Use of Transcriptional Age Grading Technique to Determine the Chronological Age of Sri Lankan Female Dengue Mosquitoes; Aedes Aegypti and Aedes Albopictus

Background: Aedes aegypti and Ae. albopictus are important vectors of human diseases such as dengue, chikungunya, and zika. In Sri Lanka, they have been responsible for transmitting the dengue virus. One of the most important parameters that affect the likelihood of arbovirus transmission is the age structure of the mosquito population. However, mosquito age is dicult to measure with accuracy. This study aims to construct Multivariate calibration models using the transcriptional abundance of three age responsive genes: Ae15848 (calcium-binding protein), Ae8505 (structural component of cuticle), and Ae4274 (zzy cell cycle/cell division cycle 20). Methods: Transcriptional age grading technique was applied to determine the chronological age of Ae. aegypti and Ae. albopictus female mosquito strains from Sri Lanka using the age responsive genes; Ae15848, Ae8505, and Ae4274. Further to investigate the inuence of temperature on this age grading technique. Expression levels of these three genes were quantied using reverse transcription qualitative PCR (qRT-PCR) and results were normalized against the housekeeping gene ribosomal gene S17 (RsP17). Results: The expression of Ae15848 and Ae8505 decreased with the age of mosquitoes and, showed the most signicant and consistent change while expression of Ae4274 increased with age. The multivariate calibration models showed more than 80% correlation between expression of these age responsive genes and the age of female mosquitoes at both temperatures. At 27 0 C the accuracy of age predictions using the models was 2.19 (±1.66) and 2.58 (±2.06) days for Ae. aegypti and Ae. albopictus females, respectively. The accuracy of the model for Ae. aegypti at 23 0 C was 3.42 (±2.74). Conclusions: An adult rearing temperature difference of 4 0 C (23 0 C to 27 0 C) did not signicantly affect the age predictions. The calibration models created during this study could be successfully used to estimate the age of wild Ae. aegypti and Ae. albopictus

