INTRODUCTION
Upon infection of its insect vector, the parasite Trypanosoma cruzi, aetiological agent of Chagas disease, joins a diverse microbial community (Eichler and Schaub, Reference Eichler and Schaub2002; Espino et al. Reference Espino, Gómez, González, do Santos, Solano, Sousa, Moreno, Windsor, Ying, Vilchez and Osuna2009; Vallejo et al. Reference Vallejo, Guhl and Schaub2009), consisting of up to eight species of bacteria (Vallejo et al. Reference Vallejo, Guhl and Schaub2009), six genera of fungi (De Moraes et al. Reference De Moraes, Reis-de-Figueiredo, Vieira-Junqueira, Lara-da-Costa, Aguiar and Cunha-de-Oliveira2001, Reference De Moraes, Junqueira, Celano, Da Costa and Coura2004; Luz et al. Reference Luz, Rocha and Nery2004), four other trypanosomatid species (Schaub, Reference Schaub1992) and at least one virus (Marti et al. Reference Marti, Balsalobre, Susevich, Rabinovich and Echeverría2015). These taxa can interact indirectly via resource competition, immune modulation, competition for immune-free space (Dobson, Reference Dobson1985; Cox, Reference Cox2001; Pedersen and Fenton, Reference Pedersen and Fenton2007) and even sometimes directly through physical attack (Azambuja et al. Reference Azambuja, Feder and Garcia2004; Castro et al. Reference Castro, Moraes, Garcia and Azambuja2007), all of which have potential consequences for the survival and reproduction of the insect.
One species of particular interest is Trypanosoma rangeli, a T. cruzi congeneric that infects several of the same mammal and triatomine species as T. cruzi. Trypanosoma rangeli is of interest in the study of Chagas disease because it shares at least 60% of its antigens with T. cruzi (Guhl and Marinkelle, Reference Guhl and Marinkelle1982; Saldaña and Sousa, Reference Saldaña and Sousa1996; Guhl and Vallejo, Reference Guhl and Vallejo2003). These antigenic similarities can lead to cross-reactions in immunogenic diagnostic tests, which can result in erroneous Chagas disease diagnoses (Guhl et al. Reference Guhl, Hudson, Marinkelle, Jaramillo and Bridge1987) and in turn interfere with the ability to predict and describe Chagas disease distribution in Chagas-endemic regions. Trypanosoma cruzi and T. rangeli are often found co-infecting together in field-caught triatomine bugs of the genus Rhodnius (Fig. 1), some of which are considered key vectors of T. cruzi to humans (Gorla and Noireau, Reference Gorla, Noireau, Telleria and Tibayrenc2010).
Fig. 1. Reported co-infection prevalence in field-caught Rhodnius triatomines (Carcavallo et al. Reference Carcavallo, Martinez Silva, Otero and Tonn1975; Vallejo et al. Reference Vallejo, Marinkelle, Guhl and de Sanchez1988; Pavia et al. Reference Pavia, Vallejo, Montilla, Nicholls and Puerta2007; Pineda et al. Reference Pineda, Montalvo, Alvarez, Santamaría, Calzada and Saldaña2008; Grijalva et al. Reference Grijalva, Suarez-Davalos, Villacis, Ocana-Mayorga and Dangles2012).
Although not pathogenic in mammals (Herbig-Sandreuter, Reference Herbig-Sandreuter1957), in triatomine bugs, T. rangeli has been observed to negatively affect the survival and development of the triatomine species Rhodnius prolixus when experimentally infected with the parasite (Grewal, Reference Grewal1957; Tobie, Reference Tobie1965; Gómez, Reference Gómez1967; Watkins, Reference Watkins1971; Añez, Reference Añez1984; Añez et al. Reference Añez, Nieves and Cazorla1987). Until recently, T. cruzi was not believed to have negative consequences for its invertebrate hosts (Schaub, Reference Schaub1989a , Reference Schaub1992, Reference Schaub1994), although this has now been shown to be variable (Elliot et al. Reference Elliot, Rodrigues, Lorenzo, Martins-Filho and Guarneri2015; Peterson et al. Reference Peterson, Graham, Dobson and Chavez2015). Little is known about the consequences of T. cruzi–T. rangeli co-infection for the triatomine bug, and to our knowledge, has been investigated just once (Añez et al. Reference Añez, Molero, Valderrama, Nieves, Cazorla and Márquez1992); that study reported delayed nymphal development and increased mortality in R. prolixus co-infected with T. cruzi and T. rangeli compared with singly-infected insects. However, the sustained effects of such co-infection on triatomines (e.g. on their reproduction or overall fitness) have never been investigated. In mammals, it was found that T. rangeli exposure in vertebrates prior to T. cruzi infection modulated the host immune response to T. cruzi, resulting in reduced disease severity in both acute and chronic T. cruzi infections (Basso et al. Reference Basso, Moretti and Votrero-cima1991, Reference Basso, Castro, Introini, Gil, Truyens and Moretti2007, Reference Basso, Moretti and Fretes2008, Reference Basso, Moretti and Fretes2014; Marini et al. Reference Marini, Moretti, Bermejo and Basso2011; Basso, Reference Basso2013). These studies suggest that T. cruzi–T. rangeli co-infection could affect triatomine fitness differently than single-species infections.
