Daily rhythms in antennal protein and olfactory sensitivity in the malaria mosquito Anopheles gambiae

We recently characterized 24-hr daily rhythmic patterns of gene expression in Anopheles gambiae mosquitoes. These include numerous odorant binding proteins (OBPs), soluble odorant carrying proteins enriched in olfactory organs. Here we demonstrate that multiple rhythmically expressed genes including OBPs and takeout proteins, involved in regulating blood feeding behavior, have corresponding rhythmic protein levels as measured by quantitative proteomics. This includes AgamOBP1, previously shown as important to An. gambiae odorant sensing. Further, electrophysiological investigations demonstrate time-of-day specific differences in olfactory sensitivity of antennae to major host-derived odorants. The pre-dusk/dusk peaks in OBPs and takeout gene expression correspond with peak protein abundance at night, and in turn coincide with the time of increased olfactory sensitivity to odorants requiring OBPs and times of increased blood-feeding behavior. This suggests an important role for OBPs in modulating temporal changes in odorant sensitivity, enabling the olfactory system to coordinate with the circadian niche of An. gambiae.

Tsetse flies, Drosophila and cockroaches have rhythmic daily changes in olfactory sensitivity [21][22][23] , and An. gambiae exhibit diel and circadian rhythms of Blood-feeding behavior 1,5 . Previous work demonstrated the presence of rhythmic proteins in mouse liver using 2D gel-electrophoresis and mass spectrometry for protein quantification and identification, respectively 24 . Importantly, while it should not be assumed that a rhythm at the mRNA level will automatically be reflected at the protein abundance level, in many cases such coordination does occur 24 . In our study of An. gambiae mosquitoes we hypothesized that rhythms in olfactory gene expression would correspond with rhythms in protein abundance, and that these rhythmic profiles would correspond with time-of-day changes in detection/sensitivity to host odorants. We used quantitative multiple reaction monitoring (MRM) proteomics to measure changes in protein abundance in adult female olfactory tissues, comparing antennae and total head appendages (THA; maxillary palps, antennae and proboscis). Further, we measured olfactory responses induced by major host-derived chemicals from female adult antennae by electroantennograms (EAGs) to detect dose-dependent and time-of-day specific differences. Finally, we provide compelling behavioral evidence that confirms the rhythmic nature of flight and blood-feeding behavior in the laboratory, and that correlates strongly with our transcriptome, proteomic and physiological evidence for time-of-day specific changes in olfactory function.

Results
Targeted quantitative proteomics. We focused our proteomics study on rhythmic olfaction genes identified in our An. gambiae transcriptome study 7,8 in olfactory tissues. We hypothesized that rhythmic levels of these transcripts will be translated into quantifiable protein rhythms, and lead to time-of-day specific changes in olfactory sensitivity and behavior. In order to assess the quality of our protein extraction, determine which olfactory proteins were likely 'visible' to a proteome, and generate empirical MS/MS spectra to be used for targeting, we performed data-dependent bottom-up proteomics on tryptic digests of mosquito heads, THAs and antennae ( Fig. 1 and Fig. S1a). Single-dimensional nano-UHPLC MS/MS was chosen as it is identical to the level of separation analyzed by the targeted proteomics approach, thus allowing direct comparisons for verification and validation of targets. Our combined injections for bottom-up proteomics yielded a total of 1022 protein and 6157 peptide identifications at a 1% false discovery rate (Supplementary Dataset 1).
