalis was not titrated to endpoint for all samples at all laboratories. At Laboratory A, 43 of 46 samples were titrated to endpoint and 3 of 46 were only tested at a dilution of 1:100 (all were positive). At Laboratory B, all 46 samples were titrated to endpoint. At Laboratory C, all 46 serum samples were titrated to a 1:6400 dilution (log₂ titer = 7) but not beyond.

Data selection and analysis

To analyze antibody cross-reactivity patterns within and between host species, we selected MAT results from animals for which there was at least one positive urine PCR or culture result, which confirms current Leptospira infection. We did separate analyses for animals with PCR- or culture-confirmed infection (n_CSL = 137; n_fox = 59; n_sku = 4) and animals with confirmed infection and PFGE-confirmed serovar (n_CSL = 19; n_fox = 11; n_sku = 2). Only two skunk samples were PFGE-positive, so we included samples from all four skunks in both analyses. We also performed an additional comparison of PCR- or culture-confirmed infections in skunks with all MAT-positive skunks to confirm that patterns were consistent. For individuals that had been sampled longitudinally, we selected the MAT result from the serum sample with a collection date closest to that of the positive urine sample. The majority of MAT results from foxes (55/59) and all from skunks (4/4) were from sera collected on the same day as the Leptospira PCR- or culture-positive urine. Sea lion serum samples used for MAT were collected within 5 days of the date that the PCR- or culture-positive urine or kidney sample was collected (range = 0–5 days, median = 0 days). To analyze relative titer magnitudes among host species, we standardized antibody titers by dividing a given antibody titer by the highest antibody titer detected against any serovar in the 5-serovar MAT panel for that host serum sample.

We tested differences in the titer profiles across species using a multivariate nonparametric analysis of similarities test (ANOSIM; [37]). We used non-parametric Kruskal-Wallis tests to evaluate differences in titer magnitude among groups (i.e., species or serovar; [38]). When appropriate, a Wilcoxon rank sum test with a Bonferroni correction was used to assess pairwise differences between groups.

We evaluated a subset of 46 fox serum samples at three certified testing laboratories as described above (see section on Sample Analysis) to compare MAT results across laboratories. Fox serum samples were chosen for this laboratory comparison based on MAT titer results from Laboratory A. For each MAT antibody titer level ranging from 1:100–1:51200, three serum samples with that MAT antibody titer against serovar Pomona, as reported by Laboratory A, were selected where possible (Table A in S1 Text). In addition to these 30 samples, we included a further 10 samples that had no detectable antibodies against serovars Pomona and Autumnalis at Laboratory A, and six samples that had no detectable antibodies against serovar Pomona but were MAT positive against serovar Autumnalis at Laboratory A. We tested for differences among laboratories using a Kruskal-Wallis test and between laboratories using a pairwise Wilcoxon rank sum test.

Lesson 1: Highest titer does not always indicate infecting serovar or even serogroup

In our study, antibody titers detected in sea lions and skunks were generally highest against serovar Pomona, while foxes typically had the highest titer against serovar Autumnalis (Figs 1 and 2, and Figs A and B in S1 Text), despite our genetic evidence showing that the infections were caused by serovar Pomona. Notably, serovars Pomona and Autumnalis are in different serogroups [20]. These results highlight that the serovar against which the highest titer is detected should not be assumed to be the infecting serovar, and that even the more cautious approach of interpreting the highest titer as the presumptive serogroup of the infecting agent can lead to erroneous conclusions. Misidentification of the infecting serovar or serogroup could result in a misunderstanding of multi-species transmission patterns with implications for disease management and control.

Lesson 2: Seronegativity must be interpreted with caution

Our longitudinal samples show that antibody titers against the infecting serovar can decay below the level of detection before those against non-infecting serovars do. Thus, a seronegative result against a given serovar does not necessarily mean it was not the infecting serovar, even when juxtaposed with positive titers against other serovars. This phenomenon could lead to misclassifying the infecting serovar if we rely on MAT for strain identification, or mistakenly ruling out the serovar that caused the infection, especially if exposure occurred in the relatively distant past.

Lesson 3: Absolute and relative titer magnitudes depend on host species

We observed significant differences in both absolute and relative MAT titer magnitudes among the three host species tested. The same infecting serovar of Leptospira gave rise to different MAT cross-reactivity profiles in different host species (Fig 1). In general, we see that sea lions have higher median titers than foxes, which in turn have higher median titers than skunks across the five serovars (Fig 2). Autumnalis is a notable exception for which sea lions and foxes exhibit a similar median titer magnitude. The mechanisms underlying these differences are unknown.

Lesson 4: Absolute and relative titer magnitudes can differ across laboratories

We observed systematic differences in absolute and relative titer magnitudes among three certified testing laboratories, including qualitative differences in which serovars elicited the highest titers from the same samples (Fig 4). Despite adherence to excellent laboratory standards and protocols, the nature of the MAT testing process means that some variation among laboratories is bound to exist. MAT is not standardized among laboratories, and variation both within and between laboratories is expected [41]. Many factors that are difficult to control can contribute to the variation of MAT results, so caution is needed when comparing MAT titers across laboratories. If such a comparison is necessary, ideally a set of samples should be run at both laboratories to calibrate the results.