As there is no speci c antiviral vaccine against dengue, e cient and environmentally friendly vector control measures are needed to prevent the disease from spread and outbreaks throughout the world.
Introduction of arti cially infected Ae. aegypti females with life-shortening Wolbachia pipienis (wMelPop) into the eld has been introduced as one of the most promising dengue mosquito eradication technique in the world due to failures in traditional vector control measures. Wolbachia has the ability either to block the dengue virus inside the mosquito body and/or decrease the female mosquito life span than the EIP of the dengue virus. This environmentally friendly method of vector control has been successfully practiced in several areas in the world especially in Australia [5]. Knowledge of the age of the vector population is important before designing and releasing wMelPop infected individuals into natural populations to determine the effectiveness of this method.
Morphological, biochemical, and molecular-based age grading techniques have been developed in determining the age of female mosquitoes. However, the use of both morphological and biochemical methods have become questionable due to their inability and inaccuracy in measuring the age of mosquitoes older than the EIP. The transcriptional age-grading technique has now been identi ed as the most accurate and precise approach in determining the chronological age of mosquitoes [3,6,7].
Quanti cation of expression levels of genes that show a variation in its expression with the age of the female mosquito is the basis of this technique. The transcription scores of these age-responsive genes of laboratory-reared mosquitoes of known ages are then feed into a multivariate calibration model which could be later used in age predictions of eld/wild individuals. The mosquitoes used in the construction of the calibration model must be from the same mosquito strain of the eld population where the researcher is planning to apply the technique [6]. The analysis of age responsive genes during these studies have shown that this technique accurately detects the age of An. gambiae [8, 9,10] and Ae. aegypti [2,3,4,11,12] older than 15 days, that is more than the EIP period. Trials, using mosquitoes reared in eld cages, have concluded that the gene expression pro les of Ae. aegypti female mosquitoes could determine the age with an accuracy of ± 5 days of the actual age [11,13,2].
The orthologues of the eight age responsive genes i.e. Ae-4274, Ae-4679, Ae 4916, Ae-6639, Ae-7471, Ae-8505, Ae-12750, Ae-15848 selected from Drosophila melanogaster have initially been used to predict the age of female Ae. aegypti mosquitoes under both laboratory and eld conditions [6,14,15,16]. According to mosquito transcriptional age grading studies CG-8505/Ae-8505/AAEL003259 (Pupal cuticle protein 78E putative) and SCP-1/Ae.-15848/AAEL008844 (calcium binding protein, putative) displayed the largest and signi cant decrease in expression levels with the age of female mosquitoes while expression levels of zzy/Ae-4274/AAEL014025 ( zzy cell cycle/ cell physiology, putative) signi cantly increased with the age. Hence these three genes; Ae-8505 and Ae.-15848 and Ae-4274 have been identi ed as the most informative age-responsive genes that could be used in the transcriptional pro ling of mosquitoes [2,11,13]. Cook et al. [6] have identi ed Ae-8505 Ae-15848 and Ae-4274 genes as the most reliable age responsive genes and recommended to use these three genes for future age determination studies. The gene RsP17 (40S ribosomal protein s17) that showed an insigni cant variation with the age has been exclusively used as the reference gene in normalizing the samples in these studies. Further, the expression of these genes has are independent of blood feeding, egg laying, digestion, and reproductive status of the mosquitoes [6,8,17].
However, this approach needs further validation and optimization based on the geographical region, as the mosquito populations may have sequence polymorphism which only affects the reliability of gene expression analysis. Therefore, Cook et al. [6] suggested creating separate models for mosquitoes in different geographical regions. Further, uctuation in the environmental parameters such as temperature, humidity, photoperiod, etc. was also to be considered as these might affect the transcriptional abundance of age responsive genes [6]. Hence, models constructed considering all or majority of these limiting factors will increase the accuracy and precision of the transcriptional age grading method.
At present dengue has become one of the major causes of hospitalization and deaths in Sri Lanka. The year 2017 reported the largest dengue outbreak (≈ 186,000 cases) in the country which is around 4.3 times an increase in the average number of cases that have been reported from 2010 to the 2016 period [18]. Control of dengue vector populations in Sri Lanka is primarily based on the application of adulticides and larvicides. However, most of these programmes have been challenged due to the development of insecticide resistance by both the dengue vectors [19,20]. Hence, the Sri Lankan government has planned to release wMelPop infected dengue vectors into the eld [21]. However, so far no attempts have been made in Sri Lanka to determine the chronological age and age structure of any of the dengue vectors which is essential before implementing this programme. Both primary and secondary dengue vectors, Ae. aegypti and Ae. albopictus are found in Sri Lanka. Although Ae. albopictus was considered as a rural secondary dengue vector in the past has now invaded urban areas as well and has become the most dominant Aedes species in most of the areas in the country [20]. However, so far none of the research in the world has focused on determining the age structure of Ae. albopictus populations.
Hence, the proposed work aimed to construct multivariate calibration models using the transcriptional abundance of three age responsive genes; Ae15848, Ae8505, and Ae4274 to determine the age structure of Ae. aegypti and Ae. albopictus female mosquito strains from Sri Lanka and to investigate the in uence of temperature on the expression levels of these age responsive genes of Ae. aegypti females.