In order to determine if proteins identified by qualitative proteomics had rhythms specific to sensory tissues, we utilized targeted quantitative MRM proteomics to both identify and determine concentration in antennae and THAs [25][26][27][28][29][30] . Biological material was collected every 4 hr for 24 hr from adult females maintained under strict LD cycle conditions (12 hr light: 12 hr dark including 1 hr dawn/dusk transitions). In antennae and THAs, we were able to defensibly quantify 21 of 25 targeted proteins, respectively, including 11 OBPs and 5 other putative chemosensory proteins (Fig. 2, Fig. S3). The profiles of protein abundance for these proteins were found to be rhythmic. Detected OBPs include OBP1 (AGAP003309), known to bind indole and DEET 11,15 , and OBP20 (AGAP005208), known to bind DEET, 6-methyl-5-heptone-2-one and indole 14 . Detected and rhythmic in our analysis, OBP1, 2 (AGAP003306), 3 (AGAP001409), 20, 22 (AGAP010409), 25 (AGAP012320), 26  (AGAP007287) are reported to have gene expression enriched in female heads/olfactory organs relative to male tissues and/or female bodies 10,12,13 . A recent RNA-seq experiment also revealed that many of these OBP genes are down-regulated in antennae following bloodfeeding and as the mosquito transitions to oviposition behavior 16 . These studies provide further evidence that these OBPs in particular are potentially involved in olfactory host-seeking behavior in females. Two other chemosensory proteins, SAPP1 and A10, were also found to be rhythmic. Additionally, the takeout proteins TO1 (AGAP004263) and TO2/3 (AGAP012703/AGAP004262), which have been implicated in chemoreception and in blood-feeding behavior 1,19 , were detected and rhythmic in antennae and THAs. Many protein rhythms were robust, with 3.2, 8.4, 4.5 and 5.3 peak-totrough fold changes observed for OBP1, OBP26, TO1 and SAPP1 antennae protein abundance, respectively. In general, olfactory proteins were at their highest abundance at Zeitgeber time (ZT) 16 and lowest between ZT4 and ZT8 (ZT12, lights off; ZT0, end of dawn transition; Fig. 2 and Fig. S3). Protein profiles were compared with corresponding time-specific mRNA profiles, reanalyzed in 24 hr format from our 48 hr time course microarray analysis 7 , and subjected to identical cosinor analysis to determine peak phase. Most of the proteins had a rhythmic phase lagging their corresponding RNA expression rhythm. With the exception of OBP10 (AGAP001189), which had antiphasic RNA and protein levels, OBP RNA rhythms peaked at ZT10.1 6 0.9 (hr 6 S.E.M.) and protein rhythms at ZT16.8 6 0.5 (measured by cosinor analysis), resulting in a peak OBP RNA to protein lag of 6.9 hr 6 0.7 hr (Fig. 2b,c). This common peak in OBPs was consistent across biological replicates and between antennae and THAs (cosinor analysis peak in antennae replicate 1, ZT18. 4 Fig. S3b-f). Such a large phase lag between RNA and protein rhythms has been observed previously 24,31 , e.g. in mouse liver the RNA-protein lag of several clock components  correspond highly with RNA expression profiles from whole-heads. Proteins are grouped into OBPs, non-OBP chemosensory proteins (AGAP007286, SAPP1, A10, TO1, and TO2/3) and non-olfactory proteins (CYP6P3, RFeSP, PPO6, and VATI). All gene symbols as listed in VectorBase except for the following genes: RFeSP (homologue to Drosophila Rieske iron-sulfur protein, AGAP008955), A10 (homologue to Drosophilla antennal protein 10, AGAP008055), VATI (predicted V-type proton ATPase catalytic subunit I, AGAP001587) and AGAP007286 (Ae. aegypti OBP43 homologue). Protein abundance normalized with tubulin. See Fig. S3 for protein rhythms in the second biological replicate antennae run and THAs. (b) Histogram of peak RNA expression phases in total heads and the corresponding peak antennal protein levels as determined by cosinor analysis from genes/proteins in panel A. (c) Cosinor analysis of antennae OBP protein levels from panel A (p , 0.001; acrophase ZT17.7); note OBP10 excluded as it has antiphasic expression compared to the other OBPs. and clock controlled genes, such as per1, per2, Id2 and Sdh1, is 6-9 hr 24,31,32 . The lag could be explained by the additional action of post-transcriptional 33 and post-translational processes (e.g. protein phosphorylation and turnover), and our previously published gene expression analyses suggest that these mechanisms are in fact rhythmically regulated in An. gambiae 7 .