Implications for interpretation of MAT results

Of the more than 300 pathogenic Leptospira serovars currently described, most diagnostic MAT panels select a maximum of 20 serovars. In fact, cost and time restrictions typically limit panels to 4–6 serovars or fewer, particularly under conditions with fewer resources and lower testing capacity. This leaves the distinct possibility that a circulating serovar (and possibly the infecting serovar) could be omitted from the MAT panel, leading to potential for sub-optimal diagnostics and misunderstanding of circulating strains and transmission linkages. In other words, the serovar (or even serogroup) associated with the maximum titer in a given panel is not necessarily the infecting agent.

Titer magnitudes are often used to assess active infections. However, given our finding of differences in absolute titers across host species, relying on titer thresholds inferred from data in one species to identify recent or active infections in another can lead to inaccurate diagnosis and poor incidence estimates. For example, longitudinally sampled sea lions acutely infected with L. interrogans serovar Pomona had initial log₂ titers against serovar Pomona ranging from 10 to 12 (i.e. 1:51200 to 1:102400) and these titers declined with a half-life of around 17 days [33]. Therefore if log₂ titer thresholds used to define active infection in humans– 3 (or 1:400)–or dogs– 4 (or 1:800)–were applied to sea lions, many would be miscategorized as current infections [15,19]; this could occur even if the infecting serovar was not included in the MAT panel and the sea lion titers arose from antibody cross-reactivity. Our longitudinal fox data show that foxes could be similarly miscategorized if the human or dog thresholds are applied to them, as some foxes infected with L. interrogans serovar Pomona persist above the log₂ titer = 4 threshold for years (Fig E in S1 Text). It is essential that any efforts to interpret absolute titers are calibrated to the system under study. When this is done, titer magnitudes (and their decay) can be used to estimate the recency of infection [16,42–44]. Modern titer kinetics approaches have the potential to include additional host-specific information about the relationship among serovars (i.e., the cross-reactivity profile of MAT titers) to estimate time since infection and improve our understanding of when outbreaks may have occurred.

We know of no prior work showing host species differences in MAT profiles. These patterns may be driven by different major histocompatibility complex (MHC) types and diversity [45,46], but more work is needed to understand how immunogenetic differences among wildlife may impact serology. It is noteworthy that the island fox population recently underwent a severe population bottleneck and exhibits very low genetic diversity (and therefore MHC diversity; [47]). Yet recent work in coyotes in southern California revealed a similar pattern–with MAT titers against serovar Autumnalis frequently exceeding those against serovar Pomona, despite known circulation of serovar Pomona in coyotes–suggesting that this effect may occur more broadly among some canids [48]. Systematically expanding surveillance across canid species and beyond could provide insights on the possible existence of a host phylogenetic effect on MAT reactivity.

It is possible that some interspecies variation in titer magnitude was due to sampling bias. Over two-thirds of sea lion samples were from animals experiencing acute leptospirosis–the disease caused by Leptospira infection. By contrast, foxes and skunks were sampled during a routine trapping program aimed at monitoring these sensitive populations, so sample collection was not biased by disease severity. This could skew our observed antibody titers higher in the sea lions as their severe clinical disease suggests a recent infection [49], but a modeling analysis of island fox titers estimated peak titers against serovar Pomona of 6 to 9 on our log₂ scale [16], consistent with values reported here for foxes, and lower than values reported for sea lions. It is clear that there is a large degree of immunological variability within and between species.

Variability in titer magnitudes has been documented across reference laboratories [50]. The International Leptospirosis Society sponsors the annual International Proficiency Testing Scheme for the Leptospirosis MAT, intended to provide information on the quality of MAT testing and improve MAT testing performance worldwide [51]. Early rounds of this program reported a wide variety of titers for the same sample and serovar [51]. Although multifactorial, variation is probably driven chiefly by two main factors. First, MAT relies on live bacterial cultures, and there may be slight strain variations among laboratories and among different batches grown within a laboratory. For trustworthy MAT results, within-culture serovar identity must be verified regularly [33,51]. Secondly, determining antibody titers by assessing agglutination under dark-field microscopy is subjective and requires significant expertise; even with best practices, some observer effect is inevitable. Altogether, many factors that are difficult to control can contribute to the variation of MAT results, so caution is needed when comparing MAT titers across laboratories. Ultimately, a difference of one serial dilution may not be a true difference at all given that convalescent testing requires a 4-fold rise to be deemed significant [32].

Overinterpretation of individual titer values can lead to misrepresentation of host relationships and circulating strains. This begs the question: given that genetic methods can provide clearer classification of pathogen strains and are becoming more accessible, why use MAT at all? Serology has many benefits that are distinct from culture-based methods with respect to duration of positivity and the potential to learn from antibody titer kinetics (with appropriately cautious interpretation). Despite issues with serological cross-reactivity and requirement of specialized laboratory and trained personnel, MAT is generally more affordable and sensitive than PCR or culture-based methods, is often an easier sample to collect, and captures information on past infections. Acute and convalescent MAT is also considerably more sensitive than organism-detection tests. Given the broad accessibility and continuing worldwide use of this diagnostic, we need to interpret its results with appropriate caution, while capitalizing on all available information. There may be an opportunity to improve assessment of the infecting serovar by exploiting consistent patterns in cross-reactivity against serovars within a host species, but more research is needed to describe these patterns within and across host species. The rising availability and falling cost of genetic methods, coupled with exciting new developments in obtaining whole genome sequences of Leptospira without culture isolates, point to a future where genetic typing adds clarity and certainty to Leptospira epidemiology and ecology.