Establishment of mosquito colonies and sample collection
Blood-fed mosquitoes collected from Sri Lanka were used to obtain eggs to establish initial colonies of Ae. aegypti and Ae. albopictus. Colonies were maintained at an insectary with ambient conditions of 27 ± 2 C temperature, 70 ± 10% relative humidity (RH), and a photoperiod of 12:12 (L:D) which is similar to the eld sites.
Another Ae. aegypti colony [using the eggs from the above Ae. aegypti colony at 27 C)] was established at 23 ± 2 C (70 ± 10% RH and of 12:12 photoperiod) to determine the effect of temperature on the transcriptional age grading technique. For all the colonies larvae were given similar conditions (similar larval densities and larval food to each tray).
RNA extraction and cDNA synthesis RNA was extracted from pools of 10 individual mosquitoes of a given age group (3 replicates per each age class) using the Arcturus® PicoPure RNA Isolation kits (Thermo Fisher Scienti c, Waltham, MA, US) following the manufacturers' protocol. All the steps were done while keeping the samples on ice and centrifugation steps were carried out at 4 0 C. These samples were treated with RNase-free DNase (Qiagen Hilden Germany) to remove the DNA contaminations from the RNA samples. Samples were stored at -80 0 C for molecular analysis.
cDNA was synthesized by reverse transcription using extracted RNA as the template. SuperScript III rst-Strand Synthesis System (Invitrogen) was used for cDNA synthesis following the manufacturers' protocol.
Validation and quanti cation of age responsive genes using qRT-PCR assay Three age responsive genes i.e. Ae. 15848 (a gene involved in calcium-binding), Ae. 8505 (a gene involved in producing a structural component of cuticle), Ae. 4274 (a gene involved in zzy cell cycle/cell physiology) and, a housekeeping gene Rsp17 were selected for qRT-PCR analysis according to previous studies [6]. Primer sequences given in Cook et al. [6] were used to amplify these three genes in Ae. aegypti. Primers for Ae. albopictus were designed using Primer 3 software (version 0.4.0) [22,23] ( Table 1). Table 1 Primer sequences and the product length of the candidate genes and housekeeping genes After validating the three age responsive genes and the housekeeping gene of both species, quantitative real-time polymerase chain reaction (qRT-PCR) assays were continued to quantify the expression levels of all these genes using the MX3005 qPCR system (Agilent Technologies). Each qRT-PCR reaction mixture (20 µl) was prepared using 1 µl of cDNA, 10 µl Brilliant green Ultra-fast SyBr Green qPCR master mix (Agilent), 0.6µl of primer (10 mM), 7.8 µl nuclease-free water. The thermal cycling conditions were conducted with denaturation at 95 0 C for 3 minutes followed by 40 cycles of 10 seconds at 95 0 C, 10 seconds at 60 0 C, and a last step 95 0 C for one minute, 55 0 C for 30 seconds, and, 95 for 30 seconds. Three biological and three technical replicates were performed for each age class and each gene.

Data analysis
Data were analyzed using MxPro qPCR software (Agilent Technologies) to obtain the Cycle Threshold (Ct) value which is the measure of the expression of genes. Fold Change (FC) of each age responsive gene at each time point, relative to the one-day old mosquitoes were calculated using the 2 −ΔΔCT method [24] incorporating the PCR e ciency. A calibration model, which explains the strength of the linear relationship between the expression of all the three candidate genes and mosquito age was constructed using the linear regression of the rst redundancy variate generated during the redundancy analysis. A nonparametric bootstrapping (1000 bootstraps) method was used to assess the sampling error [95% con dence intervals (CI)], to validate the constructed calibration models and to predict mosquito ages. The median point of the CI intervals was considered as the likely predicted ages for mosquitoes [6]. The residual value (the difference between the predicted age and the actual age) was used to assess the accuracy of each model. The expression data of one experimental design was cross-checked with other models eg. Ae. albopictus age was predicted using Ae. aegypti model and vice versa, to determine the possible use of one model for age prediction.

Results
Expression levels of age responsive genes of Ae. aegypti and Ae. albopictus at 27 0 C All three genes, Ae15848, Ae8505 and Ae4274 were differentially expressed in both species and the fold change (relative to the expression of one day old mosquitoes) obtained are presented in Fig. 1  Expression levels of the three age responsive genes; Ae15848, Ae8505 and, Ae4274 were determined at 23 0 C for ve age classes (1,5,9,13 and, 17 days) of female Ae. aegypti and the transcriptional pro les obtained are shown in Fig. 3. The expression patterns of all three genes at 23 0 C were similar to those of Ae. aegypti mosquitoes at 27 0 C, that is the expression of Ae15848 and Ae8505 decreased with the age of mosquito while Ae4274 showed an increase in its expression with the mosquito age. The expression of the Ae15848 gene was highest at all the ve-time points (log contrast value ranging from 0.073 ± 0.005 at day 29 to -0.071 ± 0.017 at day 1) compared to the other two genes. The transcriptional abundance of Ae8505 was highest in 29 day old mosquitoes (0.151 ± 0.009) and lowest in 1 day old mosquitoes (0.014 ± 0.038). The expression of Ae4274 increased with the age of mosquitoes at both temperatures (transcriptional abundance ranging from 0.138 ± 0.013 to 0.201 ± 0.008). Similar to 27 0 C results, the change in expression levels of the three genes with female mosquito age was signi cant at 23 0 C (Table 3). Ae15848 (23 0 C F = 42.44, p < 0.05) showed a more consistent change in its expression with age of mosquitoes compared to Ae8505 (23 0 C F = 11.55, p < 0.05) and Ae4274 (23 0 C F = 10.46, p < 0.05).
Further, the transcriptional abundance was greater for Ae. aegypti at 23 0 C compared to 27 0 C indicating a lower expression at 23 0 C (Fig. 3).

