Use of transcriptional age grading technique to determine the chronological age of Sri Lankan Aedes aegypti and Aedes albopictus females

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 dengue virus. One of the most important parameters influencing the likelihood of arbovirus transmission is the age structure of the mosquito population. However, mosquito age is difficult 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 (fizzy cell cycle/cell division cycle 20). Methods The transcriptional age-grading technique was applied to determine the chronological age of Ae. aegypti and Ae. albopictus female mosquito populations from Sri Lanka using the age-responsive genes Ae15848, Ae8505, and Ae4274. Furthermore, Ae. aegypti samples obtained from colonies reared at two temperatures (23 and 27 °C) were used to investigate the influence of temperature on this age-grading technique. Expression levels of these three genes were quantified using reverse transcription qualitative PCR (qRT-PCR), and results were normalized against the housekeeping gene ribosomal gene S17 (RpS17). Results The expression of Ae15848 and Ae8505 decreased with the age of mosquitoes and showed the most significant and consistent change while expression of Ae4274 increased with age. The multivariate calibration models showed > 80% correlation between expression of these age-responsive genes and the age of female mosquitoes at both temperatures. At 27 °C the accuracy of age predictions using the models was 2.19 (± 1.66) days and 2.58 (± 2.06) days for Ae. aegypti and Ae. albopictus females, respectively. The accuracy of the model for Ae. aegypti at 23 °C was 3.42 (± 2.74) days. Conclusions An adult rearing temperature difference of 4 °C (23–27 °C) did not significantly 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 mosquitoes from Sri Lanka. Graphical Abstract

Background Aedes aegypti and Ae. albopictus are important vectors of human diseases such as dengue, chikungunya, and zika. Among these diseases, dengue is the most rapidly spreading mosquito-borne viral disease and is endemic in > 120 countries worldwide, including Sri Lanka. Among several factors, the age structure of the dengue vector population has been identified as one of the most important parameters influencing the epidemiology of the disease.
Based on experimental observations of the average extrinsic incubation period (EIP), the incubation period of the dengue virus within the mosquito's body before it is transmitted to the human body is 8-12 days [1][2][3][4]. Hence, knowing the age structure of the wild populations in an area is vital in decision making, especially in vector control programmes.
As there is no specific antiviral vaccine against dengue, efficient and environmentally friendly vector control measures are needed to prevent disease outbreaks. Introduction of artificially infected Ae. aegypti females with wMel strain of Wolbachia into the field has been one of the most promising dengue control interventions in the world due to failures of traditional vector control measures. This wMel strain of Wolbachia can block the dengue virus inside the mosquito body. This environmentally friendly method of vector control has been successfully practiced in several areas in the world by introducing Ae. aegypti infected with Wolbachia (wMel strain) [5]. Knowledge of the age of the vector population is important as the age affects the fitness of the released mosquitoes in the natural environment.
Morphological, biochemical, and molecular-based age-grading techniques have been developed to determine the age of female mosquitoes. However, the use of both morphological and some biochemical methods has become questionable because of their inability and inaccuracy in measuring the age of mosquitoes older than the EIP. The transcriptional age-grading technique has now been identified as the most accurate and precise approach in determining the chronological age of mosquitoes [3,6,7].
Quantification of expression levels of genes that show variation in their expression with the age of the female mosquito is the basis of this technique. The transcription scores of these age-responsive genes of laboratoryreared mosquitoes of known ages are then fed into a multivariate calibration model, which can later be used in age predictions of field/wild individuals. The mosquitoes used in the construction of the calibration model must be from the same mosquito strain of the field 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 [2-4, 11, 12] older than 15 days, which is more than the EIP period. Trials, using mosquitoes reared in field cages, have concluded that the gene expression profiles of Ae. aegypti female mosquitoes could determine the age with an accuracy of ± 5 days of the actual age [2,11,13].
The orthologues of the eight age-responsive genes, i.e. Ae4274, Ae4679, Ae4916, Ae6639, Ae7471, Ae8505, Ae12750, and Ae15848 selected from Drosophila melanogaster, were initially used to predict the age of female Ae. aegypti mosquitoes under both laboratory and field conditions [6,[14][15][16]. According to mosquito transcriptional age-grading studies CG8505/Ae-8505/AAEL003259 (pupal cuticle protein 78E putative) and SCP-1/Ae.-15,848/AAEL008844 (calcium-binding protein, putative) displayed the largest and most significant decrease in expression levels with the age of female mosquitoes while expression levels of fizzy/Ae-4274/AAEL014025 (fizzy cell cycle/cell physiology, putative) significantly increased with age. Hence, these three genes, Ae8505, Ae15848, and Ae4274, have been identified as the most informative age-responsive genes that can be used in the transcriptional profiling of mosquitoes [2,11,13]. Cook et al. [6] have identified Ae8505, Ae15848, and Ae4274 genes as the most reliable age-responsive genes and recommended using these three genes for future age determination studies. The gene RpS17 (40S ribosomal protein s17), which shows insignificant variation with age, has been exclusively used as the reference gene for normalizing the samples in these studies. Furthermore, the expression of these genes is not affected by 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 age-responsive genes within these mosquito populations may have sequence polymorphisms that may affect the application of gene expression analysis. Therefore, Cook et al. [6] suggested creating separate models for mosquitoes in different geographical regions. Furthermore, fluctuations in environmental parameters, such as temperature, were flagged as being potentially important to the successful application of the technique as these may affect the transcription of age-responsive genes [6]. Hence, models constructed considering all or most 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 death in Sri Lanka. The largest dengue outbreak was recorded in 2017 (≈186,000 cases) in the country, which is a ≈4.3 times increase in the average number of cases that were 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 because of the development of insecticide resistance by both dengue vectors [19,20]. Hence, the Sri Lankan government has planned to release wMel-infected dengue vectors into the field [22]. 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. Multivariate calibration models constructed so far for Ae. aegypti have used gene transcription measures of mosquitoes collected from geographical locations where the environmental and climatic conditions are different from those of Sri Lanka. Therefore, it is essential to validate the model for the Sri Lankan mosquito population.
Although Ae. albopictus was considered a rural secondary dengue vector in the past, it has now invaded urban areas as well and has become the dominant Aedes species across most of the country [20]. There has been relatively little research into age grading of Ae. albopictus [21], and none is based on the transcriptional age-grading technique. Furthermore, previous studies have focused only on one species where this study focuses on two vector species belonging to the same subgenus (Stegomyia) from the same geographical location. This will allow us to study the suitability of transcriptional age-grading across different species within this medically important subgenus.
Hence, the current study 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 the Ae. aegypti and Ae. albopictus female mosquito population from Sri Lanka and to investigate the influence 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 Kandy district, 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); these are similar to the field site conditions. 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 12:12 photoperiod) to determine the effect of temperature on the transcriptional age-grading technique. The larvae in all colonies were given similar conditions (similar larval densities and larval food to each tray).