Here, we compared the fitness of triatomine bugs (R. prolixus) experimentally co-infected with T. cruzi and T. rangeli with the fitness of bugs with single-species infections of T. cruzi or T. rangeli. We defined fitness as the net contribution to future generations of each insect. We aimed to determine if there is a difference in fitness between bugs with different infection types, as we propose that the extent to which T. cruzi–T. rangeli co-infection alters the impact of each infection on individual vector fitness may in turn alter the transmission potential of the parasites. This, in turn, could have implications for vector control and Chagas disease prevention strategies.
MATERIALS AND METHODS
Experimental design
We infected 100 R. prolixus fifth instar females with just T. cruzi, just T. rangeli, or T. cruzi and T. rangeli (Table 1). A total of 33 additional uninfected insects were used as controls, for a total of 133 insects used in the experiment. After moulting into the adult stage, each female was mated with an uninfected male, and survival and reproduction were measured for up to 96–140 days. All experiments were carried out in the laboratory of the Grupo de la Biología y Control de Enfermedades Infecciosas [Biology and Control of Infectious Diseases Group (BCEI)], University of Antioquia, Medellín, Colombia.
Table 1. Treatment groups
Triatomines
All R. prolixus used in the experiment were from laboratory colonies reared in the BCEI insectary, where triatomine colonies are kept under semi-controlled climate conditions (~27 ± 1 °C and 65 ± 15% RH) and a 12 h photoperiod, and given the opportunity to feed twice weekly on hens according to the animal ethics committee regulations of the Sede de Investigación Universitaria [University Investigation Headquarters (SIU)] of the University of Antioquia. Insects used in the experiment were fed on hens once per oviposition cycle, described below. Colonies were founded by R. prolixus eggs collected in Colombia between 2000 and 2009. All insects used in the experiment were 5th instar nymphs at the time of infection. Nymphs were collected manually from the colonies, and sex was subsequently determined (prior to infection) by examining the two concentric terminal segments around the anus on the insect's ventral side under a dissecting microscope, as described in Chiang et al. (Reference Chiang, Chiang, Hoogendoorn and Lima2013) and Gillet (Reference Gillet1935).
Parasites
We used the parasite strains ‘Gal61’ (T. cruzi) and ‘Choachí’ (T. rangeli). Gal61 was originally isolated from a mouse in Galeras, Colombia, and belongs to the T. cruzi discrete typing unit (DTU) group I (Rojas et al. Reference Rojas, Caro, Lopera, Triana, Dib and Bedoya2007; Falla et al. Reference Falla, Herrera, Fajardo, Montilla, Vallejo and Guhl2009). Choachí was originally isolated from an R. prolixus individual collected in Cundinamarca, Colombia (Grisard et al. Reference Grisard, Campbell and Romanha1999; Vargas et al. Reference Vargas, Souto, Carranza, Vallejo and Zingales2000; Urrea et al. Reference Urrea, Guhl, Herrera, Falla, Carranza, Cuba-Cuba, Triana-Chávez, Grisard and Vallejo2011), and belongs to the KP1(+) kDNA (kinetoplastid deoxyribonucelic acid) group (Vallejo et al. Reference Vallejo, Guhl, Carranza, Lozano, Sánchez, Jamarillo, Gualtero, Castañeda, Silva and Steindel2002), which is associated with the Prolixus complex of Rhodnius (Urrea et al. Reference Urrea, Carranza, Cuba Cuba, Gurgel-Gonçalves, Guhl, Schofield, Triana and Vallejo2005).
Trypanosoma cruzi parasites were cultured and maintained as described in Peterson et al. (Reference Peterson, Graham, Dobson and Chavez2015). Briefly, epimastigotes were cultured at 28 °C in a RPMI-1640 liquid medium (Sigma-Aldrich, St. Louis, MO) supplemented with 10% fetal bovine serum (FBS). Epimastigotes of the T. rangeli Choachí strain were supplied by Professor Gustavo Vallejo of the University of Tolima, where they were cultured at 28 °C in NNN medium and supplemented with 10% FBS. Infectivity was maintained by cyclic R. prolixus–mouse passages every 3 months.
