Triatoma brasiliensis, a triatomine-bug vector of Chagas disease, evolved in the semiarid Caatinga, where it occupies rocky outcrops, shrubby cacti, and human dwellings. Dwellings and rocks are considered high-quality microhabitats for this saxicolous species, whereas cacti probably represent secondary, lower-quality microhabitats. This ‘microhabitat-quality hierarchy’ hypothesis predicts that T. brasiliensis populations occupying dwellings or rocks should endure harsh environmental conditions better than their cactus-living relatives.
We tested this prediction by comparing T. brasiliensis infestation (proportion of microhabitats with bugs), density (bugs per microhabitat), and crowding (bugs per infested microhabitat) in dwellings, rocks, and cacti sampled before and during the extreme drought that ravaged the Caatinga in 2012–2016. We used random-intercepts generalized linear mixed models to account for microhabitat spatial clustering and for variations in bug-catch effort; we assessed model performance and computed model-averaged effect estimates using Akaike’s information criterion. Pre-drought infestation was similar across microhabitat types; during the drought, infestation remained stable in dwellings and rocks but dropped in cacti. Pre-drought bug density declined from dwellings to rocks to cacti; an additional decline associated with the drought was largely comparable across microhabitats, albeit perhaps somewhat larger in cacti. Finally, pre-drought bug crowding was higher in dwellings than in rocks or cacti and changed little during the drought–possibly with a downward trend in dwellings and an upward trend in cacti.
Triatoma brasiliensis populations fared better in dwellings and rocks than in cacti during extreme drought. Estimates of microhabitat and drought effects on infestation, density, and crowding suggest that only a few cacti (versus many rocks and dwellings) represent good-quality habitat under such extremely harsh conditions. Our findings provide empirical support to the microhabitat-quality hierarchy hypothesis, and imply that T. brasiliensis can endure extreme climate by exploiting high-quality microhabitats, whether wild or man-made, in the semiarid Caatinga.
Triatomine bugs, the vectors of Chagas disease, feed on vertebrate blood and occupy diverse microhabitats in close association with their hosts. Each bug species is adapted to particular, ‘primary’ microhabitats. Many species, however, also use ‘secondary’ microhabitats, sometimes including man-made structures. Secondary wild microhabitats are probably of lower overall quality than primary microhabitats, whereas human dwellings can be of higher overall quality if they provide a stable and abundant blood supply. This suggests that there may be a ‘hierarchy’ of triatomine microhabitat quality–top-quality dwellings, high-quality primary microhabitats, and low-quality secondary wild microhabitats. To test this hypothesis, we measured the impact of a severe drought on Triatoma brasiliensis populations living in dwellings and wild habitats, both primary (rocks) and secondary (shrubby cacti). Dwelling and rock bug populations remained largely stable during the drought, whereas most cactus-living populations either became extinct or shrunk to an undetectable size. A few cacti, however, supported relatively large bug colonies even during the drought, suggesting that they were of higher-than-average quality. Our findings provide support to the microhabitat-quality hierarchy hypothesis, and show that T. brasiliensis (and perhaps other triatomines) can endure extreme climate by exploiting (natural or man-made) high-quality microhabitats.
Citation: Ribeiro AC, Sarquis O, Lima MM, Abad-Franch F (2019) Enduring extreme climate: Effects of severe drought on Triatoma brasiliensis populations in wild and man-made habitats of the Caatinga. PLoS Negl Trop Dis 13(10): e0007766. https://doi.org/10.1371/journal.pntd.0007766
Editor: Richard Reithinger, RTI International, UNITED STATES
Received: June 10, 2019; Accepted: September 9, 2019; Published: October 10, 2019
Copyright: © 2019 Ribeiro et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: Instituto Oswaldo Cruz (IOC), Instituto René Rachou (IRR), and Vice-Presidência de Pesquisa e Coleções Biológicas (VPPCB), Fiocruz, Brazil; Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil; this study was funded in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Brazil, Finance Code 001. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Trypanosoma cruzi, the etiological agent of Chagas disease, is transmitted primarily by blood-sucking bugs known as triatomines . Vector-borne transmission is restricted to the Americas, where recent estimates suggest that 4–5 million people carry T. cruzi . American triatomine-bug species, of which over 130 are known, have adapted to highly diverse eco-regions ranging from rainforests to deserts; within them, the bugs associate with their vertebrate hosts in similarly diverse microhabitats–from underground burrows to tree-canopy epiphytes [1,3–7]. Table 1 summarizes a rough, yet broadly accepted, classification distinguishing triatomine species that are primarily terrestrial or primarily arboreal [3,5]. Some populations of a few species can also exploit man-made habitats, where they feed on the blood of people and domestic animals; these ‘synanthropic’ triatomines are the main vectors of human Chagas disease [1,3,6,7].
One distinctive feature of synanthropic triatomine populations is that they can build very high-density colonies [1,3,6–9]. Wild populations, in contrast, are almost always found in relatively small foci, with denser colonies reported only on occasion from particularly suitable microhabitats–e.g., those in which many hiding sites are available and resident vertebrates provide a stable food supply [6,7]. This suggests that there is variation in the quality of the microhabitats a triatomine-bug species can inhabit. Such variation can be viewed as defining a gradient that runs from the many zero-quality habitats that remain unoccupied (despite being available) to the few high-quality habitats that can sustain large breeding colonies [6–10] (Fig 1). This view may be framed as the ‘microhabitat-quality hierarchy’ hypothesis. In general terms, this hypothesis predicts that triatomine-bug populations occupying higher-quality microhabitats should fare overall better than those exploiting lower-quality microhabitats–and, importantly, that the difference should become particularly evident as environmental conditions get harsher [6,7,10–12]. Here, we use a unique dataset on wild and synanthropic Triatoma brasiliensis populations studied under both regular and exceptionally harsh climatic conditions to provide an empirical test of this hypothesis.
The lower panel illustrates the hypothesis that microhabitat quality varies along a gradient running from top-quality dwellings through high-quality rock microhabitats (the primary habitat of the species) to lower-quality cacti (most likely secondary habitat); alternatively, a small minority of cacti may also represent high-quality habitat (white asterisk).
