Measuring dengue virus transmission in endemic areas is a difficult task as many variables drive transmission, and often are not independent of one another.

We aimed to determine the utility of vectorial capacity to explain the observed dengue infection rates in three hyperendemic cities in Colombia, and tested hypotheses related to three variables: mosquito density, effective vector competence, and biting rate.

We estimated two of the most influential entomological variables related to cumulative vectorial capacity, which is a modification of the traditional vectorial capacity equation, of three Colombian mosquito populations. Laboratory studies were undertaken to measure vector competence and man biting rate of local mosquito populations. In addition, the assessment of cumulative vectorial capacity also incorporated site-specific estimations of mosquito density and the probability of daily survival from previous studies conducted in those cities.

We found that the biting rates and mosquito infection rates differed among populations of mosquitoes from these three cities, resulting in differences in the site-specific measures of transmission potential. Specifically, we found that using site-specific entomological measures to populate the cumulative vectorial capacity equation was best at recapitulating observed mosquito infection rates when mosquito density was discounted compared to when we incorporated site-specific density measures.

Specific mosquito-biting rate is likely sufficient to explain transmission differences in these three cities, confirming that this parameter is a critical parameter when predicting and assessing dengue transmission in three Colombian cities with different field observed transmission patterns.

Dengue virus (DENV) is a mosquito-borne virus that circulates in more than 100 countries around the world, [

Quantifying and predicting transmission of DENV can be difficult because of the number of variables needed to capture the nuances of transmission of just a single serotype. This becomes even more complex with 1) the interactions of more than one serotype, sometimes all four at once, and 2) when heterogeneity of individual variables due to environmental or sociological factors is included. However, investigations of these variables – either singly or in combination – can still provide valuable insight into the factors that drive local transmission patterns. One approach is vectorial capacity (VC), which is a measure of the intensity of transmission in a particular area by a mosquito population [

Where

The entomological and epidemiological patterns of three Colombian cities were described previously [

Immature stages of mosquito were collected from samplings performed between mid-2012 and 2013 in the Colombian cities of Riohacha, Bello, and Villavicencio. Details about samplings, climatic and entomological characteristics are provided in Peña-García et al., 2016 [

As a control population, laboratory maintained mosquitoes belonging to Chetumal colony were raised and maintained under same conditions described above at the Arthropod-Borne and infectious disease laboratory of Colorado State University [

A strain of DENV-2 (Jamaica1409) belonging to the Asian/American genotype, the same genotype circulating in Colombia [

Infections were conducted according to Sánchez-Vargas et al. [^{6} PFU/mL. Mosquitoes of 4–6 days post-emergence from the four groups were fed on the mixture by using an artificial feeder with circulating water at a temperature of 37°C. Fully engorged mosquitoes were separated and approximately 60 mosquitoes of each city were dissected at 4, 7, 11 and 14 days after exposure by extracting midgut and heads. Samples were divided in half to be processed by immunofluorescence assay (IFA) and plaque assay. Midguts designated to immunofluorescence were fixed with 4% paraformaldehyde for at least 12 hours at 4°C and heads were squashed onto slides and fixed with acetone: PBS (3:1) for 30 minutes at –20°C. Samples processed by plaque assay were stored at –80°C until use.

To determine infection dynamics within the mosquitoes, IFA assays were performed with some modifications as done by Salazar et al. [

In addition, we used plaque assays on samples of midgut and head samples to titer the virus present in those tissues. The experiments were conducted following Sánchez-Vargas et al. [_{2} in 24-wells plates. After 1 hour of infection, cells were overlaid with agar-nutrient mixture (medium 199 1X, fetal bovine serum 7%, NaHCO_{3} 0,3%, Hank’s balanced salt solution with dextran 2%, DEAE 1%, MEM Vitamins 0,5X and MEM aminoacids 0,25X) and incubated at 37°C, 5% CO_{2} for 10 days. Plaques were counted after plates were incubated overnight with MTT 3 mg/mL solution. Data are reported as plaque forming units (PFU) per mL.

