Mosquito host-seeking diel rhythm and chemosensory gene expression is affected by age and Plasmodium stages

Malaria parasites can affect vector-related behaviours, increasing transmission success. Using Anopheles gambiae and Plasmodium falciparum, we consider the effect of interaction between infection stage and vector age on diel locomotion in response to human odour and the expression of antennal chemosensory genes. We identified age-dependent behavioural diel compartmentalisation by uninfected females post-blood meal. Infection disrupts overall and diel activity patterns compared with age-matched controls. In this study, mosquitoes carrying transmissible sporozoites were more active, shifting activity periods which corresponded with human host availability, in response to human odour. Older, uninfected, blood-fed females displayed reduced activity during their peak host-seeking period in response to human odour. Age- and infection stage-specific changes in odour-mediated locomotion coincide with altered transcript abundance of select chemosensory genes suggesting a possible molecular mechanism regulating the behaviour. We hypothesize that vector-related behaviours of female mosquitoes are altered by infection stage and further modulated by the age post-blood meal of the vector. Findings may have important implications for malaria transmission and disease dynamics.

(7 dpi), but not able to transmit the pathogen, whereas salivary gland sporozoite-infectious mosquitoes are more active (14 dpi) [10][11][12] . This modulation appears to be conserved across vectors and other pathogens, e.g., Aedes aegypti, which transmit dengue and Zika virus 5,13 . The observed changes in the odour-mediated host-seeking behaviour, demonstrated for both midgut oocyst-infected and salivary gland sporozoite-infectious mosquitoes, are reflected in the selective sensitivity of the peripheral olfactory system to key host-related volatile organic compounds 5,10,14 . Such modulation may be regulated through alterations in neural signalling and chemosensory gene expression, as demonstrated in malaria-and dengue-infected mosquitoes 2,5 .
In depth analysis of the molecular basis of olfaction in mosquitoes, has allowed for the identification and functional characterisation of chemosensory gene families, including odorant receptors (Ors), gustatory receptors (Grs), ionotropic receptors (Irs), odorant binding proteins (Obps) and chemosensory proteins (Csps) [15][16][17][18][19] . The expression of select members of these gene families in olfactory-related tissues change in response to adult maturation, age and physiological state [20][21][22][23][24][25] , including infection status 5 . These state-dependent changes in gene abundance coincide with concerted alterations in the sensitivity of olfactory sensory neurones and the behaviour of the mosquitoes 20,21,25 .
This study analysed the stage-dependent effect of P. falciparum infection on the locomotion activity of female Anopheles gambiae in the presence and absence of human odour, demonstrating that midgut oocyst-infected and salivary gland sporozoite-infectious mosquitoes display differential locomotor activity correlating with age post-blood meal and time of day. The potential molecular mechanisms, underlying the age-and infectiondependent odour-mediated locomotor activity, were assessed by transcriptome analysis, identifying differentially abundant chemosensory-related genes in the antennae coinciding with the demonstrated differences in mosquito behaviour. A better understanding of the effect of infection on locomotor activity and odour-mediated host seeking throughout the day has the potential to more accurately direct vector measures. For example, if we know when salivary gland sporozoite-infectious mosquitoes are active and host seeking, especially as these time periods differ from those of uninfected females, we can develop strategies to increase the probability of protecting people from these mosquitoes posing the greatest risk to human health, such as informing the public to take special precautions (e.g., remaining under LLINs, wearing protective clothing) during these times. Identifying the molecular mechanisms by which these behavioural traits are modulated by P. falciparum may lead to a better understanding of vectorial capacity and transmission dynamics potentially inspiring the development of innovative control strategies.