Triatoma brasiliensis is the most important vector of Chagas disease in the Caatinga of north-eastern Brazil [1,3,6,13]. The Caatinga is a semiarid scrubland/woodland eco-region where rainfall is scarce (~250–1500 mm/year) and highly irregular, with a dry season that can last up to 7–11 months in typical years . Severe droughts, in addition, are common, with one- to five-year periods in which rainfall can drop by 50% . Triatoma brasiliensis evolved in this harsh environment [3,4,16,17] and is well adapted to survive and reproduce in very dry/hot environments [17–20]. In particular, T. brasiliensis is common in rocky-outcrop habitats (Fig 1), where if feeds on wild rodents and other vertebrates [1,3,6,20–27]. In rock-free sedimentary lowlands, wild T. brasiliensis occupy Pilosocereus sp. shrubby cacti (Fig 1) in association with rodents. The species has never been reported from arboreal habitats in the wild [23,28–30]. Finally, T. brasiliensis often infest human dwellings (Fig 1) in rural areas where rock-outcrops, shrubby cacti, or both sustain wild populations that may act as the sources of dwelling-infesting and reinfesting bugs [1,3,6,13,21,23–27,31–35].
Triatoma brasiliensis belongs to the ‘T. brasiliensis complex’, a group of closely-related, saxicolous (i.e., rock-dwelling) species that almost certainly evolved in rocky habitats [1,3,4,16,17,21,29]. Rocky outcrops are therefore regarded as the primary habitat of T. brasiliensis [1,3] (Fig 1). Occupation of shrubby cacti, on the other hand, seems secondary–the originally rock-dwelling bugs likely followed their rodent hosts as they adapted to exploit cactus habitats in sedimentary lowlands [28,32]. Cacti, in addition, vary more in size and architectural complexity, and sustain overall smaller and less stable vertebrate populations, than rocky outcrops; this likely leads to sharper variation in the spatial-temporal availability of key resources (adequate shelter, vertebrate blood) for cactus-dwelling than for rock-dwelling bugs. Shrubby cacti, then, probably represent secondary, lower-quality habitat for T. brasiliensis [28,32] (Fig 1). Finally, the fact that T. brasiliensis is a notorious domestic pest, with frequent infestation of man-made structures and some high-density colonies [1,3,6,13,21,23–27,31–35], suggests that human dwellings, although not primary habitat, are high-quality microhabitats for the species (Fig 1). More generally, some human dwellings provide a particularly abundant and stable supply of shelter and food to synanthropic triatomines; in the absence of bug-control interventions, therefore, such dwellings are regarded as being of higher quality than wild microhabitats [1,3,6–9].
The microhabitat-quality hierarchy hypothesis makes some key predictions in terms of (i) the proportion of microhabitats where bugs are found (or infestation); (ii) the number of bugs caught (per unit catch-effort) per microhabitat (or bug density); or (iii) the number of bugs caught (per unit catch-effort) per infested microhabitat (or bug crowding) (see ref. ). In particular, this hypothesis makes the following predictions (see Fig 2):
Arrows in the plots emphasize drought-associated changes predicted by the microhabitat-quality hierarchy hypothesis. Note that we illustrate two alternatives for bug crowding in cacti–either it declines with the drought, suggesting that virtually all cacti are low-quality for the bugs (darker orange), or it increases, suggesting that a few cacti are high-quality and can act as ‘safe havens’ that sustain relatively large colonies (lighter orange, with a white asterisk emphasizing the link with Fig 1).
- Under regular environmental conditions, infestation by T. brasiliensis should be similar across microhabitats, yet density and crowding should decline moderately from dwellings to rock microhabitats and from rock microhabitats to cacti;
- Under harsher environmental conditions, infestation, density, and crowding should all remain largely stable (perhaps with a slight decline) in the highest-quality dwelling microhabitats, but should decline in wild habitats–moderately in the higher-quality rock microhabitats and more sharply in the lower-quality cactus microhabitats. If, however, enough resources remain available in a small minority of particularly suitable cacti (e.g., large, architecturally complex cacti that sustain stable wild-rodent populations), so that bug populations in those few cacti change little as environmental conditions get harsher, then density should decline more sharply than crowding in cactus microhabitats (Figs 1 and 2).
To investigate T. brasiliensis responses to harsher environmental conditions, we compared infestation, density, and crowding of bug populations sampled in human dwellings, rocky outcrops, and shrubby cacti before (or at the outset) and at the end of one of the most severe droughts on record in the Caatinga .
Study setting and timeline
Fieldwork took place in the municipalities of Russas (04°56’S, 37°58W) and Jaguaruana (04°50’S, 37°47’W), state of Ceará, Brazil. Both lie within the lower Jaguaribe river basin, whose rock-free sedimentary lowlands are bounded by crystalline-basement uplands in which rocky outcrops are widespread [22,28,32,33]. The original Caatinga scrubland/woodland vegetation is overall less disturbed in upland than in lowland areas, where only small preserved patches remain; Copernicia prunifera palms and Pilosocereus sp. shrubby cacti are common in those patches. Local rural dwellings usually comprise a substandard house with mud/clay or brick walls and timber-and-tile roofs plus a peridomestic area in which henhouses, corrals, pigsties, and piles of timber, tiles, and bricks are common [23,32,33,35] (Fig 1).