Insect infection
We prepared the parasites (epimastigote stage) and infected the insects as described in Peterson et al. (Reference Peterson, Graham, Dobson and Chavez2015). Briefly, parasites were counted in a Neubauer chamber, washed through centrifugation and resuspended in 1 mL of sterile phosphate-buffered saline. Insects were starved for about 2 weeks before the infection, upon which each bug was marked with a small dot of non-toxic water-based paint at the top of the pronotum, and then weighed before and after feeding to estimate the number of parasites ingested. Only females were fed infected blood, while the females from the control group and all males were fed uninfected blood. A total of 5–10 insects were grouped in small jars, which were then placed under a membrane feeder containing defibrinated, de-complemented human blood (heated to 37·5 °C) supplemented with inactivated FBS with an estimated concentration of 3·3–3·5 × 106 parasites/mL. This concentration falls within (a) the range of peak parasitaemias observed in mice and guinea pigs experimentally infected with T. cruzi (Bice and Zeledon, Reference Bice and Zeledon1970; Urdaneta-Morales and Rueda, Reference Urdaneta-Morales and Rueda1977; Perlowagora-Szumlewicz and Muller, Reference Perlowagora-Szumlewicz and Muller1982; Schaub and Losch, Reference Schaub and Losch1989a ; Schaub et al. Reference Schaub, Grünfelder, Zimmermann and Peters1989; Kollien et al. Reference Kollien, Schmidt and Schaub1998) and T. rangeli (Urdaneta-Morales and Tejero, Reference Urdaneta-Morales and Tejero1986; Zuñiga et al. Reference Zuñiga, Penin, Gamallo and de Diego1997a , Reference Zuñiga, Paláu, Penin, Gamallo and de Diego b ), and oral infectious doses used in prior published studies of T. cruzi and T. rangeli infection in triatomines (Garcia et al. Reference Garcia, Mello, Azambuja and Ribeiro1994, Reference Garcia, Machado and Azambuja2004; Mello et al. Reference Mello, Azambuja, Garcia and Ratcliffe1996; Ratcliffe et al. Reference Ratcliffe, Nigam, Mello, Garcia and Azambuja1996; Whitten et al. Reference Whitten, Mello, Gomes, Nigam, Azambuja, Garcia and Ratcliffe2001; Borges et al. Reference Borges, Machado, Garcia and Azambuja2006; Araújo et al. Reference Araújo, Cabello and Jansen2007, Reference Araújo, Waniek and Jansen2014; Nogueira et al. Reference Nogueira, Gonzalez, Gomes, de Souza, Garcia, Azambuja, Nohara, Almeida, Zingales and Colli2007; Mejía-Jaramillo et al. Reference Mejía-Jaramillo, Peña and Triana-Chávez2009; Ferreira et al. Reference Ferreira, Lorenzo, Elliot and Guarneri2010; Castro et al. Reference Castro, Moraes, Gonzalez, Ratcliffe, Azambuja and Garcia2012, Reference Castro, Peterson, Saldaña, Perea, Calzada, Pineda, Dobson and Gottdenker2014; Fellet et al. Reference Fellet, Lorenzo, Elliot, Carrasco and Guarneri2014). Trypanosoma cruzi–T. rangeli co-infections were carried out at a similar total parasite concentration, consisting of equal concentrations of each species (i.e. 1·65 × 106 of each parasite species/mL of blood, for a total of 3·8 × 106 parasites/mL of blood).
Insect reproduction
After moulting into the adult stage, we paired each female with a recently fed adult male (Buxton, Reference Buxton1930; Davey, Reference Davey1965). Males were paired with females of just one treatment group throughout the experiment to avoid cross-contamination. Copulation was determined 1 day after insects were paired from the presence of the spermatophore casing in the jar, ejected by the female (Ruegg and Davey, Reference Ruegg and Davey1979). If we did not find the spermatophore casing after the first night, then 2–3 additional males were placed in the jar with the female, and left for another night (G. Chiang, personal Communication, 2013). If copulation did not occur after three nights with several males, then we recorded the female as unmated for that oviposition cycle. Unmated individuals from the first oviposition cycle were given a second opportunity to mate for the second oviposition cycle. After mating, females were fed on hens (males were fed 3–4 days prior to copulation for sperm production). We marked each female with a small coloured dot of non-toxic, water-based paint on the pronotum (Mac Cord et al. Reference Mac Cord, Jurberg and Lima1983; Henriques et al. Reference Henriques, Castro, Gomes, Garcia and De Souza2012), weighing it before and after feeding to calculate the volume of blood ingested. We recorded oviposition and eclosion 3–4 times per week until the second oviposition cycle, 31–38 days later.