Results
Mosquito age and the stage of Plasmodium falciparum infection affect locomotor activity. Locomotor  In the presence of human odour, the age-dependent reduction in locomotor activity was observed in both uninfected (χ 2 1 = 21.12, p < 0.001) and P. falciparum carrying mosquitoes (χ 2 1 = 5.73, p = 0.02; the GLMM:lmer model was constrained by age and the random effect of experimental replication; Fig. 1b; Supplementary Figs. 1b and 2b). The infection load increased with age post-blood meal and was consistent across replicates ( Supplementary Fig. 2).
Infection status significantly affected the activity of mosquitoes, when comparing the overall locomotion profiles of 7 dpi and 14 dpi individuals with age-matched uninfected females, in the presence and absence of human odour (β-estimated = weighted mean value based on all variables plus the random effect of the replication ± SE; presence of human odour: χ 2 1 = 18.06, p < 0.001; absence of human odour: χ 2 1 = 5.73, p < 0.01). In the absence of human odour, 7 dpi females were less active compared to age-matched controls (χ 2 1 = 8.21, p = 0.004; Supplementary Fig. 1), whereas 14 dpi females were as active as their age-matched controls (χ 2 1 = 2.27, p = 0.13; Fig. 1c; Supplementary Fig. 1c). A continuous temporal activity analysis, which provides a sliding moving average of a single variable, i.e., locomotion (Supplementary Fig. 1), supported the overall GLMM statistical model, which combines multiple variables, as well as their interaction (Fig. 1), and the overall finding of this experiment. However, the continuous temporal activity analysis also identified a discrepancy, in which 14 dc females appeared to be more active than 14 dpi females during scotophase ( Supplementary Fig. 1a,c). This emphasises the statistical power of using multi-variable models for studying complex ecological interactions. The presence of human odour differentially altered the locomotor activity. While there were no significant differences between 7 dpi and age-matched females in the presence of human odour (χ 2 1 = 0.12, p = 0.73; Fig. 1c; Supplementary Fig. 1b), the salivary gland sporozoite-infectious 14 dpi mosquitoes were more active than age-matched uninfected females (14 days control; 14 dc χ 2 1 = 8.80, p = 0.002; Fig. 1c; Supplementary Fig. 1d).
Individual locomotor profiles of mosquitoes at 7 dpi and 14 dpi differed significantly from that of agematched uninfected controls (Fig. 1d). Infection significantly reduced the locomotor activity in 7 dpi females in the absence of human odour when compared with their uninfected counterparts during most time periods (ZT 0-11: Antennal transcripts were significantly differentially regulated between the two age-matched cohorts (6187), of which 3 465 were differentially abundant only between 14 dpi and 14 dc, whereas in 7 dpi and 7 dc, 850 were differentially abundant (Fig. 2b). In total, 4807 transcripts were significantly differentially regulated between the two age-matched control groups and the two groups carrying P. falciparum parasites, of which 1071 transcripts were shared between them (Fig. 2c). Within the age-matched control groups, 2614 transcripts were uniquely regulated, whereas 1122 were uniquely regulated within the two groups carrying P. falciparum parasites ( Fig. 2c; Supplementary Fig. 3). Overall, differentially abundant transcripts were not condition-dependently regulated, except those regulated post-infection during both the midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) stages (1872), of which more than 80% of the transcripts were down-regulated post-infection compared with the antennal transcripts of the uninfected, age-matched, post-blood meal females (Fig. 2b).