We sampled T. brasiliensis populations in December 2016 and January 2017, when the 2012–2016 severe drought hit its highest point (Fig 3). This sampling included 34 rural dwellings, 174 rock microhabitats, and 217 shrubby cacti (Table 2). During fieldwork, we noticed that some T. brasiliensis populations, and in particular those in wild microhabitats, were rarer and less dense than we would have expected based on our 10-year experience of field research in the region [22,23,32–35]. This led us to hypothesize a strong negative impact of the drought on T. brasiliensis populations occupying the lower-quality wild microhabitats–and an overall milder impact on populations occupying the higher-quality man-made microhabitats (Figs 1 and 2). Testing this hypothesis required comparing T. brasiliensis populations sampled in different microhabitats and under regular and severe climatic conditions (Fig 2). To do this, we used ‘pre-drought’ field data collected by our group in the same study area, yet either before the 2012–2016 drought  or right after its onset [28,32,34] (Fig 3). These pre-drought data cover six field trips conducted in 2005–2007 to sample rock microhabitats  plus dwellings and shrubby cacti studied in 2012 [28,32,34], when the impact of the drought was not yet perceptible (Fig 3). Overall, we compiled data from 606 T. brasiliensis populations– 66 sampled in human dwellings, 279 in rock microhabitats, and 261 in Pilosocereus sp. shrubby cacti (Table 2). We note, however, that some of these populations were sampled at different time-points in the same microhabitats (e.g., ); in addition, microhabitats sampled at any given occasion tended to occur in spatial clusters. Even if repeated sampling of individual microhabitats was separated by at least three months and the microhabitats we sampled were located at least ~100 m from each other, these data cannot be treated as independent. We account for these sources of non-independence via random effects in generalized linear mixed models (see below). Triatomine-bug sampling was done with two methods–direct manual searches and live-baited sticky traps [22,23,32–36]. We controlled for unequal bug-catch effort among individual microhabitats using an offset variable in our models (see below); to derive this variable, we considered one person-hour of manual searching (defined as one hour of bug searching/catching by one person) as roughly equivalent to one live-baited trap-night (defined as one trap operated over one night). We also note that we defined our ‘rock-microhabitat sampling unit’ as a discrete site that could potentially provide shelter to a T. brasiliensis colony; each ‘unit’ was usually a crevice or a space beneath a rock or a cluster of rocks (such as that shown in Fig 1). See S1 Dataset for the raw data.
Total annual and mean annual rainfall for the pre-drought (2000–2011, blue), drought (2012–2016, orange), and post-drought (2017–2018, grey) periods. Blue–orange gradation indicates that drought effects were barely perceptible in 2012, and orange–grey gradation that drought effects were still at its peak in early 2017. Lighter-blue bars indicate pre-drought years with rainfall below the overall mean (713 mm; empty bar in the right-hand panel). Pre-drought field trips are indicated by green arrowheads, and drought field trips by red arrowheads, along the x-axis. The graphs summarize daily rainfall data from the Russas rainfall station (Fundação Cearense de Meteorologia e Recursos Hídricos; data available at http://www.funceme.br/).
Data and data analyses
Our data are of two main types. First, we distinguished microhabitats in which at least one T. brasiliensis was found (coded 1) from those in which none was (coded 0); this binomial variable allowed us to test the predictions about infestation probabilities made by the microhabitat-quality hierarchy hypothesis (Fig 2). Second, we recorded the number of bugs caught in each microhabitat, and used this count variable to test predictions about changes in bug density and crowding (Fig 2). We built dummy indicator variables for microhabitat type (dwellings, rock microhabitats, or cacti) and for severe drought (coded 1 for microhabitats sampled in 2016–2017 and 0 otherwise).
We used R 3.5.1  for all our analyses. Descriptive and exploratory analyses included data tabulation/graphing and the calculation of the proportions of infested microhabitats, the mean number of bugs caught per unit effort in all microhabitats (i.e., density), and the mean number of bugs caught per unit effort in microhabitats where at least one bug had been found (i.e., crowding). We preliminarily compared these metrics across habitat types and between periods by calculating and plotting score confidence intervals (CIs) around proportions  and t-test-based CIs around means . We did these analyses using the stats 3.5.1 and Hmisc 4.1–1 R packages [37,40].
We aimed at formally testing the predictions of the microhabitat-quality hierarchy hypothesis while taking into account unequal sampling effort and the non-independence of spatially-clustered observations. To do this, we fitted a series of generalized linear mixed models (GLMMs) and evaluated them using second-order Akaike’s information criterion scores (AICc) [41,42]. We modeled infestation with binomial GLMMs; AICc selected the complementary log-log (cloglog) link-function instead of the canonical logit link. Preliminary analyses also showed that the negative binomial distribution fits our bug-count data better than the Poisson; because crowding data contain no zeros, we modeled this variable using a zero-truncated negative binomial distribution . All our models included (i) the log-transformed bug-catch effort as an offset, which made bug detection/catch data comparable across microhabitats, and (ii) a 10-cluster random effect (coded ‘C_1’ to ‘C_10’ in S1 Dataset) to account for the likely non-independence of temporally-replicated or spatially-clustered observations. The models fall in the following categories (see Table 3):
- Null models, representing the hypothesis that neither microhabitat type nor drought have any effects on infestation, density, or crowding;
- Microhabitat models, representing the hypothesis that infestation, density, or crowding may vary between microhabitats, but without measurable drought effects;
- Drought models, representing the hypothesis that the drought, but not microhabitat type, has measurable effects on infestation, density, or crowding;
- Joint microhabitat/drought models, including (a) additive models representing the hypothesis that drought effects are consistent across microhabitats and microhabitat effects are similar in pre-drought and drought periods, and (b) interaction models representing the hypothesis that drought effects differ among microhabitats and habitat effects may vary with the drought.
For example, a rather sophisticated view of the microhabitat-quality hierarchy hypothesis is represented by interaction microhabitat × drought models stating, e.g., (i) that the drought should not have measurable effects on infestation, density, or crowding in human dwellings; (ii) that infestation, density, and crowding should decline moderately in rock microhabitats with the drought; and (iii) that infestation and density should decline moderately in cacti–where crowding, however, may even increase with the drought (Fig 2, Table 3). This can be represented by the ‘full’ model in which Y represents the response variable (infestation, density or crowding), × represents an interaction between microhabitat type and drought, the offset adjusts for unequal bug-catch effort, the random-intercepts term accounts for cluster-level non-independence of observations, and ε is residual error. An alternative model could represent the competing specific hypothesis that drought effects should only be apparent in the lower-quality cactus microhabitats, with the drought having negligible effects on populations occupying high-quality dwelling or rock microhabitats (Table 3):
Our models, including model type, general model structure, and a brief outline of the specific hypothesis each represents, are summarized in Table 3. We fitted and evaluated infestation and density models using lme4 1.1–19 , and crowding models using glmmADMB 0.8.3.3 . For a given dataset (infestation, density, or crowding), models with lower AICc scores have more support from the data; as a rough guidance, models with AICc scores ~4–6 units larger than the lowest-score model have little to no support from the data [41,42]. We also used AICc model weights (wi) as a measure of the evidence in favor of each model . We finally quantitatively evaluated, for each outcome variable, microhabitat and drought effects (and their interactions) as estimated (i) by the most competitive models (AICc < 2.0 units from the top-ranking model’s AICc) in each model set and (ii) across all models in each set by computing model-averaged coefficients and unconditional CIs  in the package MuMIn 1.42.1 . We note that, for consistency with our AIC-based approach (see ref. ), we present all point estimates with 85% CIs.