We measured reproduction as fecundity (egg production) and the percentage of oviposited eggs that hatched. Fecundity in R. prolixus is correlated with the quantity of blood ingested and weight before feeding (Friend et al. Reference Friend, Choy and Cartwright1965), and the standard index used when comparing fecundity in R. prolixus is the E value (Ruegg and Davey, Reference Ruegg and Davey1979). The E value is calculated as the total number of eggs produced by a given individual divided by the product of the blood meal volume multiplied by its pre-feeding weight. This represents the efficiency with which the insect converts nutrition (blood) into food, while normalizing for blood and insect mass, allowing for comparison across feedings. The E value is independent of the timing of the oviposition cycle in an insect's lifetime. In analysing the E values, we did not include insects that died before an oviposition cycle began (i.e. resulting in an E value of 0), in order to compare E value independent of mortality rate. In addition to these measurements, time-dependent reproductive values were also generated for each individual in our fitness analyses, described below.
Infection confirmation
After insect death, we extracted total DNA from each insect using Qiagen DNeasy blood and tissue kit. Additionally, we extracted DNA from pooled males and pooled offspring to check for horizontal and vertical transfer of parasites. We amplified DNA in an RT–PCR (StepOnePlus Real-Time PCR System, Applied Biosystems), with the T. cruzi primer pair [TcZ1/2 (Cummings and Tarleton, Reference Cummings and Tarleton2003)] and R. prolixus reference gene primer (RP18S, Paim et al. Reference Paim, Pereira, Di Ponzio, Rodrigues, Guarneri, Gontijo and Araújo2012). To obtain a T. rangeli-specific primer of the optimal size [<150 base pairs (bp)] that did not cross-amplify T. cruzi, we designed a primer denoted as ‘PEEL5’ −F (5′-TGCTTTCGTAGTTGGCACTG-3′) and −R (5′-ACGCACCTCCTCCTCTCTCT-3′), which amplifies a 93 bp fragment of T. rangeli telomeric DNA. We designed this primer from the T. rangeli clone TrTel 10 telomeric sequence (GenBank ID: AF426020·1), using the Primer3 plus software (Untergasser et al. Reference Untergasser, Nijveen, Rao, Bisseling, Geurts and Leunissen2007).
Statistical analyses
We carried out all statistical analyses using the R statistical computing environment software version 3.03 (R Core Team, 2014) using non-parametric tests to avoid normality assumptions. We tested for differences between treatments in the amount of parasites or blood ingested per unit of insect mass using the Kruskal–Wallis rank sum tests. We tested for differences in the amount of parasites or blood ingested per unit of body weight using Wilcoxon Rank Sum tests. We applied the ‘kruskalmc’ function from the ‘pgirmess’ package (Giraudoux, Reference Giraudoux2013) to carry out multiple comparisons and control for family wise error when a difference was found in Kruskal–Wallis tests. This function implements comparisons between treatments, and one- and two-tailed comparisons vs control. We accepted P-values under 0·05 as statistically significant.
We analysed survival function for each treatment group using the Kaplan–Meier (K–M) method in the R ‘survival’ package (Therneau and Grambsch, Reference Therneau and Grambsch2000; Therneau, Reference Therneau2015). We compared survival function (the probability of total time until failure) between treatment groups using the ‘survdiff ’ function in the ‘survival’ package, a two-tailed test for censored data that implements the G–ρ family of tests (Harrington and Fleming, Reference Harrington and Fleming1982), where deaths at various times are weighted by a factor of S(t)^ρ (S = K–M estimate; t = time), and ρ is a scalar parameter that determines the type of test used. When set at 0, all deaths are weighted equally across time and a log-rank test is used. When set at 1, deaths at the beginning of the time period are more heavily weighted, and the Peto and Peto test (Peto and Peto, Reference Peto and Peto1972) is employed. We set ρ at 1, to offset insect death events related to senescence. We carried out pairwise comparisons between K–M survival curves with Chi-squared (χ 2) distribution tests and adjusted P-values to control for the familywise error rate using the Holm–Bonferroni correction method (Holm, Reference Holm1979).