To characterise the functional ontology of the differentially abundant genes in the antennae of midgut oocystinfected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) mosquitoes with their age-matched controls, a gene ontology (GO) analysis of molecular function (level three) was conducted (Fig. 3). Of the 3 685 differentially abundant transcripts between the two uninfected control groups, the majority (> 75%) were functionally classified as structural constituent of cuticle (GO: 0042302) and enzyme inhibitor activity (GO: 0004857; Fig. 3a). Both of these classes were more abundant in older compared to the younger individuals. None of the age-dependent GO terms identified in the uninfected cohort comparison (Fig. 3a) were detected in the pairwise comparisons of the antennal transcriptomes among midgut oocyst-infected (7 dpi), salivary gland sporozoite-infectious (14 dpi) and their age-matched controls. The three most represented functional classes in the pairwise comparisons between the midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) groups with their agematched controls, as well as between the midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious ( www.nature.com/scientificreports/ Principal component analysis (a) of the overall transcript abundance in the antennae of midgut oocystinfected (7 days post-infection, dpi) and salivary gland sporozoite-infectious (14 dpi) mosquitoes compared to age-matched uninfected mosquitoes (7 days control, dc and 14 dc). Plasmodium falciparum infection load is indicated (see Supplementary Fig. 1). Proportional Venn diagrams (b, c) depicting pairwise comparisons among the antennal transcriptomes of midgut oocyst-infected, salivary gland sporozoite-infectious, and age-matched uninfected females. Overlapping regions represent the subsets of transcripts that are shared between the different conditions. Significant differences in transcript abundance were determined as a fold change of greater than 1.5, and an FDR-corrected p-value < 0.05.  . Gene ontology classes that were included comprised greater than 1% of the total number of differentially abundant genes identified. Those with less than 1% are grouped together in "other". www.nature.com/scientificreports/ in both midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) samples, while those in the carbohydrate derivative binding class were differentially regulated only in infectious samples (Fig. 3b).

Scientific Reports
A detailed analysis of the major chemosensory gene families associated with the odorant binding functional class, Ors, Irs, Grs, Csps and Obps, was conducted. As the mosquitoes aged post-blood meal, 34% of the chemosensory genes were significantly regulated in the antennae of uninfected females, with chemosensory receptors demonstrating higher abundance in the older females, while the binding proteins were both up-and downregulated (Fig. 4). Following P. falciparum infection, the abundance of 18 Obps reduced with age post-blood meal, while only one chemosensory gene, Ir41a, demonstrated an increased abundance in the antennae of older females (Fig. 4). The differential abundance of these 18 Obps, along with three others, appear to be a result of an increased abundance at 7 dpi compared to the age-matched controls. The other chemosensory gene that was regulated at this age, Ir7u, was down-regulated upon infection.
The post-blood meal, age-dependent regulation of chemosensory genes was affected following P. falciparum infection, as 73% of those genes that were up-regulated with age (Fig. 4, column 1), were shown not to increase in abundance post-infection (Fig. 4, columns 2 and 3), resulting in a higher abundance of these transcripts in the controls compared with those at 14 dpi (Fig. 4, column 4). The only exception to this was Ir7u, which was down-regulated at 7 dpi, exacerbating the decreased abundance observed at 14 dpi. Interestingly, when the 7 dpi in the antennae of midgut oocyst-infected (7 days post-infection; dpi) and salivary gland sporozoite-infectious (14 dpi) female mosquitoes, along with their age-matched controls (7 days control; 7dc and 14 dc) were analysed in pairwise comparisons displayed as log 2 fold change in abundance. Chemosensory genes are arranged into gene families, odorant binding proteins (OBPs), chemosensory proteins (CSPs), odorant receptors (Ors), ionotropic receptors (Irs) and gustatory receptors (Grs), in descending order of abundance (grey gradient bar, log 10 TPM). Significant differences in abundance (fold change > 2; FDR-corrected p < 0.05) are indicated by black dots. www.nature.com/scientificreports/ mosquitoes were compared to 14 dpi, the abundance of the above-mentioned Obps were reversed, returning to pre-infection levels (Fig. 4, column 2).

Discussion
The stage of infection of P. falciparum in An. gambiae has a significant impact on vector-related behaviour, which affects malaria transmission and disease dynamics 2,5,27,28 . This study confirms that midgut oocyst-infected (7 dpi) mosquitoes are less active than salivary gland sporozoite-infectious (14 dpi) females, and furthermore identifies this change in behaviour to be a result of an interaction between infection stage and age post-blood meal, which can be modulated by the presence of human odour and diel rhythms. We demonstrate that transcriptomic profiling of antennal tissue of P. falciparumoocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) An. gambiae identifies members of several chemosensory gene families as changing with age post-blood meal and infection status, suggesting a molecular mechanism, which may regulate the sensitivity of the peripheral olfactory system of the vector to human odour during the sporogonic cycle. As age and a blood meal are intrinsic to the infection status of the female, controlling for age post-blood meal provides a more accurate indication of which behaviours and molecular correlates are regulated by P. falciparum infection.