This study is part of a research program on Chagas disease ecoepidemiology reviewed and approved by the Fiocruz Research Ethics Committee (CEP/Fiocruz 139/01), by the Fiocruz Committee on Animal Research (CEUA/Fiocruz L-026/2017), and by the Brazilian Environmental Agency (IBAMA/Sisbio 14323–6). All household heads provided informed consent prior to dwelling inspections.
Descriptive and exploratory analyses
Table 2 presents a summary of our observations and exploratory analyses; the main results are graphically shown in Fig 4. Pre-drought infestation was similarly high (~50–60%) across microhabitat types. Severe drought was associated with a slight decrease of dwelling infestation and with no measurable change in rock infestation; in shrubby cacti, in contrast, infestation fell by one order of magnitude–from nearly 50% to about 4% (Table 2, Fig 4). Pre-drought bug density and crowding decreased from dwellings to rocks to cacti. Density then decreased across all microhabitat types during the drought; it did so moderately in dwellings and rocks, yet very sharply (again by an order of magnitude) in cacti (Table 2, Fig 4). Finally, bug crowding also decreased moderately in dwellings and rocks, paralleling changes in density, but increased, even if slightly, in cacti–from 1.15 to 1.44 bugs per unit catch-effort per infested cactus (Table 2, Fig 4).
Summary metrics by microhabitat (dwellings = triangles; rocks = squares; cacti = circles) both before (‘pre-drought’, blue) and during the drought (‘drought’, orange), with 85% confidence intervals (CI). Note that y-axes are on log10 scale.
Generalized linear mixed models–infestation
We fitted 14 binomial GLMMs to evaluate the relative support for each of our specific infestation hypothesis (see Table 3). Table 4 shows that the better-performing models were joint interaction models; simpler microhabitat-only, drought-only, or joint additive models had essentially no support from the data. The top-ranking model (wi = 0.882) had much more support than the second-ranking model (ΔAICc > 4; Table 4). We therefore found strong evidence [41,42] in favor of the specific hypothesis stating that the effects of the drought on infestation were different in higher-quality (dwellings and rocks) vs. lower-quality (cacti) microhabitats–relative to the competing hypotheses in Table 3 and after accounting for unequal sampling effort and cluster-level non-independence of observations. The numerical output of the top-ranking model is presented in Table 5, and model-averaged coefficients for all variables in Fig 5. These quantitative results suggest (i) that microhabitat effects were relatively small, with pre-drought infestation probably slightly less frequent in cactus than in dwellings or rocks, and (ii) that the drought had a strong negative effect on cactus infestation–but only a small impact, effectively indistinguishable from zero, on dwelling and rock infestation (Table 5, Fig 5).
Microhabitat effects are represented in yellow, drought effects in red, and microhabitat × drought interactions in orange; intercepts (grey) are included for reference; the dashed horizontal line at zero indicates no effect.
The only competitive model (ΔAICc < 2.0) is in bold typeface.
Generalized linear mixed models–density
Bug-density negative binomial models, ranked by AICc, are presented in Table 6. Only two joint models (one additive, one with interactions) were competitive (ΔAICc < 2.0; Σwi = 0.892); their numerical output is presented in Table 7. These analyses suggest that bug density varied across microhabitat types and was also affected by the drought, with perhaps some moderate variation of drought effects across different microhabitats; as with infestation, microhabitat-only and drought-only models had little support from the data (Table 6). Model-averaged effect estimates are shown in Fig 5. These quantitative results suggest (i) that there were strong microhabitat effects, with pre-drought bug density declining from dwellings to rocks to cacti, and (ii) that the drought had moderate negative effects on bug density–with probably a somewhat stronger effect on cactus populations than in rock or dwelling populations (Table 7, Fig 5).
Competitive models (ΔAICc < 2.0) are in bold typeface.
Generalized linear mixed models–crowding
Table 8 shows that there was more model-selection uncertainty, as measured by the slower rate of decline of wi values, in crowding than in infestation or density models (see Tables 4 and 6). The two top-ranking models (ΔAICc < 2.0; Σwi = 0.565) were the same joint models that AICc selected as the best-performing density models (Tables 6 and 8). This suggests that, as for density, bug crowding varied across microhabitat types and was also affected by the drought–with this latter effect likely differing to some extent among different microhabitats. The third-ranking crowding model (‘cm2’, ΔAICc = 2.34), however, was a habitat-only model distinguishing man-made from wild microhabitats, and the fifth-ranking model (‘cm1’, ΔAICc = 2.59) was another habitat model distinguishing all microhabitat types (Table 8), thus providing some evidence that microhabitat effects, and in particular dwelling effects, might be stronger for bug crowding than for density. The numerical output of the competitive crowding models (Table 9) and model-averaged estimates (Fig 5) substantiate these views, showing (i) that pre-drought microhabitat effects were strong, with a sharp difference between the less crowded wild microhabitats and the much more crowded dwellings, and (ii) that drought effects were overall small and probably different across microhabitat types–somewhat negative in dwellings, close to zero in rocks, and somewhat positive in cacti (Table 9, Fig 5).
Competitive models (ΔAICc < 2.0) are in bold typeface.
In this study we have examined the notion that some of the habitats that Chagas disease vectors can occupy are of higher quality than others–i.e., that there is a ‘hierarchy’ of triatomine-bug microhabitat quality (Fig 1). We confronted some of the key predictions of this hypothesis with extensive field data gathered before and during an extreme climate event–the severe drought that ravaged the Caatinga in 2012–2016. Using T. brasiliensis as a case-study, we found compelling evidence for the often-assumed, yet seldom tested, idea that substandard rural dwellings are of higher quality for the bugs than the species’ wild microhabitats. Further, our findings indicate that rock microhabitats, where the primarily saxicolous T. brasiliensis probably evolved, are of higher quality than shrubby cacti–which most likely represent secondary habitat. However, our results also suggest that, even if cacti are of low average quality, a few high-quality cacti can act as ‘safe havens’ where relatively large T. brasiliensis populations survive through periods of extremely harsh environmental conditions; the same appears to be true, in contrast, of most rock and dwelling microhabitats (Figs 1 and 2).