To estimate this variable, we developed a new methodology by hypothesizing that the probability of a single mosquito feeding depends on the day after emergence of that mosquito. Specifically, we hypothesized that there was a greater likelihood of blood feeding at later days after emergence. Thus, we evaluated the likelihood of blood feeding of female mosquitoes at 1, 3, 5, and 6 days after emergence on healthy volunteers. Adult female mosquitoes were supplied with water and sugar

Five non-infected/non-exposed females of the same age were released in a cage of 0.43 m^{2} (120 cm long, 53 cm width and 68 cm tall). A healthy volunteer introduced an arm in the cage for 15 minutes based on observations from Canyon and Hii (1997) [

In order to obtain a model for the bite probability of mosquitoes through time, a logistic model was fit to the explanatory variable of time (given as days) to represent days post emergence. The resulting model was further used to predict the day at which 99% of mosquitoes would have fed given simultaneous emergence. The total man-biting rate was estimated by calculating the area under curve resulting from the model by using a regular trapezoid method through to 14 days. This value (area under the curve) was divided by 14 to obtain an average bites per day, but does not take into account mosquito age. Thus, our final model assumes that mosquitoes at all ages are circulating at the same time with a constant (average) biting rate.

In addition, we estimated the time (in minutes) that an individual mosquito takes to bite. This was done by recording time since introduction of the volunteer’s arm in the cage until a mosquito introduces its proboscis into the skin. This variable was called “time to bite” and was analyzed using Kruskal-Wallis test to check if there were differences among volunteers or populations. Also, we recorded the time since introduction of proboscis in the skin until it was removed; this variable was called “total feed time”. The “total feed time” data was log-transformed to be analyzed by ANOVA. We analyzed these variables in order to find differences among populations and to analyze if volunteer has a differential effect on bite.

Estimates of mosquito density are taken directly from fieldwork data reported by Peña-Garcia and co-workers in 2016 [

The average number of mosquitoes per house was then used to obtain a ratio of mosquitoes per person as an estimate of mosquito to human host density [

Because there was such a disconnect in observed mosquito density and DENV incidence, we further investigated the sensitivity of our measures as relates to mosquito density and transmission by consideration of two density values in vectorial capacity: 1) Mosquito density estimated as described before in this paper (section Estimation of mosquito density) and 2) taking a value of 1, which results in nullifying the mathematical contribution of the variable [

Life tables for the three mosquito populations were constructed in parallel (Pérez L et al., in preparation). The number of adult mosquitoes surviving each day along the entire adult survival curve was used to estimate a probability of mosquito to survive each day. The total probability of daily survival was estimated by averaging the probability of all days evaluated. Data of life tables from each city were developed at two different temperatures, we averaged both datasets to obtain a unique value per population.

Since plaque assays quantify infectious viral particles, we used plaque assay data from head tissues at four studied days to determine effective vector competence (EVC) according to a modification of the methodology developed by Christofferson and Mores [

We modified the EVC methodology by fitting vector competence values at day post exposure to a logistic regression instead of a linear function. Thus, the function we used follows the form:

Where _{1} is the change in vector competence (_{0} is the y-intercept. The subsequent iterative calculations include the estimation of area under curve taking just the final two time points, which include the cumulative proportions of competent mosquitoes. Thus, the EVC was estimated taking an upper limit of

And consequently, the estimation of cVC is defined by the following equation:

Where

Ethical Approval (Act N° 13, 29/07/2015) was obtained from the Bioethics Committee of the University of Antioquia, Medellin, Colombia.