As part of our proteomics analysis we additionally targeted three proteins with rhythmic mRNA levels 7,8 that are specifically involved in metabolic detoxification (CYP6P3, AGAP002865), immunity/cuticle pigmentation (PPO6, AGAP004977) and vesicular type H 1 ATPase (VATPase) activity (VATI, AGAP001587): In all cases, protein levels were found to be rhythmic (Fig. 2, Fig. S3), and therefore suggests functional changes. As CYP6P3 metabolizes pyrethroids, such rhythmicity may confer time-of-day specific changes in insecticide resistance; PPO6 is a key enzyme involved in the melanization process, an important immune sequestering process; and changes in VATPase activity, amongst several things, contributes to synapse function and neurotransmitter recycling, see 7,8 . It is plausible that rhythmicity in this process might contribute additionally to the temporal changes observed in olfactory sensitivity (see electrophysiological analysis, below). In THAs, in addition to proteins reliably detected in antennae, we also defensibly detected and profiled an additional three non-olfactory proteins which were rhythmic at the transcript level 7 (Fig. S3d): The citric acid cycle enzyme isocitrate dehydrogenase (IDH; AGAP006660), the glutathione-Stransferase metabolic detoxification protein GSTD7 (AGAP004163), and Actin5C (AGAP000651). Note that the Drosophila Act5C gene promoter has been used frequently for transgenic mosquito experiments.
Not everything in our preparations had a protein profile with a peak during the early to mid-night phase (ZT12-20). In fact we observed profiles suggesting rhythmic peaks earlier in the 24 hr day; e.g. in THAs GSTD7 has a cosinor peak phase estimate of ZT9.4 and a peak expression measured at ZT4. There were several patterns of protein and RNA where this phase coordination was notably different: In both antennae and THAs RFeSP had a constitutively expressed RNA profile but rhythmic protein levels, while in THAs, GSTE3 (AGAP009197) displayed the opposite phenomenon, with rhythmic mRNA levels but corresponding constitutive protein levels. GSTE3 has been implicated in insecticidal detoxification 34 . Furthermore, in THAs the Actin5C protein profile was antiphasic to its encoding mRNA (cosinor peak phase estimate ZT3.0 and ZT16.7 for RNA and protein, respectively). Such a distinct absence of phase coordination between RNA and protein has been reported before 24,33 . For example, Reddy et al. similarly found half of cycling mouse liver proteins lacked a corresponding rhythmic mRNA; a phenomenon they attributed to post-transcriptional processes 24,33 . However, considering that our mRNA profiling was performed in total heads and protein analysis from specific olfactory tissues, we cannot exclude the possibility of tissue-specific mRNA rhythms that might correspond better with observed protein results.
Using proteomics we could not defensibly detect the obligatory co-receptor required for all odorant receptor (OR) function, Odorant Receptor Coreceptor Agam\Orco, also known as Or7 (AGAP002560) 35,36 . RNA levels of Agam\Orco are rhythmic, peaking at ZT10 (prior to dusk and the onset of nocturnal behavioral activities involving olfaction) [1][2][3][4][5]7 . We previously hypothesized that rhythmic Agam\Orco protein levels could serve as a mechanism for rhythmic control of olfactory sensitivity 7 . However, immunoblot analysis revealed constitutive Agam\Orco levels (i.e. no significant differences in protein abundance were detected by Kruskal-Wallis ANOVA; cosinor analysis, n.s.; Fig. 3), indicating Agam\Orco is unlikely to gate rhythmic olfactory sensitivity. Note that we were unable to detect any odorant-specific ORs in our microarray analysis 7 or proteomic analyses, presumably due to their low abundance 16 .