We used Cox proportional hazards (PH) models (Cox, Reference Cox1972) to examine the main effects and two-way interactions of parasite treatment, parasite dose and blood ingested on treatment hazard rates (the instantaneous rate of failure at any given time, given that the individual has survived up until that time). The PH assumption, (i.e. hazards were proportional over time) was tested with the Coxph function in the ‘survival’ package. We selected model covariates using Akaike's Information Criterion (AIC) with the stepAIC function in the ‘MASS’ package (Venables and Ripley, Reference Venables and Ripley2002), and manual one-variable-at-a-time reduction.
We log2 transformed parasite dose data, and centred them on the log2 transformation of 5·0 × 105 parasites, the round number closest to the mean. We used the Predict function from the ‘rms’ package (Harrell, Reference Harrell2014) to estimate log relative hazards and their 95% confidence intervals based on 1000 simulations of the model.
We ran the Cox model with three variations. In the first variation, we investigated the interaction between treatment and blood:weight ratio, and compared the parasite treatment group hazards with the control hazard. In the second and third variations, we included only parasite treatment groups to investigate relative hazard. To control for a possible effect of absolute number of parasites vs relative number of each parasite species in the mixed parasite species dose, we ran the model with data for the absolute number of parasites ingested by the mixed group in the second variation. In the third variation, we ran data for the mixed group as the relative number of each parasite species ingested. This does not change the power of the model or the summary statistics; the change was reflected only in effect size. Cox PH model outputs are in Tables S1–S3 in the Supplementary Materials.
Fitness estimates
We used individual survival and reproduction data to construct an age-classified population projection matrix for each insect (McGraw and Caswell, Reference McGraw and Caswell1996; Twombly et al. Reference Twombly, Clancy and Burns1998). Each matrix was 3 × 3, with age-specific survival (P i ) on the sub-diagonal [always 0 or 1 in individual matrices (McGraw and Caswell, Reference McGraw and Caswell1996)], and age-specific realized reproductive output (F i ) in the first row. All other matrix elements were zeros. Each time step (t i ) in the matrix represented one month (with t 0 being the day of insect infection). The model for each individual A was constructed as:
$$A=\matrix{0 \hfill \quad {\rm F}_2 \hfill \quad {\rm F}_3 \cr 1 \hfill \quad 0 \hfill \quad 0 \cr 0 \hfill \quad 1 \hfill \quad 0}$$
The dominant eigenvalue (λ) of each matrix is a maximum-likelihood estimate of individual fitness, with values above one indicating population growth, and values below one indicating population shrinkage. The dominant left eigenvector of each matrix is an estimate of individual reproductive value v i for each time step. We calculated dominant eigenvalues (λ) using the eigenfunction in the R base package, and reproduction values were calculated by hand based on these values, as in McGraw and Caswell (McGraw and Caswell, Reference McGraw and Caswell1996; based on Fisher, Reference Fisher1930). The reproductive value for t 1 (v 1) is scaled to one, and other values are given relative to v 1. In an individual population projection model where F 1 is equal to 0, v 2 is equal to lambda. The model assumes a closed population with unlimited resources, no genetic structure, and does not account for effects of population density.
RESULTS
Parasites ingested
Insects ingested between 30·1 and 337·9 mg of blood (mean 214·8 mg), and an estimated 62 000–1 079 000 total parasites (mean 708 000). The ratio of the volume of blood ingested to insect pre-feeding weight ranged from 0·99 to 14·25 (mean 8·23), and the ratio of the estimated number of parasites ingested per mg of insect biomass ranged from 2000 to 48 000 parasites (mean 28 000). There were no differences between treatments in the absolute parasites dose, nor were there any linear relationships between the parasite dose and death day, E value, reproductive value or estimate of total fitness. There was a significant difference between treatments in the ratio of the volume of blood ingested per mg of insect biomass (Kruskal–Wallis, blood: P = 1·67 × 10−4; parasites: P = 0·01), with the mixed group ingesting significantly more blood than the T. cruzi or control groups (Fig. 2; KruskalMC, P < 0·05 for comparisons).
Fig. 2. The distribution of the ratio of the volume of blood consumed in the infective blood meal to mg of insect biomass, across treatments. The mixed group blood:weight ratio was significantly higher than that of the T. cruzi and control groups.
Reproduction
87·8–97·6% of insects in each group laid eggs, and there was no significant difference between treatment groups in this respect. The E values were significantly different between treatments in both the first and second oviposition cycles (Kruskal–Wallis; cycle 1: P = 8·98 × 10−8; cycle 2: P = 3·24 × 10−4, Fig. 3A and B). In both cycles, E values for the T. cruzi or T. rangeli treatments were significantly lower than the co-infected treatment E values (Kruskalmc, P < 0·05). The T. cruzi treatment had a significantly lower E value than the control group in cycle 1 only (Kruskalmc, P < 0·05). The mean percentage of oviposited eggs that hatched ranged between 79·4 and 84·3% for cycle 1; 62·4–81·8% for cycle 2; and 77·6–83·7% overall. The percentage of eggs that hatched was not significantly different between treatments. Additionally, there was no association between E value and per cent of eggs hatched.