As a control group, we determined that Chetumal colony mosquitoes had high mid-gut infection rates as determined by IFA: 48.1%, 79%, 60%, and 63% at 4, 7, 11 and 14 days post-exposure (dpe), respectively (Table

Midgut infection percentages of mosquitoes from the three field-caught origins and one laboratory colony and the log-odds of getting a positive midgut compared to the baseline of colony origins (Chetumal colony) at 4, 7, 11 and 14 days post-infection (dpi) according to IFA.

Percentage of midgut Infection (n) | Log-Odds | |||||||
---|---|---|---|---|---|---|---|---|

dpi | 4 | 7 | 11 | 14 | 4 | 7 | 11 | 14 |

Chetumal | 0.48 (27) | 0.79 (29) | 0.6 (30) | 0.63 (27) | . | . | . | . |

Riohacha | 0.33 (27) | 0.47 (30) | 0.6 (30) | 0.5 (30) | –0.148 | –0.326* | <0.0001 | –0.13 |

Bello | 0.56 (25) | 0.48 (33) | 0.53 (28) | 0.48 (27) | 0.079 | –0.308* | –0.0064 | –0.15 |

Villavicencio | 0.58 (26) | 0.56 (25) | 0.58 (24) | 0.2 (30) | 0.095 | –0.233 | –0.016 | –0.43* |

* statistically significant at 95% confidence.

We found that at 4 dpi, there was no significant difference in the log-odds of getting a positive midgut or head between any of the cities and the colony mosquitoes (Tables

Head infection percentages of mosquitoes from the three field-caught origins and one laboratory colony and the log-odds of getting a positive head compared to the baseline of colony origins (Chetumal colony) at 4, 7, 11 and 14 days post-infection (dpi) according to IFA.

Percentage of head Infection (n) | Log-Odds | |||||||
---|---|---|---|---|---|---|---|---|

dpi | 4 | 7 | 11 | 14 | 4 | 7 | 11 | 14 |

Chetumal | 0.03 (30) | 0.2 (30) | 0.5 (30) | 0.63 (30) | . | . | . | . |

Riohacha | 0 (30) | 0.13 (30) | 0.4 (30) | 0.47 (30) | –0.033 | –0.067 | –0.1 | –0.167 |

Bello | 0 (30) | 0 (30) | 0.07 (30) | 0.17 (30) | –0.033 | –0.2* | –0.43* | –0.467* |

Villavicencio | 0.03 (30) | 0 (30) | 0.2 (30) | 0.23 (30) | <.0001 | –0.2* | –0.3* | –0.4* |

* statistically significant at 95% confidence.

Pairwise comparisons among population viral titers in the midgut and head tissues reveal that the highest discrepancies were between Riohacha and the populations of Bello and Villavicencio, where Riohacha had higher viral titers, most notable at 7 and 14 dpi (Figure

Viral titers at 4, 7, 11 and 14 days post-infection in the midgut and head tissues of mosquitoes from Riohacha, Bello, Villavicencio and the control population of Chetumal. Number of processed mosquitoes per city per dpi equals to 30. Significant differences are represented by *(

All three cities had significant Pearson correlation coefficients (Riohacha, Pearson = .55; Villavicencio, Pearson = .51; Bellow, Pearson = .45), which suggests that the susceptibility of all three populations to developing a disseminated infection was relatively predicted by the ability to become infected.

A Kruskal-Wallis test did not find significant differences in “time to bite” among cities (^{2} = 2.61, df = 2, ^{2} = 8.02, df = 4,

However, a significant difference in the frequency of successful biting among populations from the three cities was observed through Kruskal-Wallis test (^{2} = 9.65, df = 2,

When the data were fitted to a logistic model, the average estimated bites per day was higher in Bello (0.76 bites per day), followed by Villavicencio (0.62 bites per day), and Riohacha (0.59 bites per day) (Figure

Curves and mathematical functions fitted to bite data from Riohacha

Bello had the lowest BI as observed previously [

To obtain updated values of mosquito density, each of these values was divided by the respective average number of people per house as reported by DANE (4.5 for Riohacha, 3.8 for Bello and 3.7 for Villavicencio). Thus, the vector to host ratio at household level of Riohacha, and Villavicencio were quite similar at 0.42 and 0.43, respectively. However, Bello was estimated to have a lower ratio at household level of 0.16 mosquitoes per person.