Host seeking in mosquitoes is dependent on age 20,21,23,29,30 and physiological state, including infection [3][4][5] . Our results demonstrated that the overall locomotor activity of uninfected females significantly reduced with age post-blood meal, in line with previous studies on mosquitoes 21,23 , further emphasising the role of age in regulating host seeking. As mosquitoes age, energy reserves decrease, resulting in older females reducing their search activities and becoming more selective to reduce their energy requirements 31,32 . We find that the presence of human odour did not alter the overall locomotor activity, similar to that described by Cator et al. 10,28 , suggesting that the difference in the level of activity with age is independent of the presence or absence of human odour. While previous studies on the effects of Plasmodium infection on mosquito host seeking have shown a reduced activity in midgut oocyst-infected (7 dpi) compared with salivary gland sporozoite-infectious (14 dpi) mosquitoes, these have not controlled for the age of the insects 10,11 . Our results indicate an interaction between age post-blood meal and infection in midgut oocyst-infected females (7 dpi), which display a similar locomotor activity as an age-matched cohort, is not seen in salivary gland sporozoite-infectious individuals (14 dpi). When controlling for age post-blood meal and observing the effect of infection, midgut oocyst-infected females reduced their overall locomotor activity compared to age-matched controls in the absence of human odour, while in the presence of human odour there was no discernible difference in activity. In contrast, salivary gland sporozoiteinfectious (14 dpi) mosquitoes increased their overall locomotor activity in the presence of human odour, but not in its absence, compared to age-matched controls. These findings provide an explanation for the discrepancies between past studies on the behaviour of oocyst-infected (7 dpi) and sporozoite-infectious females (14 dpi) 10,33 . The mechanisms underlying the changes in host-seeking and blood-feeding behaviours are still unclear, however several hypotheses have been advanced, chief among these is that the response in midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious females (14 dpi) is a result of a general immune response, and/or a direct manipulation of the vector behaviour by the pathogen 3,34-36 . The increased response to human odour, by salivary gland sporozoite-infectious mosquitoes (14 dpi), coincides with increased feeding frequency, probing rate and meal size 28,35,37,38 , and has substantial effects on malaria transmission ecology and overall disease dynamics 36 .
Diel locomotor activity in mosquitoes is regulated by the age post-blood meal and infection status of female malaria vectors 26,39,40 . In the presence and absence of human odour, younger uninfected mosquitoes (13 days post-emergence, dpe) were more active than the older mosquitoes (19 dpe) during photophase, scotophase and peaking at dawn 41 . We find that the peak of activity for older females, on the other hand, occurred at dusk. We note, however, that the continuous temporal analysis of the locomotor activity, in which other factors are not taken into consideration, revealed an increased activity of older females during scotophase in the absence of human odour. Post-blood meal, the age-dependent change in diel locomotor activity was not modulated by the presence of human odour, which is in line with the diel patterns of the peripheral olfactory system responsiveness [42][43][44] . The shift in peak activity of older females co-occurs with an increased propensity of older females to visit nectar sources at dusk 45 , indicating a behavioural compartmentalisation due to an age-related reduction in energy reserves 31,46 . Here, we show infection with P. falciparum altered the diel locomotor activity in both the presence and absence of human odour, generally decreasing activity in midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious females (14 dpi) in the absence of human odour, similar to that described in An. stephensi infected with Plasmodium yoelii 47 . In the presence of human odour, midgut oocystinfected (7 dpi) and salivary gland sporozoite-infectious females (14 dpi) demonstrated a disrupted activity pattern associated with host seeking. In particular, infectious females (14 dpi) shifted their odour-associated activity to scotophase and dawn, which is in line with observations of field-caught mosquitoes carrying sporozoites 39,48 . By selectively increasing their activity during times of low-risk of predation and high access to human hosts, the infectious females increase the probability of a successful blood meal and pathogen transmission [49][50][51][52][53] . While the infectivity of the bite is not affected by diel rhythms 54 , the altered host-related activity pattern in infectious females may have explicit implications on the most effective deployment of vector control tools, and thus our ability to affect malaria transmission dynamics efficiently 36 . Whether this is a result of an active manipulation by the pathogen or a general immune response remains to be clarified 3,[34][35][36]55,56 .