Electrophysiological analysis. We next determined if the temporal rhythms described in protein levels correspond with olfactory sensitivity. We used electroantennogram (EAG) analysis to measure the olfactory responses induced by host-derived odorant chemicals. At ZT4 (morning), ZT8 (afternoon), ZT12 (dusk) and ZT16 (night), mosquito antennae were challenged with major host-derived odorant chemicals (nonanal, indole, geranyl acetone, and a mixture of hexanoic acid, geranyl acetone, nonanal, indole, 3methylindole and p-cresol). As several of the odorants are hydrophobic, it is expected that OBPs will be necessary for the mosquito to detect these stimuli. Since OBPs are rhythmic in protein abundance, we predicted rhythmic sensitivity in detection of these compounds. Indeed, we found time-of-day specific olfactory responses to all four stimuli, with sensitivity peaking at ZT16 (night) and least at ZT4 or ZT8 (day) (Fig. 4a-b). In all cases, the time-of-day specific differences in EAG amplitudes were comparable to the differences observed due to 10-fold changes in odorant concentration (mean peak-to-nadir difference in each time series: Geranyl acetone 44%, indole 50% and nonanal 61%; mean increase in response with each 103 increase in odorant concentration: Geranyl acetone 15%, indole 67%, and nonanal 81%). It is therefore plausible that the #8.4-fold rhythmic changes observed in OBP abundance (antennae mean 6 SEM fold-change, 3.4 6 0.4; THAs, 3.2 6 0.6) could account for such differences in EAG responses. This is especially so given that in some cases multiple OBPs can bind a single odorant, and vice versa 11,14,15,17 . These findings are broadly consistent with our observed OBP protein rhythms (Fig. 2, Fig. S3) and mosquito behaviors described here (Fig. 6) and by others 1,5 . Specifically, protein expression and olfactory sensitivity were also lowest at ZT4-ZT8 and highest at ,ZT16, and blood-feeding and flight behaviors were higher at night than during the daytime. These data are particularly interesting for indole as there is significant evidence that OBP1 contributes to antennal sensing of this odorant 11,17 . OBP1 protein rhythm has a nadir at ZT4-ZT8 and a peak at ZT16 (Fig. 1, Fig. 2 corresponds with the lowest and highest sensitivity to indole in our electrophysiological measurements. Next, we looked at olfactory responses to nonanal (host odorant that elicited highest EAG amplitude response from our panel) at ZT16 and at ZT8 from normally reared mosquitoes, but tested under identical lighting conditions. In order to control for any acute effects of light or dark on influencing the odorant response, the ZT8 group was pretreated with 4 hr of darkness prior to EAG recording, and both ZT8 and ZT16 animals were tested under red light (Fig. 5a). It is plausible that light could influence olfactory function, either acting via the compound eye, or acting directly on the antennae via the blue light photoreceptor cryptochrome 1 (CRY1, AGAP001958), which is likely to be expressed locally in chemosensory organs of the mosquito 7,37,38 . Consistent with the experiments conducted under normal LD cycle conditions ( Fig. 4a-b), this analysis revealed both a concentration dependent and a time-of-day effect: Mosquitoes at ZT16 retained a higher EAG sensitivity than dark-treated ZT8 mosquitoes. This indicates that reduced sensitivity at ZT8 is not an acute effect of light on the function of the antennae, but in fact a reflection of an underlying circadian clock and/or diel regulated mechanism (driven by the environmental 24 hr LD cycle) 8 . Further, 24 hr rhythmic changes in electrophysiological sensitivity persist in constant dark (DD) conditions in Drosophila and cockroaches 21,23 , and are driven by peripheral clocks located in the Drosophila antennae themselves 39 . Therefore, it is likely that the observed time-of-day changes in EAG sensitivity in An. gambiae are driven not by an acute effect of light or darkness, i.e. a masked response, but by an underlying clock and/or diel mechanism. This corresponds with our previous work that suggests that OBP gene expression is dependent upon both mechanisms working in concert 8 .