Triatomines probably evolved from nest-dwelling, arthropod-feeding ancestors [1,4,48], and now occupy protected microhabitats in close association with their vertebrate hosts [1,3,5,7] (Table 1). The quality of those microhabitats is thought to depend on (i) whether the physical substrate is adequate for the bugs’ hiding, resting, foraging, mating, and egg-laying requirements; (ii) whether the microclimate is adequate for bug development, survival, and reproduction; and (iii) whether the vertebrate-blood supply is adequate in terms of quality, amount, and temporal stability [1,3,7,49]. Here, ‘adequate’ is defined relative to the physical substrate, microclimate, and blood sources each bug species is adapted to–i.e., relative to the characteristics of the ‘primary’ microhabitats where the species evolved. Many triatomine species, however, can also breed in ‘secondary’ microhabitats–including, in some cases, man-made habitats. Theory predicts that, all else being equal, secondary habitat should be of lower overall quality than the primary habitat each species is adapted to . If, however, adequate substrate, microclimate, and blood supply are available in some secondary microhabitats, then they might become high-quality–possibly even higher-quality than the species’ primary habitat if, for example, food availability is enhanced or if there is ‘ecological release’ from predators or competitors [51,52]. In the absence of bug-control interventions, this appears to be the case of some human dwellings [1,3]. Further, the highly aggregated distribution of wild triatomine populations, with many small and very few large colonies , suggests that natural microhabitats, either primary or secondary, may also vary in their quality. While these observations clearly hint at the existence of a ‘hierarchy’ of triatomine-bug microhabitat quality, from top-quality dwellings through good-quality primary to poor-quality secondary wild habitats (Fig 1), a formal test of this hypothesis was apparently unavailable before this report.
Our results provide empirical support for the view that human dwellings can be higher-quality habitat for T. brasiliensis than the species’ natural habitats, whether primary (rock microhabitats) or secondary (shrubby cacti) (Figs 1 and 2). First, pre-drought infestation with T. brasiliensis was about as frequent in dwellings as it was in rocks or cacti, yet bug density and crowding were clearly higher in dwellings (Table 2, Figs 4 and 5). This suggests that, in regular-climate periods (and in the absence of systematic vector control), the bugs use dwellings about as often as they use their natural habitats, yet synanthropic colonies grow larger, on average, than wild ones. Second, dwelling infestation varied little, perhaps with a slight decline, as environmental conditions got extremely harsh; this was similar to what was seen in the bugs’ primary rocky habitat, but not in secondary cactus habitats, where infestation dropped by about 10-fold (Table 2, Figs 4 and 5). Dwelling bug density and crowding declined moderately with the severe drought, yet both remained above values recorded in wild microhabitats even under regular climate; drought effects on crowding, however, appeared to be particularly negative in dwellings (Fig 5). Thus, our results suggest that synanthropic T. brasiliensis populations (i) were as common as, yet larger than, their wild counterparts under regular climate, and (ii) also fared overall better than wild populations during the 2012–2016 severe drought. These findings are in line with observations that T. dimidiata dwelling-breeding colonies remained largely stable after an extreme-climate event, hurricane Isidore, struck the Yucatán peninsula in 2002, yet house-invasion by adult bugs transiently increased–thus suggesting that the bugs were flying from heavily disturbed wild habitats (which, however, were not investigated in that study) towards the more stable or buffered man-made habitats .
We found that rock microhabitats were often infested by T. brasiliensis before the drought, with moderate bug density and crowding, and that none of those metrics had changed much at the end of the five-year severe drought (Table 2, Figs 4 and 5). Thus, we found no evidence that rock-microhabitat pre-drought infestation had declined, and model-based estimates suggest moderate reductions of density and crowding that were most likely similar across microhabitat types (Tables 4–9, Figs 4 and 5). These results clearly suggest that extreme drought had, if anything, a modest impact on wild T. brasiliensis populations occupying their primary rocky habitat–the frequency of infestation did not change and the size of the colonies declined only slightly-to-moderately.
Pilosocereus sp. shrubby cacti sampled right after the onset of the severe drought (Fig 3) were often occupied by T. brasiliensis, yet both density and crowding were already apparently lower than in rocks and dwellings (Table 2, Figs 4 and 5). We found that, when sampled during the last months of the drought, cacti in the same area were much less likely to be infested and supported a much less dense bug population (Tables 2–7, Figs 4 and 5). Crowding, however, either did not change or perhaps even increased slightly in the very few cacti that were infested during the drought (Tables 2, 8 and 9, Figs 4 and 5, S1 Fig). This suggests that most cacti were so poor-quality, as overall expected from secondary habitat, that their T. brasiliensis populations either went locally extinct with the drought or became so small as to be undetectable. On the other hand, our analyses suggest that relatively large populations managed to survive through the drought in a few cacti–which may, therefore, represent fairly high-quality habitat (Figs 1 and 2). Our data show that bug-catch per unit effort was > 2.0 in three cacti sampled during the drought and in two sampled before the drought, yet infestation was > 10 times higher before than during the drought (Table 2, Fig 4 and S1 Fig).