Fig. 3. E value distributions in each treatment group for oviposition cycle 1 (left) and oviposition cycle 2 (right). In both cycles, the mixed group E values were significantly higher than the T. cruzi and T. rangeli treatment group E values. The control group E values were significantly higher than the T. cruzi and T. rangeli treatments in cycle 1. In cycle 2, the control group is higher than just the T. rangeli treatment.
Survival function
K–M survival curves (representing survival function, i.e. the probability of total time until failure), were significantly different from each other (χ 2 = 8·4, 3 df, P = 0·03, Fig. 4). The T. cruzi treatment group had a significantly shorter time to failure than the mixed treatment group (χ 2 distribution comparisons, P < 0·05).
Fig. 4. K–M survival curves for each treatment group. The T. cruzi treatment survival function was significantly different than that of the mixed group.
Hazards analysis
The Cox model variation investigating the interaction of treatment with blood:weight ratio was significant (Likelihood ratio test, 24·67, 7 df, P = 8·67 × 10−4; Supplementary Materials Table S1), suggesting hazard (i.e. instantaneous risk of death) was not the same between treatment groups even when blood meal and body size were taken into account. Investigating the blood:weight ratio allowed us to control for differences in insect size by measuring the effect of the quantity of blood (and also therefore, number of parasites for infected groups) per unit of body mass. Quantity of blood was used in the calculation rather than parasite dose to be able to include the control group. The main effects of T. cruzi treatment were significant, with a hazard 2·17 times that of the control group (eβ = 2·17, P = 4·33–04). The control and mixed treatments interacted significantly with the blood:weight ratio, but in opposite directions; the control group hazard increased as the blood:weight ratio increased, while the mixed group hazard decreased with increases in the blood:weight ratio (control: eβ = 1·26, P = 1·55 × 10−3; mixed: eβ = 0·74, P = 5·64–03, Fig. 5).
Fig. 5. Interaction of treatment with the blood:weight ratio of the infective blood meal. Hazards were predicted after 1000 simulations of the model. Figures are centred on the mean ratio, 8·23. Grey shading indicates 95% confidence intervals. Just the interactions in the bottom row (the control and mixed treatment groups) were significant.
The Cox model investigating the main and interaction effects of parasite dose was also significant (Likelihood ratio test, 29·63, 5 df, P = 1·74 × 10−5). The patterns and significant effects were the same in both variants of the model (examining the effect of absolute vs relative parasite dose), with effects being slightly larger in the model investigating absolute parasite dose. In both model variations there were no differences in the main effects of treatment on hazard. Main effects of parasite dose were significant for T. rangeli and marginally significant for T. cruzi, with a 3-fold increase in hazard at a dose of 1 million parasites from the hazard at 500 000 parasites (T. rangeli: eβ = 3·27, P = 4·33–04; T. cruzi: eβ = 3·07, P = 6·5 × 10−2). Effects of the interaction between treatment and parasite dose were significant for the mixed group in both model variations (absolute and relative parasite doses of the mixed group). At 250 000 parasites, the mixed group hazard was significantly higher than either single-species infection treatment, while at 1 million parasites the mixed group hazard was significantly lower (mixed vs T. cruzi: P = 0·025; Mixed vs T. rangeli, P = 0·00006; full summary in Supplementary Materials, Tables S2 and S3). Interaction effects were not significant when comparing the T. cruzi treatment with the T. rangeli treatment, suggesting their hazards were not significantly different from each other at any parasite dose.
Fitness
Fitness estimates (λ) and reproductive values v2 and v3 (corresponding to 60 and 90 days) were significantly different between treatments (Kruskal–Wallace; λ and v 2: P = 1·69 × 10−7; v3: P = 1·42 × 10−2), with T. cruzi and T. rangeli treatment groups having significantly lower λ and v 2 values than the mixed and control groups (KruskalMC, P < 0·01, Fig. 6). The reproductive value at 90 days (v 3) was significantly different between the T. cruzi and mixed group, with T. cruzi being lower (KruskalMC, P < 0·05). The T cruzi and T. rangeli treatment group fitness estimates and reproductive values were not significantly different from each other at any time point.
Fig. 6. Distribution of fitness estimates in each treatment group. The control and mixed groups had significantly higher fitness estimates than the T. cruzi and T. rangeli groups.