Finally, we tested the time-of-day specific olfactory sensitivity of mosquitoes to hexanoic acid, a hydrophilic host odor constituent that does not require OBPs for detection. Mosquito olfactory responses were measured with increasing doses of hexanoic acid, and compared to nonanal, which served as a positive control. As expected, responses to the nonanal control exhibited time-of-day sensitivity, peaking at ZT16, while responses to hexanoic acid did not change (two way ANOVA; Tukey post hoc tests, n.s.; Fig. 5b). Acid responses were however dose dependent (all concentrations are significantly different from each other; Tukey post hoc tests, p , 0.001). Earlier investigations of female Culex pipiens mosquitoes also did not reveal a time-of-day specific change in electrophysiological sensitivity to lactic acid 40 , which presumably does not require an OBP for detection.
Flight and blood-feeding behavior rhythms. Having established a clear time-of-the day modulation of abundant olfactory components in adult female An. gambiae using transcriptomic and proteomic approaches and sensory physiological evidence, we investigated if the An. gambiae strain from our laboratory maintained diel flight and blood-feeding behaviors typical of mosquitoes observed in the wild. Flight activity of host-seeking mosquitoes was monitored using an infra-red beam break system 3 , and our data revealed a clear onset of mosquito behavior at the beginning of the night (Fig. 6a), a finding consistent with laboratory and field observations 5,41 . Blood-feeding behavior has previously been reported to be under circadian regulation in laboratory conditions using a membrane feeder 1 and to be nocturnal in the field 42 . In the current study, blood-feeding preference was measured in our laboratory by exposing a human arm every 4 hr to individual cages of host-seeking females. After 6 min exposures, mosquitoes were visually examined for the presence of blood in their abdomens. We found that under diel conditions, blood-feeding was almost exclusively restricted to the night phase, with peak biting occurring in the early night (Fig. 6b), and coincident with timing of flight activity (Fig. 6a). The rhythm in blood feeding preference was found to persist for 2 days under DD conditions (Fig. 6c), demonstrating that this temporal behavior is driven by the endogenous circadian clock. We note the peak in bloodfeeding preference (ZT12-16) is coincident with the peak time of olfactory sensitivity as measured by EAG. It is also coincident with the peak in abundance of olfactory proteins such as the takeout proteins, known to be involved in regulating feeding behavior, and OBPs.

Discussion
This work provides additional compelling evidence of the important role of daily rhythms in the sensory biology of An. gambiae, the primary malaria vector. It has been hypothesized that expression dynamics of chemosensory genes may result in overt behavioral phenotypes 16 . Here, we identify the pre-dusk/dusk peak in expression of OBPs detected by microarray analysis 7,8 (genomics) corresponds with peak protein abundance (proteomics), and is coincident with the time of increased olfactory sensitivity to host odorants thought to require OBPs for detection (electrophysiology), and times of increased biting (blood-feeding) behavior. We also find that neither olfactory sensitivity to an odorant that does not require OBPs for detection (i.e. hexanoic acid), nor protein abundance of the olfactory co-receptor OR, Agam\Orco, changes across the 24 hr day. Rhythmic expression of OBPs peak at the time of highest olfactory sensitivity to hydrophobic odorants (,ZT16). Therefore, these results strongly suggest that rhythms in OBPs contribute to the changes observed in sensory and behavioral function. This is not surprising given that the knockdown of a single OBP, OBP1, results in a significant decrease in the sensitivity of major chemostimuli, including indole, in An. gambiae 11 and in Cx. quinquefasciatus 17 .