Our results demonstrate age post blood-meal and P. falciparum infection significantly affected overall antennal transcript abundance, with infection systemically down-regulating gene abundance, and differentially affecting the functional ontological class odorant binding (GO:0005549). Among the differentially abundant transcripts, antennal abundance generally increased with age across all of the major olfactory-related gene families, whereas only Obps, the most abundantly expressed transcripts of the chemosensory gene families, exhibited higher abundance in midgut oocysts-infected (7 dpi) compared with salivary gland sporozoites-infectious females (14 dpi). www.nature.com/scientificreports/ Odorant binding proteins are involved in binding and transporting odorants to the receptors and in the overall sensitivity of the peripheral olfactory system by providing gain control 22,57,58 . The Obps appear to play a major role in regulating the sensitivity and selectivity of the peripheral olfactory system of An. gambiae in response to ageing and infection status, as 67% of the Obp transcripts reliably expressed in the antennae were age-and infection-dependently regulated. As the females age, older individuals exhibited a higher transcript abundance in two thirds of the differentially expressed classic, and all of the plus-C classified Obps 17,57 . The other third of the differentially regulated classic Obps, along with all 13 of the two-domain Obps 17,57 , were expressed abundantly in 7 dpi antennae, compared with not being reliably expressed in their age-matched controls and salivary gland sporozoite-infectious females (14 dpi). Bimolecular protein interaction models indicate that the most energetically favourable ligands of the reliably expressed classic Obps, including those that were differentially abundant, are distinct sets of carboxylic acids, aromatics and terpenes 17 , known to regulate host and floral seeking 45,46 . While no clear pattern of ligand binding was apparent associated with ageing, those classic Obps increasing in abundance 7 dpi generally did not include carboxylic acids among their modelled ligands, likely resulting in a reduced sensitivity of the peripheral olfactory system to these salient human odorants by midgut oocyst-infected females (7 dpi) [59][60][61] . This corresponds to the reduction in human odour-associated activity exhibited by midgut oocyst-infected females (7 dpi). The functionally characterised differentially abundant olfactory receptors, Or11, Or39 and Ir41a, all respond to human odour constituents 16,21,62 . The receptor Or39 has been postulated to modulate the onset of host seeking in Anopheles coluzzii through the concerted downregulation of the receptor and the sensitivity to sulcatone of its cognate olfactory sensory neuron 21 . Whether the lower abundance of Or39, in conjunction with Or11, observed in the antennae of salivary gland sporozoite-infectious (14 dpi) compared with uninfected females, underlies the observed increased locomotor activity of salivary gland sporozoite-infectious females (14 dpi) in the presence of human odour, remains to be investigated. The only receptor to have a higher abundance in the antennae of salivary gland sporozoite-infectious (14 dpi) compared to midgut oocyst-infected females (7 dpi), Ir41a, responds to the human spermous cyclic amine, pyrrolidine 16,63 , suggesting that this receptor is involved in the increased attraction to human odour 14 dpi. The differential regulation of transcript abundance of genes modulating the chemosensory pathway in response to parasite development within the insect vector provides a mechanism by which the physiological 14 and behavioural sensitivity 4,10,11,33,35 to human odour can be modulated during the sporogonic cycle of P. falciparum.