Multivariate calibration models and age predictions
Aedes aegypti and Ae. albopictus from 27 0 C Calibration models were generated separately for each species using multivariate canonical redundancy analysis of the transcriptional abundance of the three genes; Ae15848, Ae8505, and Ae4274. Figure 4a and 4b show the graphs drawn between the rst redundancy variate and the actual age of Ae. aegypti and Ae. albopictus females.
A nonparametric bootstrapping (1000 bootstraps) procedure was used to validate the model generated for each species and, to obtain the 95% con dence intervals (CI) and age predictions (Cook et al., 2007).
According to the calibration model results, the actual age of both Ae. aegypti (R 2 = 0.8108, p < .0001) and Ae. albopictus (R 2 = 0.8990, p < .0001) females showed a strong positive correlation with the rst redundancy variate. The graphs between the predicted age derived from the model and the actual age of Ae. aegypti and Ae. albopictus are shown in Figs. 4c and 4d (Fig. 4f).
Aedes aegypti from 27 0 C and 23 0 C Linear age prediction models were generated separately for female Ae. aegypti at 23 0 C and 27 0 C using the normalized expressions of the age responsive genes for 5 age classes, 1, 5, 9, 13 and 17 days. Similar to the calibration model for mosquitoes at 27 0 C the model developed at 23 0 C also showed a strong linear positive correlation with the age of female mosquitoes (23 0 C; R 2 = 0.9222, p < 0.0001) (Fig. 5a and 5b).
To test the effect of temperature on the age prediction accuracy, cross predictions were conducted. When the age of Ae. aegypti females reared in 27 0 C were cross-checked with the model generated for Ae.