Though our data provides compelling evidence suggesting OBPs and other chemosensory proteins modulate olfactory sensitivity and behavioral outputs, we are aware that our quantitative transcriptome and proteome analyses did not detect another major chemosensory family, the ORs. Recent investigations have demonstrated a modulation of OR and OBP gene expression based on An. gambiae physiology, and how such differences are coincident with a switch from host-seeking to oviposition behaviors. However, these changes in OR expression were modest as compared to the OBPs 16 . Our own investigation into the role of ORs was inconclusive: The most abundant OR, Agam\Orco, did show a weak RNA rhythm in our microarray analysis 7 , but we found no evidence of rhythmicity in protein abundance. Furthermore, additional downstream signal transduction components, as previously suggested 39 , or rhythmic trafficking of olfactory components 43 may also modulate olfactory responses. The rhythmic blood-feeding responses described here are likely driven by several factors, including time-of-day changes in sensitivity to host odorants (which may be driven by rhythmic OBP levels we describe here); rhythmic responses to CO 2 , as reported in the hematophagous Triatoma infestans 44 ; and rhythmic protein abundance of behavioral/chemosensory factors, such as the takeout proteins TO1 and TO2/3, that we demonstrate to be rhythmic. Working alone or in concert with these above mentioned regulatory factors, it still remains a compelling hypothesis that rhythms observed in sensory organ OBP and takeout protein abundance could confer these dramatic time-of-day specific changes in odorant sensitivity and feeding behavior. The coincident times of peak protein abundance, olfactory sensitivity and behavior reflect the extraordinarily fine-tuned control of mosquito physiology, with OBP and other chemosensory protein abundance and high olfactory sensitivity up-regulated when needed (at night) and down-regulated when not required (daytime).
Our work highlights the important role of circadian/diel biology in the mosquito. Improved understanding of biological timing at the molecular level that underlies key physiological aspects of An. gambiae may prove to be important for the successful implementation of existing or novel control methods and future experimental design. Further, greater understanding of the rhythmic nature of blood-feeding is important. There is growing evidence that the use of insecticide-impregnated bed nets is acting as a selective pressure, potentially modifying both age and genetic composition of the mosquito population 45 . Thus, selection for mosquitoes that host-seek at times of the day when humans are not protected under bed nets may be occurring. were defensibly identified at this stage after searching the bottom up generated development data. OR protein abundance was presumably below the detection limits of the experimental design; or the preparation used excluded their extraction. Note that OBP3 (AGAP001409) and OBP10 (AGAP001189) were only defensibly detected in the antennae in the second biological replicate time course: For the purposes of presentation, these data are shown with the first replicate time course (Fig. 2). Raw protein identification data, including data from which the MRMs were derived, have been deposited into the PeptideAtlas: http://www.peptideatlas.org/PASS/ PASS00300.

Methods
MRM development, validation and acquisition. MRM transitions were determined largely from the empirical MS/MS data obtained from the bottom-up proteomics approach. Our analytical work was comparable to that employed in our previous MRM efforts 27,28,49 and in agreement with the guidelines described by the Aebersold group 29,50 . A combination of empirical (MS/MS), MRM-initiated full-scan sequencing and published spectra were the source for the MRM peptide transitions. Specific transition development was performed using Skyline (MacCoss Lab) and MRMPilot (ABSciex) combined with manual refinement. An example of each type of peptide optimization is shown in Fig. S1b-g; including MRM-initiated full scan sequence confirmation performed as previously demonstrated 51 .
Briefly, 4-9 transitions were chosen for each peptide used to describe a protein of interest for as many as 4-peptides per protein (several proteins only yielded one reproducibly detectable peptide for quantification). Transition validation and determination was made according to our previously published work and published sample validation methods 29   www.nature.com/scientificreports the most part inversely proportionate to the observed signal from the development data. This was performed to decrease the cycle-time thereby improving the sampling rate of the data (Supplementary Dataset 1). These were subsequently re-confirmed and tested using bulk mosquito-head extracts as a complex and complete sample stress-test. For each peptide analyzed, 3-4 transitions were chosen, similar to what we and others have observed 28,30,50 . Tubulins, due to their 24 hr constitutive RNA and protein expression, are frequently used in insect circadian experiments as an immunoblot loading control, including butterfly antennae 38,52 . b-tubulin (AGAP010929) was used as an internal standard for normalizing protein levels across the time courses: Peptide transition EIVHIQAGQ[C]GNQIGAK was used for antennae and YLTVAAVFR for THA normalization.