In sum, our data and analyses suggest, taken together, that most of the dwellings we investigated were top-quality habitat for T. brasiliensis, with bug populations remaining largely stable through an extreme-drought event. In the wild, most rocks were probably high-quality habitat, with pre-drought density and crowding lower than in dwellings but higher than in cacti–and with the drought having, at most, modest effects on those populations. Most cacti, in contrast, were probably low-quality habitat for wild T. brasiliensis–low enough as to be apparently unable to sustain bug populations during severe drought. Some bugs, however, survived under such extreme conditions in a few cacti that possibly had the defining features of high-quality microhabitats–adequate physical substrate, adequate microclimate, and adequate blood supply. These hypothesis-driven findings are based on a unique, large dataset and on a thorough analytical strategy that accounted for the particular error distribution of each dependent variable, adjusted for unequal sampling effort and non-independence of spatially-clustered observations, and incorporated model-selection uncertainty via model averaging. There are, however, several important caveats that should be kept in mind when interpreting our results. First, in this report we do not address why some specific microhabitats were of higher quality than others. Instead, we assume (i) that physical substrate, as suggested by, e.g., refs. [21–23,25,31,32,35], was minimally adequate for T. brasiliensis in most of the microhabitats we sampled, and did not change much with the drought; (ii) that the already dry pre-drought microclimate was overall adequate, as suggested by high infestation frequency and by previous studies on T. brasiliensis [18–20] and other triatomines , and changed more during the drought in poorer-quality microhabitats such as small cacti or less-protected rocky habitats; and (iii) that the blood supply, as suggested by our extensive field observations, was larger in dwellings than in wild microhabitats–and changed less with the drought (a) in dwellings (where people plus domestic and synanthropic vertebrates remained available) than in rock microhabitats (where wild rodents and other small mammals, birds, reptiles, or free-ranging goats were common even during the drought) and (b) in rock microhabitats than in cacti (where wild rodent populations visibly declined during the drought). A detailed assessment of physical-substrate and host-availability effects on wild T. brasiliensis populations is ongoing and will be the topic of a future report. Second, and along the same lines, we concentrated our sampling on microhabitats that, based on prior experience, we believed could sustain at least small T. brasiliensis populations. For example, we did not sample small or arboreal (vs. creeping) cacti, small isolated boulder outcroppings, dwellings that were of a higher-than-average standard  or had been sprayed with insecticides less than about one year before fieldwork, or trees or palms [23,28–30]. Our conclusions, therefore, refer to microhabitats that were considered to be, a priori, minimally suitable for T. brasiliensis, and we make no claims as to how the drought might have affected populations occupying a priori less suitable microhabitats–although we would expect particularly strong, negative drought effects. Finally, we note that our analyses assume that any particular microhabitat in which we found no T. brasiliensis was truly unoccupied; this assumption, however, is probably unrealistic in man-made [32,34] and cactus microhabitats , and remains untested in rock microhabitats–a topic we also plan to address in future work. Microhabitats in which we caught no bugs, therefore, may have been either uninfested or occupied by very small populations that went undetected. Because a local population so small as to go undetected is likely to be on the brink of extinction [11,12,50–52], frequent non-detections may also signal a negative effect of microhabitat type, drought, or their interaction. We note, in addition, that adjusting for bug-catch effort and cluster-level dependencies made our results more directly comparable across microhabitat types and climate periods. In sum, we are confident that our study, with its strengths and limitations, provides a reliable (and original) appraisal of the key predictions of the microhabitat-quality hierarchy hypothesis.
Conclusions and outlook
Triatoma brasiliensis populations fared overall better in dwellings and rocky outcrops than in cacti during a five-year extreme drought. Estimates of microhabitat and drought effects on infestation, density, and crowding suggest that only a few cacti, vs. many rocks and dwellings, represent good-quality habitat for the species under such extremely harsh conditions. Thus, and as predicted by the microhabitat-quality hierarchy hypothesis, T. brasiliensis microhabitats appear to lie along a ‘quality gradient’ that runs from top-quality rural dwellings through high-quality, primary rocky habitats and to the lower-quality, secondary habitat represented by shrubby cacti–a few of which, however, are probably high-quality. What specific traits determine, and to what extent, microhabitat quality for T. brasiliensis remains unclear. Limited data suggest that, if minimal physical-substrate and microclimate requirements are met, vertebrate-host availability may be key to microhabitat quality in both dwellings  and cacti ; this might, then, also be the case for rock microhabitats. Ascertaining the relative roles of physical substrate, microclimate, and host availability on triatomine-bug microhabitat quality may help us understand whether and how environmental management could synergize insecticide-based control of Chagas disease vectors (see, e.g., ref.  and references therein). At any rate, the results of our hypothesis-driven assessment clearly imply that T. brasiliensis can endure extreme climate by exploiting high-quality microhabitats, whether wild or man-made, in the semiarid Caatinga. This provides a hint of the species’ potential to adapt to the widespread increase in aridity expected under climate change across northern South America . We anticipate that, as long as substandard dwellings remain available to the bugs in the semiarid Brazilian north-east , T. brasiliensis will keep using them as high-quality microhabitats in which to survive and breed–and will likely do so even under extreme climate.
S1 Fig. Bug-catch per unit effort in each microhabitat and period.
Black bars highlight microhabitat where no bugs were caught. The insets in the right-hand-side panels show cacti with non-zero bug catches, which are barely visible, given the scale of the y-axes, in the main graphs.
We thank the people from the study communities for their kindness and patience; the Municipal Health Departments of Russas and Jaguaruana for technical assistance and physical facilities; and the Ceará State Health Department for providing transportation and field support. We particularly thank Carlos Jorge de Lima, Francineudo Nogueira de Oliveira, and Marcos Eduardo Melo, vector control-surveillance agents, for their invaluable assistance during fieldwork.
- 1. Lent H, Wygodzinsky P. Revision of the Triatominae (Hemiptera: Reduviidae) and their significance as vectors of Chagas’ disease. Bull Am Mus Nat Hist. 1979; 163: 123–520.
- 2. GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018; 392: 1789–1858. pmid:30496104
- 3. Barrett TV. Advances in triatomine bug ecology in relation to Chagas disease. In: Harris KH, organizer. Advances in disease vector research. New York: Springer-Verlag; 1991. pp. 143–176.
- 4. Monteiro FA, Weirauch C, Felix M, Lazoski F, Abad-Franch F. Evolution, systematics, and biogeography of the Triatominae, vectors of Chagas disease. In: Rollinson D, Stothard JR, editors. Advances in parasitology, vol 99. Cambridge, MA: Academic Press; 2018. pp. 265–344. https://doi.org/10.1016/bs.apar.2017.12.002 pmid:29530308
- 5. Gaunt M, Miles M. The ecotopes and evolution of triatomine bugs (Triatominae) and their associated trypanosomes. Mem Inst Oswaldo Cruz. 2000; 95: 557–565. pmid:10904415
- 6. World Health Organization. Control of Chagas disease. Second report of the WHO Expert Committee. World Health Organ Tech Rep Ser. 2002; 905: 1–109.