Infection status at death
The difference between treatment groups in the proportion of samples that amplified in the qPCR was marginally non-significant (Fisher's Exact Test for Count Data, P = 0·09), although there were no significant differences after performing individual comparisons between each treatment and adjusting the P-values for multiple comparisons. 90% of T. cruzi treatment group samples amplified; 76·92% of T. rangeli samples amplified; 61·53% of samples from the mixed treatment group amplified T. cruzi; and 84·61% amplified T. rangeli. There was no parasite DNA amplification for the pooled male and offspring groups.
DISCUSSION
Co-infection: advantageous for the host and parasite?
We observed that insects co-infected with T. cruzi and T. rangeli had higher survival, reproduction and overall fitness, suggesting that T. Cruzi–T. rangeli co-infection could reduce negative life history consequences of a single infection with T. cruzi or T. rangeli for R. prolixus. This could in turn, lead to increases in the transmission potential of T. cruzi and/or T. rangeli. Additionally, this could be a way that virulent strains persist, especially T. rangeli, which, as mentioned, is known to be pathogenic to R. prolixus. Reported prevalences of T. cruzi–T. rangeli co-infection in field-caught triatomines have been found to be higher than single infections of T. rangeli in R. prolixus (Groot, Reference Groot1951; Vallejo et al. Reference Vallejo, Marinkelle, Guhl and de Sanchez1988), R. pallescens (Pineda et al. Reference Pineda, Montalvo, Alvarez, Santamaría, Calzada and Saldaña2008; Calzada et al. Reference Calzada, Pineda, Garisto, Samudio, Santamaria and Saldaña2010; Gottdenker et al. Reference Gottdenker, Chaves, Calzada, Peterson, Santamaría, Pineda and Saldaña2016); and R. colombiensis (Pavia et al. Reference Pavia, Vallejo, Montilla, Nicholls and Puerta2007), which would support the idea of a co-infection advantage for T. rangeli (Fig. 1). However, more data on fitness in trypanosome-infected field-caught triatomines are needed to support this result.
Additionally, we found a threshold parasite dose below which insects infected with a single species infection had a lower instantaneous hazard rate (i.e. risk of death) and above which co-infected insects had a lower risk. This might increase the transmission potential of the parasites if the parasite dose in the blood meal were associated with higher numbers of parasites transmitted by the bugs. However, T. cruzi infective dose does not correlate with the number of parasites excreted (Wood, Reference Wood1954; Urdaneta-Morales and Rueda, Reference Urdaneta-Morales and Rueda1977; Chowdury and Fistein, Reference Chowdury and Fistein1986; Azambuja et al. Reference Azambuja, Feder and Garcia2004, Reference Azambuja, Garcia and Ratcliffe2005), and the total trypanosome population size and composition (proportion of each form present) within a triatomine will fluctuate with feeding status; significant decreases in parasite numbers can occur within 4 h after feeding by as much as 50% in some parts of the bug (Schaub and Lösch, Reference Schaub and Lösch1988; Schaub, Reference Schaub1989b ; Kollien and Schaub, Reference Kollien and Schaub1998a ). Thus, it seems unlikely that the higher infective doses tolerated by co-infected insects increase the parasites’ transmission potential, aside from increasing the transmission probability by keeping the insect alive longer.
Insect reproduction: quality vs quantity
While the efficiency of egg production seemed to be affected by parasite treatment, the per cent of oviposited eggs that hatched was not. It is known that the processes of egg growth and oviposition are controlled separately in R. prolixus (Mundall, Reference Mundall1978). Oviposition of badly formed eggs, which has been observed in Cimex species, is rare, even in cases of insect malnutrition (Buxton, Reference Buxton1930). This investment in egg quality over quantity could be a mechanism of insecticide resistance, which has been observed in T. infestans eggs (Toloza et al. Reference Toloza, Germano, Cueto, Vassena, Zerba and Picollo2008), and could be one factor that explains residual populations in human homes after insecticidal spraying.