Discussion
Transcriptional age grading is a molecular-based technique with high precision and accuracy that could be used to determine the chronological age of mosquitoes. However, the models developed using this approach should be validated and optimized for mosquito strains from different geographical areas and adjusted for variations of environmental factors, especially temperature prevail in such areas [6]. Hence, the present study was conducted to develop a multivariate calibration model for Ae. aegypti and Ae. albopictus from Sri Lanka using transcriptional age grading technique that could be later used in estimating the age structure of wild mosquito populations. Further, the study attempted to understand the effect of temperature on this age grading technique.
In Sri Lanka, the highest number of dengue incidences and deaths every year is reported from the dry zone where the annual temperature is 27.5 0 C covers a major part of the island and this is the area that reports. Further, the entire country experiences relatively high humidity averaging around 80% throughout the year [25]. Hence, the mosquitoes used in the present study were maintained in insectaries with 27 0 C ± 2 and 80% ± 10 relative humidity simulating the eld conditions. Transcriptional age grading studies on Ae. aegypti [2,4] and An. gambiae [9] have clearly shown that the models constructed for mosquitoes reared in laboratory or semi-eld conditions could be successfully used to assess the age of wild mosquito populations. Therefore, the multivariate calibration models that were constructed during this study will be a useful tool in predicting the age of wild populations of Ae. aegypti and Ae. albopictus in Sri Lanka.
For both Sri Lankan Ae. aegypti and Ae. albopictus the pattern of gene expression was similar to the previous reports i.e. reduction in the expression of Ae15848 and Ae8505 genes and an increase in Ae4274 gene expression, with the age of the female mosquito [3,4,6,8,11]. In all the experimental designs, the change in expression of these three genes was signi cant with the age of female mosquito, further proving the fact that these are among the most informative age responsive genes that could be successfully used in the transcriptional age grading approach.
The expression levels of all three genes were high in both Ae. aegypti and Ae. albopictus strains from Sri Lanka than the mosquito strains from other countries. Further, during the current study, the expression of Ae15848 was signi cantly higher than the other two genes and showed a consistent change with the mosquito age. This gene has shown a similar strong negative correlation with the chronological age of female Ae. aegypti [2,4] and An. gambiae [9] mosquitoes from other parts of the world. Aedes aegypti has shown a four-fold increase in the log contrast of the same gene from 1 to 29 day old mosquitoes in a study carried out in Northern Australia [4], and around 1.5 fold increase in mosquitoes from Central Vietnam [11] and Queensland, Australia [3]. Gene, Ae8505 of both species showed a rapid decrease in gene expression from day 1 to day 5 and more consistent change thereafter, similar to previous reports.
The expression change of gene Ae4274 with the age of mosquitoes was signi cant although it vaires along with a small range as observed for Ae. aegypti females from Australia and Vietnam [3,4,11]. Previous research work has reported a 4 0 C temperature difference as the minimum range that could have a signi cant effect on the survival of both mosquito larvae and adult [26,27,28,29]. Hence, the model developed for Ae. aegypti colonies maintained at 27 0 C was compared with mosquitoes reared at 23 0 C (4 0 C difference) to check the effect of temperature on this molecular based age grading approach.
According to the results, the expression of the three age responsive genes for mosquitoes from 23 0 C was lower than mosquitoes from 27 0 C. Around 92% of the gene expressions were age related at 23 0 C while it is 81.01% at 27 0 C. According to the analysis conducted using nonparametric bootstrap, Studies will be conducted in the future to determine the age of wild mosquitoes using the multivariate calibration models generated during this study.

Conclusion
This is the rst report from Sri Lanka on the use of the transcriptional age grading technique to determine the age of mosquitoes and the rst study to develop a model for Aedes albopictus. The species-speci c multivariate calibration models created using the age responsive genes, Ae15858, Ae8505, and Ae4274 could successfully estimate the chronological age of wild Ae. aegypti and Ae. albopictus with higher accuracy than previously reported in other mosquito species. The age of mosquitoes could be predicted with an accuracy of nearly ± 3 days of their actual age, across the age spectrum. A drop in temperature from 27 0 C to 23 0 C did not have a very strong effect on the multivariate calibration models.   Logcontrast values obtained for the three age responsive genes Ae15848, Ae8505 and Ae4274 at each time point of Ae. aegypti and Ae. albopictus female mosquitoes reared in 27°C temperature.

List Of Abbreviations
Page 18/21 Logcontrast values obtained for the three age responsive genes, Ae15848, Ae8505 and Ae4274 at ve age classes (1,5,9,13 and 17 days) of Ae. aegypti females reared at 27°C and 23°C. Calibration models generated for a) Ae. aegypti and b) Ae. albopictus female mosquitoes using the transcriptional pro les of the three age responsive genes; Ae15848, Ae8505 and Ae4274 c); Age predictions of c) Ae. aegypti d) Ae. albopictus female mosquitoes at eight age class; Cross validation, e) Ae. albopictus data checked on Ae. aegypti model f) Ae. aegypti data checked on Ae. albopictus model (The dash lines indicate where predicted age equals actual age). Calibration models generated for female Ae. aegypti mosquitoes reared at a) 230C and b) 270C using the transcriptional pro les of the three age responsive genes; Ae15848, Ae8505 and Ae4274 c); Age predictions of female mosquitoes Ae. aegypti reared at c) 23°C d) 27°C at ve age class; Cross validation, e) 27°C data checked on 23°C model f) 23°C data checked on 27°C model (The dash lines indicate where predicted age equals actual age).