Behavioural activity of An. gambiae and its expression of antennal chemosensory-related genes are altered in an age-and P. falciparum stage-dependent manner. Irrespective of the mechanism regulating these changes, be it active manipulation by the parasite and/or a general response to immune challenge, the differential behavioural response to human odour of females as they age or become infected post-blood meal significantly affects vectorial capacity and malaria transmission. Our ability to increase the resolution concerning which behavioural change is affected by age and/or infection, allows us to fine tune the epidemiological models for malaria transmission 33 , and integrate both factors and their interactions into the development and implementation of control measures.

Methods
Ethics. Human blood (type O) was provided in citrate-phosphate-dextrose-adenine anti-coagulant/preservative, and serum (type AB) was obtained from the blood transfusion service (blood bank) at the Karolinska University Hospital (Solna, Sweden), in accordance with the Declaration of Helsinki, and approved by the Ethical Review Board in Stockholm (2011/850-32). The human materials were obtained anonymously from the blood bank and did not require an informed consent statement. The Swedish Work Environment Authority (Stockholm, Sweden, SU FV-2.10.2-2905-13/31-01-2022) approved the class 3 biological agent laboratory and its practices, including insectary design and equipment to work with P. falciparum-infected mosquitoes. The regulations are mainly based on the EU directive 2000/54/EC on the protection of workers from risks related to exposure to biological agents at work.

Mosquito rearing and blood feeding.
Anopheles gambiae (Keele) were reared and maintained at 27 ± 1 °C, 80 ± 1% relative humidity and a photoperiod of 12 h light: 12 h dark cycle, 11 h full light of ~ 300 lx and 11 h darkness, separated by 1 h dawn and dusk transitions, respectively. The Keele colony was established at Keele University as a result of balanced interbreeding of four An. gambiae s.l. strains: G3 from MacCarthy Island, the Gambia; and three Tanzanian strains, ZAN U from Zanzibar, Ifakara strain from Njage, and KIL from Marangu 64 , and set up at Stockholm University in 2014. For experiments, pupae (ca. 350) were collected and transferred to cages (custom made: 30 cm × 20 cm × 20 cm) to emerge as adults, which were fed ad libitum on 5% glucose and 0.05% para-amino benzoic acid (PABA, Sigma-Aldrich, Steinheim, Germany). Females (5 dpe) were provided with a meal of cultured red blood cells (RBCs) through glass membrane-feeders connected to a custom-built heating system kept at 37 °C. The RBCs were washed with Roswell Park Memorial Institute (RPMI-1640) medium (Gibco, ThermoFisher Scientific, Drieich, Germany), and stored in RPMI-1640 at 50% haematocrit at 4 °C prior to use.
Parasite culture and mosquito infection. The strain of P. falciparum used in this study, NF 54 -SU, was initially isolated in the Netherlands and donated by Klavs Berzins. In vivo culturing of P. falciparum NF 54 -SU followed the standard protocol 65 . In short, P. falciparum NF 54 -SU was cultured using washed RBCs (5% haematocrit) from type-O blood donors, and 10% serum mixed with RPMI-1640 medium (Gibco™, ThermoFisher Scientific), supplemented with 5.94 gl −1 HEPES buffer (Sigma Aldrich), 2.1 g l −1 sodium hydrogen carbonate (Sigma Aldrich), and 0.05 g l −1 hypoxanthine (Sigma Aldrich www.nature.com/scientificreports/ parasitaemia, 6% haematocrit in RPMI-1640 medium. On the day of experimental infection, uninfected RBCs were supplemented with P. falciparum gametocytes to a final gametocytaemia of ca. 0.7%, which generates a high infection prevalence (> 50%) in An. gambiae s.s. 65 . For each infectious meal, mixtures of gametocytes from 14-and 17-day gametocyte cultures were used. Twenty-eight groups of ca. 100 mosquitoes were fed on mature gametocyte-infected RBCs or on uninfected washed RBCs using a membrane feeder. All age-matched control females (7-and 14-day control; 7 and 14 dc) were handled in the same manner as the infected and infectious cohorts, including access to mates, blood meals and oviposition substrates.