MRM data were acquired on a QTrap 5500 (ABSciex) running in triple-quadrupole or hybrid ion-trap mode as previously published 28 . MIDAS experiments (MRMtriggered Data-Dependent) were acquired using the hybrid triple quadrupole-ion trap transition mode. 2 ml of the samples as prepared earlier from each time point were injected in-triplicate onto a 100 mm 3 100 mm C18BEH column (Waters) running at 600 nl/min. A 90 min gradient from 2-35% acetonitrile-water (0.1% FA) was used with pre-run solvent blanks injected. Mid-run washes and quality control peptide injections were performed every 12 injections. Including tuning, approximately 100 LC/MS/MS (MRM) acquisitions were acquired. An example chromatogram of a complete MRM trace, for THAs collected at ZT16, is shown in Fig. S1d. Overall peak retention and reproducibility were excellent. Some runs exhibited a slight , 1 min (1-2%) systematic drift in retention-time (RT) throughout the .100 hr of acquisition, but this was readily corrected for with integration parameters. Most runs exhibited excellent stability. For example, RT stability for the btubulin peptide .FPGQLNADLR. (2y8) was 0.31%CV for all injections of antennae samples. This is typical of peptides observed in the antennae.
Peak Processing. Peak areas were integrated using MultiQuant (ABSciex) as described in Li et al 49 . Briefly, 30 s RT windows were used, and a 3-point Gaussian smooth was applied to all transitions. Peak area integrations of the quantifier MRM transition were converted to area ratio/area response by dividing by the peak area of the associated internal standard peptides from b-tubulin. Additionally, identical numbers of antennae or THAs were processed at each time point ensuring reasonable reproducibility of biological replicates. b-tubulin peak area varied by 5-8% within technical replicates (instrument variability) and less than 2-fold within biological replicates (extraction, antennae reproducibility). Average %CV for all peptides was excellent, at ,25% (uncorrected peak area) for antennae analysis and comparable for heads/THAs.
CircWave cosinor analysis. CircWave v1.4 software (www.huttlab.nl, www.euclock.org, a cosinor analysis program, courtesy of Dr. Roelof Hut,) was used to analyze the rhythmicity of gene and protein levels by fitting a Fourier-curve (one sine wave) to the data 53 . The p values reported are the result of F test from software.
EAG recordings. Female An. gambiae (4 days post emergence, mated but not bloodfed [i.e. host-seeking state]) were immobilized on ice. The head of the mosquito was excised with surgical microscissors (World Precision Instruments, Sarasota, FL) and mounted on an indifferent electrode. Electrodes contained chloridized silver wires in drawn-out glass capillaries filled with 0.1% KCl and 0.5% polyvinylpyrrolidone (PVP). The recording electrode accommodated the two antennae of the excised head after the tips of the antennae were clipped to provide a better contact. Electrode placement (5-10 min) in the dark was made possible by a 16 LED red light source (MaxMax, Carlstadt, NJ; 640 nm peak; 0.35-1.6 mW/in 2 at the level of the mosquito). Signals were amplified and directly recorded via an IDAC4-USB box (Syntech, Germany). Recordings were analyzed with EAG Pro version 1.1 software (Syntech). The preparation was held in a charcoal filtered and humidified continuous air stream (2.4 liters/min) delivered via a glass tube. A stimulus pulse (1.8 liters/min) was added to the air stream for 0.5 s. To prevent changes in air flow during stimulation, a charcoal-filtered air flow (0.6 liters/min) was delivered via another solenoid valve through a blank syringe into the glass tube and at the same distance from the preparation. Any change in antennal deflection induced by the stimuli or control puffs was recorded for 10 s. Animals tested at ZT12 and ZT16 were exposed to the normal 1 hr dusk dimming cycle. ZT16 group animals were maintained in a dark box (0-1 lux red light) until time of electrophysiological preparation. Red light provided by fluorescent lamps (Philips TLD 36 W/15, 660 nm peak, 4.5 mW/in 2 at level of mosquito) was used to aid experimental procedures during the dark phase of the LD cycle. In the dark-adapted experiment, the ZT8 group was placed into a dark box at ZT4 and tested in red light at ZT8 (thereby matching the conditions of the ZT16 group).