RNA extraction and cDNA synthesis
RNA was extracted from pools of ten individual mosquitoes of a given age group (3 replicates per each age class) using the Arcturus ® PicoPure RNA Isolation kit (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer's protocol. RNA was eluted in 30 µl of elution buffer provided with the isolation kit. All the steps were carried out while keeping the samples on ice, and centrifugation steps were carried out at 4 °C. After adding the wash buffer, the RNA purification columns were treated with RNase-free DNase (Qiagen Hilden Germany) and incubated for 15 min at room temperature to remove DNA contamination from the RNA samples. Samples were stored at − 80 °C for molecular analysis.
cDNA was synthesized by reverse transcription using 1 ng extracted RNA as the template. SuperScript III First-Strand Synthesis System (Invitrogen) was used for cDNA synthesis following the manufacturer's protocol. One microlitre of Oligo (dt) 20 (50 µM), 3 µl of DEPC-treated water, and 1 µl of 10 mM dNTP mix were added to the sample and the mixture was incubated for 5 min at 65 °C. Four microlitres of 5× first-strand buffer, 1 µl of 0.1 M DDT, 1 µl of RNaseOUT, and 1.5 µl of Superscript III RT were added to the initial reaction. Sample was then incubated at 25 °C for 5 min, followed by 50 °C for 60 min and 70 °C for 15 min. The c-DNA product was treated with 1 µl Rnase H and incubated at 37 °C for 20 min to remove RNA.