- 7. Noireau F, Dujardin JP. Biology of Triatominae. In: Telleria J, Tibayrenc M, editors. American trypanosomiasis: Chagas disease one hundred years of research. Burlington: Elsevier; 2010. pp. 149–168.
- 8. Schofield CJ. Density regulation of domestic populations of Triatoma infestans in Brazil. Trans R Soc Trop Med Hyg. 1980; 74: 761–769. pmid:7010697
- 9. Gürtler RE, Cohen JE, Cecere MC, Chuit R. Shifting host choices of the vector of Chagas disease Triatoma infestans in relation to the availability of hosts in houses in north-west Argentina. J Appl Ecol. 1997; 34: 699–715.
- 10. Mortelliti A, Amori G, Boitani L. The role of habitat quality in fragmented landscapes: a conceptual overview and prospectus for future research. Oecologia. 2010; 163: 535–547. pmid:20414787
- 11. Ehrlich PR, Murphy DD, Singer MC, Sherwood CB, White RR, Brown IL. Extinction, reduction, stability and increase: the responses of checkerspot butterfly (Euphydryas editha) populations to the California drought. Oecologia. 1980; 46: 101–105. pmid:28310633
- 12. Franklin AB, Anderson DR, Gutierrez RJ, Burnham KP. Climate, habitat quality, and fitness in northern spotted owl populations in northwestern California. Ecol Monogr. 2000; 70: 539–590.
- 13. Costa J, Almeida CE, Dotson EM, Lins A, Vinhaes M, Silveira AC, et al. The epidemiologic importance of Triatoma brasiliensis as a Chagas disease vector in Brazil: a revision of domiciliary captures during 1993–1999. Mem Inst Oswaldo Cruz. 2003; 98: 443–449. pmid:12937751
- 14. Leal IR, da Silva JMC, Tabarelli M, Lacher TE Jr. Changing the course of biodiversity conservation in the Caatinga of northeastern Brazil. Conserv Biol. 2005; 19: 701–706.
- 15. De Nys E, Engle N, Magalhaes AR, editors. Drought in Brazil: proactive management and policy. Boca Raton: Taylor & Francis; 2017.
- 16. Monteiro FA, Donnelly MJ, Beard CB, Costa J. Nested clade and phylogeographic analyses of the Chagas disease vector Triatoma brasiliensis in Northeast Brazil. Mol Phylogenet Evol. 2004; 32: 46–56. pmid:15186796
- 17. Costa J, Dornak LL, Almeida CE, Peterson AT. Distributional potential of the Triatoma brasiliensis species complex at present and under scenarios of future climate conditions. Parasit Vectors. 2014; 7: 238. pmid:24886587
- 18. Guarneri AA, Lazzari C, Diotaiuti L, Lorenzo MG. The effect of relative humidity on the behaviour and development of Triatoma brasiliensis. Physiol Entomol. 2002; 27: 142–147.
- 19. Lorenzo MG, Guarneri AA, Pires HH, Diotaiuti L, Lazzari CR. Aspectos microclimáticos del hábitat de Triatoma brasiliensis. Cad Saude Publica. 2000; 16(Suppl. 2): 69–74.
- 20. Catalá S, Bezerra CM, Diotaiuti L. Thermal preferences and limits of Triatoma brasiliensis in its natural environment–field observations while host searching. Mem Inst Oswaldo Cruz. 2015; 110: 793–796. pmid:26517659
- 21. Costa J, de Almeida JR, Britto C, Duarte R, Marchon-Silva V, Pacheco RS. Ecotopes, natural infection and trophic resources of Triatoma brasiliensis (Hemiptera, Reduviidae, Triatominae). Mem Inst Oswaldo Cruz. 1998; 93: 7–13.
- 22. Sarquis O, Carvalho-Costa FA, Oliveira LS, Duarte R, D’Andrea PS, et al. Ecology of Triatoma brasiliensis in northeastern Brazil: seasonal distribution, feeding resources, and Trypanosoma cruzi infection in a sylvatic population. J Vector Ecol. 2010; 35: 385–394. pmid:21175946
- 23. Sarquis O, Carvalho-Costa FA, Toma HK, Georg I, Burgoa MR, Lima MM. Eco-epidemiology of Chagas disease in northeastern Brazil: Triatoma brasiliensis, T. pseudomaculata and Rhodnius nasutus in the sylvatic, peridomestic and domestic environments. Parasitol Res. 2012; 110: 1481–1485. pmid:21979785
- 24. Bezerra CM, Cavalcanti LP, Souza RC, Barbosa SE, Xavier SC, Jansen AM, et al. Domestic, peridomestic and wild hosts in the transmission of Trypanosoma cruzi in the Caatinga area colonised by Triatoma brasiliensis. Mem Inst Oswaldo Cruz. 2014; 109: 887–898. pmid:25410992
- 25. Almeida CE, Faucher L, Lavina M, Costa J, Harry M. Molecular individual-based approach on Triatoma brasiliensis: inferences on triatomine foci, Trypanosoma cruzi natural infection prevalence, parasite diversity and feeding sources. PLoS Negl Trop Dis. 2016; 10: e0004447. pmid:26891047
- 26. Lilioso M, Folly-Ramos E, Rocha FL, Rabinovich J, Capdevielle-Dulac C, Harry M, et al. High Triatoma brasiliensis densities and Trypanosoma cruzi prevalence in domestic and peridomestic habitats in the state of Rio Grande do Norte, Brazil: the source for Chagas disease outbreaks? Am J Trop Med Hyg. 2017; 96: 1456–1459. pmid:28719275
- 27. Bezerra CM, Barbosa SE, Souza RCM, Barezani CP, Gürtler RE, Ramos AN Jr, et al. Triatoma brasiliensis Neiva, 1911: food sources and diversity of Trypanosoma cruzi in wild and artificial environments of the semiarid region of Ceará, northeastern Brazil. Parasit Vectors. 2018; 11: 642. pmid:30558643
- 28. Valença-Barbosa C, Lima MM, Sarquis O, Bezerra CM, Abad-Franch F. A common Caatinga cactus, Pilosocereus gounellei, is an important ecotope of wild Triatoma brasiliensis populations in the Jaguaribe valley of Northeastern Brazil. Am J Trop Med Hyg. 2014; 90: 1059–1062. pmid:24710611
- 29. Carbajal-de-la-Fuente AL, Dias-Lima A, Lopes CM, Emperaire L, Walter A, Ferreira A, et al. Behavioural plasticity of Triatominae related to habitat selection in North-East Brazil. J Med Entomol. 2008; 45: 14–19. pmid:18283937
- 30. Abad-Franch F, Lima MM, Sarquis O, Gurgel-Gonçalves R, Sánchez-Martín M, Calzada J, et al. On palms, bugs, and Chagas disease in the Americas. Acta Trop. 2015; 151: 126–141. pmid:26196330
- 31. Diotaiuti L, Faria Filho OF, Carneiro FC, Dias JC, Pires HH, Schofield CJ. Aspectos operacionais do controle do Triatoma brasiliensis. Cad Saude Publica. 2000; 16(Suppl. 2): 61–67.