Mosquito odour-mediated locomotor activity. For recording the behavioural activity of the mosquitoes, DAM2 Drosophila activity monitors (TriKinetics Inc., Waltham, MA, USA) were used ( Fig. 1a; Supplementary Fig. 4). Two hours prior to the experiments, mosquitoes were placed individually in glass tubes (65 mm × 7 mm), using an aspirator, for pre-conditioning. Cotton plugs saturated with 5% glucose with 0.05% PABA (as above) were placed at one end of the glass tubes to provide ad libitum access to sustenance. Human odour was supplied by a previously worn cotton sock (ca. 24 h) placed inside a 200 ml glass wash bottle (Sigma Aldrich). The odour was introduced, via Teflon tubing, into the glass tubes through a MAN2 gas distribution manifold (TriKinetics Inc.). A clean unworn sock was similarly used as a control. Each experiment was replicated three times with midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) mosquitoes assessed in parallel with age-matched controls, which were blood fed at the same time as those which were infected, for a total of 384 mosquitoes assayed (96 per condition). Behavioural experiments were conducted in the class 3 biological agent laboratory (BSL3) within a climate chamber kept at 27 ± 1 °C, 70% relative humidity with a photoperiod of 11 h light: 11 h dark with 60 min of both dawn and dusk lighting conditions. Both assays were conducted over 24 h, with activity data recorded every 30 s. In this assay, the activity of female Anopheles gambiae, 7 days post-infection (dpi) and 14 dpi with P. falciparum, compared to age-matched uninfected females (days control, dc), in the presence (with human odour) or absence of human odour (w/o human odour) were assessed. Individual locomotor profiles (n = 384), assessed for variation using a redundancy analysis, demonstrated a differential effect of age post-blood meal ( Fig. 1a; Supplementary Fig. 4a,b). The influence of age, the presence and absence of human odour, and the effect of the sporogonic stages of P. falciparum on the An. gambiae (7 and 14 days post infection; 7 and 14 dpi) diel locomotor activity was analysed using general linear mixed models (GLMM), in which the effect of replication (experimental blocks) and the weighting of multiple replications (random variables) were taken into account. For all results, the significance of the best maximal model containing all explanatory effects was evaluated by using likelihood ratio test. All analyses were performed using R (R Core team 66  To assess the average overall locomotor activity of all the mosquitoes across the 24 h ( Supplementary  Fig. 1a-d) of recording, the Williams mean was used. This geometric mean (log (n + 1)) was calculated for all recordings every 30 min, and minimizes the influence of low/high values and inflated numbers of zeros on the data distribution. To compare the Williams means, a repeated measures two-way ANOVA with a Tukey's post hoc analysis was used to compare the overall locomotion activity every 30 min at 7 dpi and 14 dpi. The significance of the test is given by the Williams mean, which is simple moving average, separately for only one fixed explanatory variable at a time. This method does not allow for, or include the variation from, an extra factor, such as experimental replication (random variable in the case of biological and ecological studies). In contrast, GLMM statistical modelling (the output of a mixed model) takes into account the whole data set by including the effect of replication (i.e., experimental blocks, infection, age, diel activity pattern and host odour). Therefore, GLMM: mixed model (lmer) is an appropriate method for analysing the complex biological and ecological studies when the response variable, in this case locomotion, is affected by multiple factor interactions as it accounts for multiple replicates and explanatory variable interactions. In all analyses, all of the data met the assumptions of the test for normality and error homogeneity. Backward elimination was used for sequential removal of non-significant variables, to obtain the minimal statistically significant model. All graphical visualisations and statistical analyses were preformed using GraphPad Prism 68  Transcriptome analyses. Pairs of