Stimuli. Chemicals of highest purity were used. Indole (98% purity), 3-methylindole (98%), hexanoic acid (98%), geranyl acetone (98%) and p-cresol (99%) were from Sigma-Aldrich; nonanal (95%) was purchased from Fluka. Chemicals were dissolved in the solvents dichloromethane (DCM; HPLC grade) or hexane (glass distilled) to make a stock solution of 100 mg/ml and decadic dilutions were made. An aliquot (10 ml) of a stimulus was loaded onto a filter paper strip, the solvent was evaporated for 30 s, and the strip was placed in a 5 ml polypropylene syringe from which various volumes were delivered onto the EAG preparation.
Data Analysis. Each recording was first baseline normalized, smoothed with a 10 min running average, amplitude determined and the value of the appropriate blank (solvent) control subtracted. Statistical analyses on the effects of time of day and chemical concentration were performed using SigmaPlot 12 (Systat Software, Chicago, IL) and GraphPad Prism 5 (GraphPad Software, La Jolla, CA). Non-parametric statistical analyses were used when Shapiro-Wilk normality test failed (p , 0.05) or data were square root transformed to correct for non-normal distributions.
Immunoblot analysis. Three mosquito heads from each time point were homogenized in 1X lysis buffer (30 mM Tris-HCl, pH 6.8, 10% SDS, 0.0002% bromophenol blue, 5% b-mercaptoethanol, 10% glycerol). Lysate from the equivalent of ,1 An. gambiae head was loaded and separated on a Novex NuPAGE 4-12% Bis-Tris gel (Life Technologies, Carlsbad, CA) and transferred to a polyvinylidene difluoride membrane. Membranes were blocked and probed overnight with a 151000 dilution of rabbit Agam\Orco (Or7, AGAP002560) antiserum 54 . Agam\Orco was detected by horseradish peroxidase-linked goat anti-rabbit IgG (152000, GE Healthcare Life Sciences) and developed with the ECL Western Blotting Detection System (GE Healthcare Life Sciences, Piscataway, NJ) per manufacturer's protocol. Two major bands were observed at ,50 kDa (consistent with the predicted molecular weight of 54 kDa) and ,100 kDa, which we hypothesize is an Agam\Orco heterodimer. As a loading control, membranes were stripped and re-probed with a 152000 dilution of mouse anti-b-actin mAb JLA 20 (Developmental Studies Hybridoma Bank). Actin was detected by horseradish peroxidase-linked goat antimouse IgG (152000, GE Healthcare Life Sciences) and developed. Densitometry was conducted using Image J software version 1.42q (NIH, Bethesda, MD).
Behavioral assays. Individual mosquito locomotor/flight activity was measured with a Locomotor Activity Monitor 25 (LAM 25) system (TriKinetics, Waltham, MA), as previously detailed 3 . Briefly, individual mated but not blood-fed adult female 4-6 days post-emergence (i.e. host-seeking state) mosquitoes were placed in 25 3 150 mm clear glass tubes with access to 20% HFCS in the tubes provided ad libitum. Flight activity recorded as infrared beam breaks per minute. All recordings occurred in a light-proof box with its own lighting system in a 12 hr/12 hr LD cycle (11 hr full light, 11 hr darkness (0 lux), and 1 hr dawn and 1 hr dusk transitions) with full light measured at the level of the LAM 25 between 69 and 119 lux. Mosquitoes were monitored for 7 full days. Locomotor flight activity was visualized in actogram format using ClockLab version 2.61 (Actimetrics, Wilmette, IL).
For blood-feeding preference assays, 4-8 days post-emergence mated adult female not blood-fed (i.e. host-seeking state) mosquitoes were separated into individual containers. Blood-feeding began the next morning. Mosquitoes were exposed (within inches) to human arm (the investigator's) for 6 min. Mosquitoes were then anesthetized with CO 2 and visually assayed for the presence of any blood in the abdomen. To exclude potential human circadian effects, similar assays were performed with mosquitoes raised under reverse LD cycle conditions and then exposed to the human investigator at various phases of the circadian cycle. Similar time-of-day specific results were attained.