- 32. Valença-Barbosa C, Lima MM, Sarquis O, Bezerra CM, Abad-Franch F. Modeling disease vector occurrence when detection is imperfect II: drivers of site-occupancy by synanthropic Triatoma brasiliensis in the Brazilian northeast. PLoS Negl Trop Dis. 2014; 8: e2861. pmid:24811125
- 33. Coutinho CF, Souza-Santos R, Lima MM. Combining geospatial analysis and exploratory study of triatomine ecology to evaluate the risk of Chagas disease in a rural locality. Acta Trop. 2012; 121: 30–33. pmid:22015423
- 34. Abad-Franch F, Valença-Barbosa C, Sarquis O, Lima MM. All that glisters is not gold: sampling-process uncertainty in disease-vector surveys with false-negative and false-positive detections. PLoS Negl Trop Dis. 2014; 8: e3187. pmid:25233352
- 35. Sarquis O, Sposina R, de Oliveira TG, Mac Cord JR, Cabello PH, Borges-Pereira J, Lima MM. Aspects of peridomiciliary ecotopes in rural areas of northeastern Brazil associated to triatomine (Hemiptera, Reduviidae) infestation, vectors of Chagas disease. Mem Inst Oswaldo Cruz. 2006; 101: 143–147. pmid:16830706
- 36. Noireau F, Abad-Franch F, Valente SA, Dias-Lima A, Lopes CM, Cunha V, et al. Trapping Triatominae in sylvatic habitats. Mem Inst Oswaldo Cruz. 2002; 97: 61–63.
- 37. R Core Team. 2014. The R project for statistical computing. www.Rproject.org
- 38. Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med. 1998; 17: 857–872. pmid:9595616
- 39. Nakagawa S, Cuthill IC. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biol Rev Camb Philos Soc. 2007; 82: 591–605. pmid:17944619
- 40. Harrell FE. Hmisc: a package of miscellaneous R functions. 2014. Programs available from http://biostat.mc.vanderbilt.edu/Hmisc.
- 41. Burnham KP Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach, 2nd edition. Berlin: Springer-Verlag; 2002.
- 42. Aho K, Derryberry D, Peterson T. Model selection for ecologists: the worldviews of AIC and BIC. Ecology. 2014; 95: 631–636. pmid:24804445
- 43. Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM. Zero-truncated and zero-inflated models for count data. In: Mixed effects models and extensions in ecology with R. Statistics for Biology and Health. New York: Springer; 2009. pp. 261–293.
- 44. Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw. 2015; 67: 1–48.
- 45. Bolker B, Skaug H, Magnusson A, Nielsen A. Getting started with the glmmADMB package. 2012. glmmadmb.r-forge.r-project.org/glmmADMB.pdf
- 46. Bartoń K. MuMIn: Multi‐model inference. https://CRAN.R-project.org/package=MuMIn.
- 47. Arnold TW. Uninformative parameters and model selection using Akaike’s information criterion. J Wildl Manage. 2010; 74: 1175–1178.
- 48. Hwang WS, Weirauch C. Evolutionary history of assassin bugs (Insecta: Hemiptera: Reduviidae): insights from divergence dating and ancestral state reconstruction. PLoS One. 2012; 7: e45523. pmid:23029072
- 49. Abad-Franch F, Santos WS, Schofield CJ. Research needs for Chagas disease prevention. Acta Trop. 2010; 115: 44–54. pmid:20227378
- 50. Gomulkiewicz R, Holt RD, Barfield M, Nuismer SL. Genetics, adaptation, and invasion in harsh environments. Evol Appl. 2010; 3: 97–108. pmid:25567911
- 51. Estoup A, Ravigne V, Hufbauer RA, Vitalis R, Gautier M, Facon B. Is there a genetic paradox of biological invasion? Annu Rev Ecol Evol Syst. 2016; 47: 51–72.
- 52. Colautti RI, Alexander JM, Dlugosch KM, Keller SR, Sultan SE. Invasions and extinctions through the looking glass of evolutionary ecology. Philos Trans R Soc Lond B Biol Sci. 2017; 372: pii: 20160031.
- 53. Guzman-Tapia Y, Ramirez-Sierra MJ, Escobedo-Ortegon J, Dumonteil E. Effect of Hurricane Isidore on Triatoma dimidiata distribution and Chagas disease transmission risk in the Yucatán Peninsula of Mexico. Am J Trop Med Hyg. 2005; 73: 1019–1025. pmid:16354805
- 54. de la Vega G, Schilman PE. Using ecophysiological traits to understand the realized niche: the role of desiccation tolerance in Chagas disease vectors. Oecologia. 2017; 185: 607–618. pmid:29067559
- 55. Lima MM, Carvalho-Costa FA, Toma HK, Borges-Pereira J, de Oliveira TG, Sarquis O. Chagas disease and housing improvement in northeastern Brazil: a cross-sectional survey. Parasitol Res. 2015; 114: 1687–1692. pmid:25673077
- 56. Gürtler RE, Yadon ZE. Eco-bio-social research on community-based approaches for Chagas disease vector control in Latin America. Trans R Soc Trop Med Hyg. 2015; 109: 91–98. pmid:25604759
- 57. Zaninelli PG, Menendez CG, Falco M, Lopez-Franca N, Carril AF. Future hydroclimatological changes in South America based on an ensemble of regional climate models. Clim Dyn. 2019; 52: 819–830.