Validation and quantification of age-responsive genes using qRT-PCR assay
Three age-responsive genes, 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 fizzy cell cycle/cell physiology), and a housekeeping gene, RpS17, 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) [23,24] (Table 1).
After validating the three age-responsive genes and the housekeeping gene of both species, quantitative real-time polymerase chain reaction (qRT-PCR) assays were used 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 cDNA, 10 µl Brilliant green Ultra-fast SyBr Green qPCR master mix (Agilent), 0.6 µl of primer (10 mM), and 7.8 µl nuclease-free water. The thermal cycling conditions were conducted with denaturation at 95 °C for 3 min followed by 40 cycles of 10 s at 95 °C, 10 s at 60 °C, and a last step of 95 °C for 1 min, 55 °C for 30 s, and 95 °C for 30 s. 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 1-day-old mosquitoes, was calculated using the 2 −ΔΔCT method [25] incorporating the PCR efficiency. Gene expression value/Ct value of each candidate gene at each time point was normalized to the reference gene (RpS17) by calculating the log contrast values using the equation given below [6].
Where, X i = mean Ct value of gene X. X ref = mean Ct of the reference gene RpS17. X total = sum of the Ct values for all genes from an individual.
One-way ANOVA was conducted (using the transcriptional abundance data/log contrast values of each experiment) to determine any significant variation in the expression of the gene with the age of mosquitoes using MiniTab 15.
A multivariate calibration method described by Cook et al. [3,6] was used to predict mosquito age. Canonical redundancy analysis was conducted for each experimental design using the normalized gene expression values (log contrast) of all three candidate genes using the SAS statistical software (SAS University edition). Syntax previously written by Cook et al. [6] was modified and used in all the analyses. A calibration model, which explains the strength of the linear relationship among the expression of all three candidate genes and mosquito age, was constructed using the linear regression of the first redundancy variate generated during the redundancy analysis. A nonparametric bootstrapping (1000 bootstraps) method was used to assess the sampling error [95% confidence intervals (CI)] to validate the constructed calibration models and to predict mosquito pool ages. The median point of the CI intervals was considered as the likely predicted ages of mosquitoes [6]. The residual value (difference between the predicted age and actual age) was used to assess the accuracy of each model. The expression data of one experimental design were cross-checked with other models, e.g. Ae. albopictus age was predicted using the Ae. aegypti model and vice versa, to determine the possible use of one model for age prediction.

Expression levels of age-responsive genes of Ae. aegypti and Ae. albopictus at 27 °C
All three genes, Ae15848, Ae8505, and Ae4274, were differentially expressed in both species and the fold changes relative to the expression of 1-day-old mosquitoes obtained are presented in Fig. 1. Underexpression of both Ae15848 and Ae8505 in all the age groups compared to day 1 and a strong drop in expression were observed from day 1 to day 5 for both species. The fold changes of genes Ae15848 and Ae8505 of 5-day-old mosquitoes compared to day 1 of Ae. aegypti were respectively 0.2 and 0.01. For Ae. albopictus day-5 FC values relative to day 1 for Ae15848 and Ae8505 respectively were 0.10 and 0.03. For Ae. aegypti, the FC of gene Ae4274 was Table 1 Primer sequences and product length of the candidate genes and housekeeping genes  The expression of Ae4274 increased with the age of mosquitoes unlike the other two genes. For both mosquito species the lowest expression of Ae4274 was at day 1 (0.138 ± 0.025 and 0.143 ± 0.013 respectively for Ae. aegypti and Ae. albopictus). The highest expression was at 21-day-old age class, and the log contrast values were 0.087 ± 0.012 for Ae. aegypti and 0.89 ± 0.007 for Ae. albopictus. The expression levels of this gene were lower compared to Ae15848 and Ae8505 for almost all time points tested.
According to one-way ANOVA test results, all three genes showed a significant change in expression with the age of female mosquitoes (Table 2). Aedes aegypti showed a significant and more consistent change in expression of genes Ae15848 (F = 12.34, p = 0.0001) and Ae8505 (F = 19.62, p = 0.0001) than Ae4274 (F = 8.50, p = 0.001). and Ae4274 F = 9.12, p = 0.001). Analysis using twoway ANOVA showed that the expression of Ae15848 (F = 34.99, p = 0.0001) and Ae8505 (F = 36.69, p = 0.0001) was significantly greater in Ae. aegypti than that for Ae. albopictus while Ae4274 did not show any significant difference in its expression between these two species (F = 0.37, p = 0.547).