The probability of daily survival was observed to be similar among the three mosquito populations. Riohacha showed marginally lower survival with a probability of daily survival of 0.9295, while Bello and Villavicencio had similar values of 0.9304 and 0.9306, respectively.

When we calculated the EVC for each population, our results indicate that the highest effective vector competence was in the mosquito population from Riohacha (0.82), followed by Bello (0.57), and finally Villavicencio (0.34) (Figure

Distribution of bootstrapped EVC values for mosquitoes from Riohacha

The role of density in cVC estimation.

When we combined all of the site-specific parameters (Table

Parameters used to estimate cVC of each of the three cities.

City | EVC ( |
Avg. bites per day ( |
Density ( |
Probability of daily survival ( |
---|---|---|---|---|

Riohacha | 0.82 | 0.59 | 0.42 | 0.9295 |

Bello | 0.57 | 0.76 | 0.16 | 0.9304 |

Villavicencio | 0.34 | 0.62 | 0.43 | 0.9306 |

* Head positivity by plaque assay used as a proxy for transmission potential of DENV-2.

In a previous study, we demonstrated that mosquito infection rates of natural populations were highest in Bello, followed by Villavicencio and Riohacha [

Measuring the kinetics of arboviruses in

It is worth noting that since the development of vectorial capacity concept, several modifications of the original equation has been developed [

Further, when we used site-specific density estimates, we found that our measures did not follow that of observed field infection rates and that density-independent measures of cVC more accurately reflected field observations. This could suggest that a subset of mosquitoes is responsible for the majority of transmission events that is not readily captured in high-level measures of population density. We hypothesize that there is a critical threshold of transmitting mosquitoes necessary for virus perpetuation, but the factors that define this subset of mosquitoes has yet to be determined. It is critical that the role of density and the subset of infectious mosquitoes be further explored to add precision to future estimates and predictions of transmission. This would have obvious implications for vector control strategies that include population control (by insecticide spraying) versus in-door mosquito targeting, and may explain the disconnect between vector control implementation and a lack of decrease in dengue in some instances [

In addition, comparisons of site-specific transmission patterns of DENV-2 can be captured through the use of a mosquito density-independent cVC measure, and reconfirm that biting rate of the mosquito populations is an important and critical driver of transmission. Traditional approaches to measuring the man biting rate involve either indirect measures (human landing rates) [

The reasons why the mosquito population of Bello have higher biting rates compared to the other populations are unknown. Possible explanations include a lack of intraspecies competition for breeding sites close to domiciles, as well as the hosts themselves [

Our study is not without limitations. First, the use of a laboratory strain of DENV-2 cannot account for site-specific heterogeneity of the viral population, as there is vector competence variability both within and among strains and serotypes [

The transmission system of dengue virus is multifactorial, where numerous factors influence the probability of infection (e.g. social factors, environmental conditions) [

In summary, these results offer insights into the complex transmission patterns in three Colombian cities. We offer a refined methodology for estimating the site-specific man-biting rate of focal mosquito populations. We also demonstrate the necessity of estimating vector competence from field-caught mosquitoes by demonstrating intraspecific heterogeneity from relatively proximal communities, as well as explore circumstances under which site-specific consideration of mosquito density may not be as necessary. Ultimately, we demonstrate that studies of

Authors want to thank to COLCIENCIAS through project number 1115-725-53478 and to Universidad de Antioquia, UdeA for financial support. Also, this work was partially funded through grant: NIH/NIGMS R01GM122077.

COLCIENCIAS through project number 1115-725-53478, Universidad de Antioquia, UdeA; and grant NIH/NIGMS R01GM122077.

The authors have no competing interests to declare.