T. cruzi vs T. rangeli virulence
As mentioned, T. rangeli is considered to be pathogenic to triatomines of the genus Rhodnius, while T. cruzi has been described in several publications as ‘subpathogenic’ (Schaub, Reference Schaub1989a , Reference Schaub1990, Reference Schaub1992; Schaub and Losch, Reference Schaub and Losch1989a ), i.e. pathogenic only in the presence of external stress. In this light, it is surprising that the fitness of the treatment group infected with T. cruzi was not significantly higher than the fitness of the T. rangeli treatment group. However, the majority of studies supporting the subpathogenic theory of T. cruzi in triatomines have been carried out in the species T. infestans (Schaub, Reference Schaub1988a , Reference Schaub b ; Schaub and Lösch, Reference Schaub and Lösch1988; Schaub and Losch, Reference Schaub and Losch1989a , Reference Schaub and Losch b ; Kollien and Schaub, Reference Kollien and Schaub1998a , Reference Kollien and Schaub b ; Kollien et al. Reference Kollien, Schmidt and Schaub1998). Most studies investigating effect of T. cruzi on R. prolixus life history have found a mild effect (D'Alessandro and Mandel, Reference D'Alessandro and Mandel1969; Neves and Peres, Reference Neves and Peres1975; Fellet et al. Reference Fellet, Lorenzo, Elliot, Carrasco and Guarneri2014), and effects have also been observed in Panstrongylus megistus (Lima et al. Reference Lima, Borges-Pereira, Albuquerque Dos Santos, Teixeira Pinto and Vianna Braga1992) and Mepraia spinolai (Botto-Mahan, Reference Botto-Mahan2009). Additionally, Añez et al. (Reference Añez, Molero, Valderrama, Nieves, Cazorla and Márquez1992) also found no significant difference in development or mortality between insects infected with T. cruzi and insects infected with T. rangeli. Moreover, recent studies have found that T. cruzi can negatively affect R. prolixus life history outcomes, depending on temperature (Fellet et al. Reference Fellet, Lorenzo, Elliot, Carrasco and Guarneri2014; Elliot et al. Reference Elliot, Rodrigues, Lorenzo, Martins-Filho and Guarneri2015) and parasite strain (Peterson et al. Reference Peterson, Graham, Dobson and Chavez2015). This could be due to increased parasite replication rates at higher temperatures (Wood, Reference Wood1954; Asin and Catalá, Reference Asin and Catalá1995). However, the insects in this study were reared under climate conditions similar to those found in R. prolixus-endemic areas of Colombia (Hoyos et al. Reference Hoyos, Pacheco, Agudelo, Zafra, Blanco and Triana2007; Gutierrez et al. Reference Gutierrez, Trujillo Güiza and Escobar Martínez2013), thus, if temperature were an underlying factor in T. cruzi virulence, it would suggest that T. cruzi may also be virulent to free-living Colombian R. prolixus.
Our survival results are not in agreement with the other published study of R. prolixus survival when co-infected with T. cruzi and T. rangeli, which found that insects with mixed infections had higher mortality (Añez et al. Reference Añez, Molero, Valderrama, Nieves, Cazorla and Márquez1992). This could be due to differences in temperature between the studies (ours was carried out at higher temperatures), insect stage and/or parasite strains. Considering the high degree of polymorphism found within both the T. cruzi and T. rangeli species, it seems possible that the outcome of triatomine infection with either or both trypanosomes could lie in a wide range of outcomes from mildly virulent to positive. In our work, we have observed a wide range of survival in insects infected with different T. cruzi DTU I strains (Peterson et al. Reference Peterson, Graham, Dobson and Chavez2015).
Concluding remarks
Due to the inherent limitations of laboratory experiments, the extrapolation of effects observed in the laboratory to their meaning in the natural system must be carried out cautiously. That said, our findings suggest that some T. Rangeli–T. cruzi co-infections could ameliorate the negative effects of single-species infections, allowing more virulent strains to persist and potentially increasing the transmission potential of both parasites. Further research into T. cruzi–T. rangeli co-infections in other triatomine systems and in field-caught bugs will provide more insight into this topic.
SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at http://dx.doi.org/10.1017/S0031182016000615.
ACKNOWLEDGEMENTS
We would like to extend our gratitude to Germán Rodriguez from the Office of Population Research at Princeton University for help with the survival analyses, and Professor Gustavo Vallejo from the University of Tolima for providing the T. rangeli culture used in this work.
FINANCIAL SUPPORT
This work was supported by funding from the Universidad de Antioquia (to O.T.C.), the Princeton Program in Latin American Studies (to J.K.P.) and the Princeton Institute for International and Regional Studies (to J.K.P.).
COMPETING INTERESTS
The authors declare that they have no competing interests.
AUTHORS’ CONTRIBUTIONS
J.K.P. conceived and designed the study, carried out all assays, performed the statistical analysis, and drafted the manuscript. A.L.G. participated in the study design, helped with statistical analysis and critically revised the manuscript. R.J.E. helped with insect and molecular assays and helped with the analyses of the results; A.P.D. participated in the study design and revised the manuscript. O.T.C. participated in the design of the study, helped to coordinate the experiments, and critically revised the manuscript.