Expression levels of age-responsive genes of Ae. aegypti at 23 °C
Expression levels of the three age-responsive genes, Ae15848, Ae8505, and, Ae4274, were determined at 23 °C for five age classes (1,5,9,13, and 17 days) of female Ae. aegypti and the transcriptional profiles obtained are shown in Fig. 3. The expression patterns of all three genes at 23 °C were similar to those of Ae. aegypti mosquitoes at 27 °C, that is, the expression of Ae15848 and Ae8505 decreased with mosquito age while Ae4274 showed an increase in its expression with the mosquito age. The expression of the Ae15848 gene was highest at all five 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 mosquitoe age at both temperatures (log contrast values ranging from 0.138 ± 0.013 to 0.201 ± 0.008). Similar to 27 °C results, the change in expression levels of the three genes with female mosquito age was significant at 23 °C (Table 2). Ae15848 (23 °C F = 42.44, p = 0.0001) showed a more consistent change in its expression with mosquitoe age compared to Ae8505 (23 °C F = 11.55, p = 0.0001) and Ae4274 (23 °C F = 10.46, p = 0.001). Furthermore, the transcriptional abundance was greater for Ae. aegypti at 23 °C compared to 27 °C indicating a lower expression at 23 °C (Fig. 3).

Multivariate calibration models and age predictions Aedes aegypti and Ae. albopictus from 27 °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 b shows the graphs drawn between the first 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% confidence intervals (CI) and age predictions for the mosquito pools [6]. According to the calibration model results, the actual  Fig. 3 Log contrast values obtained for the three age-responsive genes, Ae15848, Ae8505, and Ae4274, at five age classes (1,5,9,13, and 17 days) of Ae. aegypti females reared at 27 and 23 °C age of both Ae. aegypti (R 2 = 0.8108, p = 0.0001) and Ae. albopictus (R 2 = 0.8990, p = 0.0001) females from the tested mosquito pools showed a strong positive correlation with the first redundancy variates. The graphs between the predicted ages derived from the model and the actual age of Ae. aegypti and Ae. albopictus are shown in Fig. 4c (Fig. 4f ).

Aedes aegypti from 27 °C and 23 °C
Linear age prediction models were generated separately for female Ae. aegypti at 23 °C and 27 °C using the normalized expressions of the age-responsive genes for five age classes, 1, 5, 9, 13, and 17 days. Similar to the calibration model for mosquitoes at 27 °C, the model developed at 23 °C also showed a strong positive linear correlation with the age of female mosquitoes (23 °C; R 2 = 0.9222, p = 0.0001) ( Fig. 5a and b). According to the nonparametric bootstrap method, the residual value for Ae. aegypti at 27 °C was 2.19 (± 1.66) days and at 23 °C was 3.42 (± 2.74) days ( Fig. 5c and d).
To test the effect of temperature on the age prediction accuracy, cross predictions were conducted. When the age of Ae. aegypti females reared at 27 °C were crosspredicted with the model generated for Ae. aegypti females at 23 °C, the age of Ae. aegypti was overestimated [residual value 3.93 (± 3.00) days] (Fig. 5e). The 27 °C model under-estimated (mainly the younger age classes showed the highest deviation) the age of Ae. aegypti females from 23 °C [residual value of 5.82 (± 3.57)] (Fig. 5f ).