antennae from midgut oocyst-infected (7 dpi), salivary gland sporozoite-infectious (14 dpi) and age-matched control mosquitoes, all of which received a blood meal at the same age, were collected from individual females following flash freezing on dry ice for < 30 s, and placed into 24-well culture plates (ThermoFisher Scientific) in RNAlater (500 µl; Thermo Fisher Scientific). Individual carcasses (bodies missing antennae) of midgut oocyst-infected (7 dpi) and salivary gland sporozoite-infectious (14 dpi) mosquitoes were also collected for subsequent qPCR analysis, to determine the parasite abundance of individual females (see below). Following the quantification of the parasite load, the antennae of successfully infected individuals were pooled into tubes of 150 pairs of antennae, according to the level of infection, and stored at − 80 °C. Four replicates of each condition were collected. RNA libraries were constructed using a TruSeq RNA Library Prep Kit (Illumina, Berlin, Germany) from total RNA from the pooled antennae following extraction and DNase I digestion (RNeasy Mini kit, Qiagen, Hilden, Germany). Quality and quantity control of total RNA aliquots was performed using a Nanodrop spectrophotometer (Thermo Fisher Scientific), an Agilent 2100 Bio-analyser www.nature.com/scientificreports/ (Santa Clara, CA, USA) and a Qubit 2.0 Fluorometer (Thermo Fisher Scientific), before sending to BGI Genomics (MGI Tech Co., Ltd., China) for Illumina paired-end indexed sequencing, using the Illumina HiSeq2000.

RNA-Seq and differential expression analyses.
High-quality reads were determined using the following criteria: low quality reads were clipped from the start and end of each read using a sliding window and reads shorter than 40 nt were removed. The cleaned reads were mapped using CLC Genomics Workbench version 11 (https:// digit alins ights. qiagen. com) to the An. gambiae genome Agam4 with reference to the Agam4.10 gene set (www. vecto rbase. org). On average 94.11% of all reads mapped to the genome. Transcript abundance was reported as transcript per million (TPM) with a threshold level of abundance above 1 TPM. Differential transcript abundance was determined using the β-binomial general linear model algorithms in CLC Genomics Workbench 11. To control for false discovery rate (FDR), the Benjamini-Hochberg correction was applied 69 . This analysis generated weighted fold changes (FC) and FDR-corrected p-values that were used to detect differential expression. Significantly differential gene abundance was determined based on a FC ≥ 2 and an FDRcorrected p-value < 0.05.
Quantitative real-time polymerase chain reaction analysis. Quantitative real-time polymerase chain reaction (qPCR) was used to estimate the abundance of P. falciparum in individual mosquitoes at 7 and 14 days after being subjected to an infectious meal, and used for both behavioural and transcriptomic analyses, allowing quantification of the numbers of parasite genomes. In all analyses, a single body of an uninfected female mosquito was used as a negative control. Reactions were performed on a Roche Light Cycler using SYBR Green (Roche, Mannheim, Germany). DNA standards containing known numbers of P. falciparum parasites were produced from asexual cultures, and used to generate standard curves of P. falciparum. A similar statistical approach, as described for the behavioural analyses, was used to test for variation in the number of parasites (oocyst load and total number of sporozoites per whole mosquito body) 70 . Given the highly over-dispersed nature of parasite abundance data, a negative binomial distribution was assumed in these analyses (GLMM: glmmADMB, nlme package; R Core team 66 : R v.3.2.3 and RStudio 1.1.463 66 ). A backwards elimination approach was used to test for the significance of all fixed effects and interactions, while controlling for random variation due to replicate, as for the analyses of oocyst and sporozoite prevalence.

Data availability
All data supporting the findings of this study are available within the article and its Supplementary Information files. The sequencing data are available at the NCBI database (BioProject ID PRJNA756244), and are available from the corresponding authors upon request.