Discussion
Several chemical and molecular methods have been developed to measure the age of insect vectors such as measuring pteridine concentration [26], detecting the proportions of different cuticular hydrocarbons [27], protein expression profiling [28], and near-infrared spectroscopy [28,29]. Among these, transcriptional age grading is a molecular-based technique with high precision and accuracy that can be used to determine the chronological age of mosquitoes. However, the models developed using this approach should be validated and optimized for mosquito populations from different geographical areas and adjusted for variations of environmental factors; especially temperature prevails 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, which could later be used in estimating the age structure of wild mosquito populations. Furthermore, the study attempted to understand the effect of temperature on this agegrading technique. This study basically followed the protocol described by Cook et al. [6] with some changes in the RNA extraction and qRT-PCR technique. Also RNA was extracted from pools of mosquitoes, similar to several other studies [2,3,10]. Therefore, the transcriptional abundance represented by the graphs represents broad trends among a pool of mosquito.
In Sri Lanka, the highest number of yearly dengue incidences and deaths is reported from the dry zone where the annual temperature is 27.5 °C and covers a major part of the island. Furthermore, the entire country experiences relatively high humidity averaging around 80% throughout the year [30]. Hence, the mosquitoes used in the present study were maintained in insectaries with 27 ± 2 °C and 80% ± 10 relative humidity, simulating the field 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 semifield conditions could be successfully used to assess the age of wild mosquito populations. Therefore, the multivariate calibration models constructed during this study will be a useful tool for 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 those of 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 significantly related to the age of female mosquitoes, further proving that these are among the most informative age-responsive genes that can be successfully used in the transcriptional age-grading approach.
The expression levels of all three genes were higher in both Ae. aegypti and Ae. albopictus from Sri Lanka than in mosquito populations from other countries. However, it cannot be confirmed that these results are solely due to population differences since there are some experimental protocol differences in the present study compared to that of the previous work. Furthermore, during the current study, the expression of Ae15848 was significantly higher than in the other two genes and showed a consistent change with 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 significant although it vaired along a small range as observed for Ae. aegypti females from Australia and Vietnam [3,4,11].
Unlike the calibration models developed previously using individual mosquitoes of Ae. aegypti by Cook et al. [3] (R 2 = 0.73) and Hugo et al. [4] (R 2 = 0.72), the two multivariate calibration models developed for both species during the present study had a strong correlation to the age of the mosquito (R 2 for Ae. aegypti = 0.8108, p = 0.0001 and R 2 for Ae. albopictus = 0.8990, p = 0.0001). Furthermore, in this study the accuracy of the models was greater for both species, 2.19 (± 1.66) days for Ae. aegypti and 2.58 (± 2.06) days for Ae. albopictus. These facts indicate the high precision and accuracy of the age prediction models generated during the current study. Hence, these multivariate calibration models generated from this study will be refined using data collected from mosquitoes maintained under a more natural environment. The refined models will be tested using field-collected mosquitoes. Furthermore, these findings provide strong evidence that the transcriptional age grading must be validated and optimized for mosquito strains from different geographical locations as stated by Cook et al. [6].
The age of Ae. albopictus was overestimated when the data were cross-predicted with the Ae. aegypti model and the age prediction accuracy decreases significantly from 2.58 (± 2.06) days to 6.77 (± 1.41) days. Ae. aegypti age will be underestimated and age prediction accuracy decreases from 2.19 (± 1.66) to 7.4 (± 3.41) days when cross-predicted with Ae. albopictus model. This analysis strongly supports the species-specific nature of the multivariate calibration models (although of the same genus) generated using the transcriptional age-grading technique.
Previous research work has reported a 4 °C temperature difference as the minimum range that could have a significant effect on the survival of both mosquito larvae and adults [31][32][33][34]. Hence, the model developed for Ae. aegypti colonies maintained at 27 °C was compared with mosquitoes reared at 23 °C (4 °C difference) to check the effect of temperature on this molecular-based agegrading approach. According to the results, the expression of the three age-responsive genes for mosquitoes from 23 °C was lower than for mosquitoes from 27 °C. Around 92% of the gene expressions were age related at 23 °C while it is 81.01% at 27 °C. According to the analysis conducted using nonparametric bootstrapping, the accuracy of age predictions did not show a considerable difference between the two temperatures (residual value 2.19 ± 1.66 days at 27 °C and 3.42 ± 2.74 days at 23 °C).
According to the results of the cross-validation, the age prediction accuracy of mosquitoes from 27 °C decreased slightly from 2.19 (± 1.66) days to 3.93 (± 3.00) days and from 3.42 (± 2.74) to 5.82 (± 3.57) days for mosquitoes reared in 23 °C. It could be stated that a temperature difference of 4 °C affects expression of age-responsive genes to a lesser extent than the difference exists between two species under the same temperature. Similar observations have been reported by Hugo et al. [4] and have suggested generating separate calibration models for inter-seasonal variations for different regions that show a greater variation in climatic factors. Sri Lanka is a small island and does not show drastic changes in temperature between areas within the country. Hence, the models developed during this study could be used to determine the age structure of wild mosquito populations from the whole country. Depending on the average temperature at the time of investigation the most appropriate model to be used can be determined.
Studies will be conducted in the future to determine the age of wild mosquitoes using the multivariate calibration models generated during this study after validating the models with mosquitoes reared under semi-field/field conditions.

Conclusion
This is the first report from Sri Lanka on the use of the transcriptional age-grading technique to determine the age of mosquitoes and the first study to develop a model for Aedes albopictus. The species-specific 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 to 23 °C did not have a very strong effect on the multivariate calibration models.