Cultured skin microbiota attracts malaria mosquitoes
© Verhulst et al. 2009
Received: 23 June 2009
Accepted: 17 December 2009
Published: 17 December 2009
Skip to main content
© Verhulst et al. 2009
Received: 23 June 2009
Accepted: 17 December 2009
Published: 17 December 2009
Host-seeking of the African malaria mosquito, Anopheles gambiae sensu stricto, is guided by human odours. The precise nature of the odours, and the composition of attractive blends of volatiles, remains largely unknown. Skin microbiota plays an important role in the production of human body odours. It is hypothesized that host attractiveness and selection of An. gambiae is affected by the species composition, density, and metabolic activity of the skin microbiota. A study is presented in which the production and constituency of volatile organic compounds (VOCs) by human skin microbiota is examined and the behavioural responses of An. gambiae to VOCs from skin microbiota are investigated.
Blood agar plates incubated with skin microbiota from human feet or with a reference strain of Staphylococcus epidermidis were tested for their attractiveness to An. gambiae in olfactometer bioassays and indoor trapping experiments. Entrained air collected from blood agar plates incubated with natural skin microbiota or with S. epidermidis were analysed using GC-MS. A synthetic blend of the compounds identified was tested for its attractiveness to An. gambiae. Behavioural data were analysed by a χ2-test and GLM. GC-MS results were analysed by fitting an exponential regression line to test the effect of the concentration of bacteria.
More An. gambiae were caught with blood agar plates incubated with skin bacteria than with sterile blood agar plates, with a significant effect of incubation time and dilution of the skin microbiota. When bacteria from the feet of four other volunteers were tested, similar effects were found. Fourteen putative attractants were found in the headspace of the skin bacteria. A synthetic blend of 10 of these was attractive to An. gambiae.
The discovery that volatiles produced by human skin microorganisms in vitro mediate An. gambiae host-seeking behaviour creates new opportunities for the development of odour-baited trapping systems. Additionally, identification of bacterial volatiles provides a new method to develop synthetic blends, attractive to An. gambiae and possibly other anthropophilic disease vectors.
The African malaria mosquito Anopheles gambiae sensu stricto (hereafter referred to as An. gambiae) preferably feeds on human beings inside houses and is therefore one of the most effective vectors of Plasmodium malaria parasites . Although visual and physical cues play a role in the host-seeking behaviour of An. gambiae, host-seeking is mainly accomplished by odour-mediated anemotaxis in which volatile organic compounds (VOCs) of human origin provide essential cues . Humans are differentially attractive to mosquitoes because of the odours they emit [2–4]. The skin microbiota plays an important role in the production of human body odours  and without bacteria human sweat is odourless . Many volatile compounds seem to be widespread among bacteria, although others are strain-specific. Some strains can produce up to 60 different volatile compounds . Differences in foot odour production can be explained by micro-floral differences between humans  and a recent study by Xu et al  provides more evidence that there is a connection between the microbial composition on human skin and chemical signature of humans. If host-selection by An. gambiae is based on the species composition, metabolic activity and/or density of the skin microbiota, then this will bear a direct impact on the number of bites received per person and the resulting risk of infection . The non-random nature of host selection remains poorly understood, yet has an important impact on exposure to disease .
A study on Limburger cheese volatiles revealed the putative role of bacteria in mosquito olfaction [11, 12]. Knols et al  suggested that bacteria involved in the ripening of Limburger cheese may have originated from human skin and hence that these bacteria are responsible for the production of 'human-specific' VOCs that mediate the host-seeking process of malaria mosquitoes. Washing the feet with a bactericidal soap significantly altered the selection of biting sites of An. gambiae on a motionless naked volunteer . In addition, human eccrine sweat is attractive to An. gambiae, but only after incubation for one or two days . Microorganisms on the skin are responsible for the conversion of fresh sweat into sweat attractive to An. gambiae [6, 14].
The human odour profile consists of more than 350 compounds [15, 16]. It was examined whether human skin microbiota produces attractive VOCs (kairomones)  for An. gambiae when cultured in vitro and whether analysis of entrained odours collected from these microorganisms can lead to a synthetic blend attractive to An. gambiae. In this paper an attractant is defined as a compound or blend of compounds, which causes insects to make oriented movements towards its source .
The An. gambiae s.s. culture originated from Suakoko, Liberia (courtesy Prof. M. Coluzzi). Mosquitoes have been cultured in the laboratory since 1988 and received blood meals from a human arm twice a week. Adults were maintained in 30-cm cubic gauze-covered cages in a climate-controlled chamber (27 ± 1°C, 80 ± 5% RH, LD 12:12). They had access to a 6% (w/v) glucose solution on filter paper. Eggs were laid on wet filter paper and placed in tap water in plastic trays and fed daily with Tetramin® baby fish food (Melle, Germany). Pupae were collected daily and placed in 30-cm cubic cages for emergence.
Skin microbiota samples were taken from a human foot, because there is evidence that this body part produces VOCs that influence the selection of biting sites by An. gambiae . Each volunteer (all healthy males, Caucasian, aged 23, 25, 28, 29, and 31 years) was asked not to shower, drink alcohol and eat spicy food 24 hours before the experiment and not to use soap during the last shower. Volunteers were provided a nylon sock, which had to be worn 24 hours before the experiment. Samples were taken from the foot of each volunteer by using a sampling ring and washing buffer as described by Taylor et al . A sterile Teflon sampling ring (internal diameter 2.9 cm) was placed in the centre of the underside of the foot, and 0.75 mL of full-strength wash fluid (75 mM sodium phosphate buffer (pH 7.9) + 0.1% (v/v) Triton X-100, Merck, The Netherlands) was added. The surface of the skin, within the ring, was gently scrubbed with a sterile glass stick for 1 min and the fluid was pipetted in a 2 ml sample tube (Eppendorf®). Immediately thereafter the process was repeated at the same site, and the two samples were pooled and diluted 5× in half strength wash fluid . Diluted microbiota samples (100 μl) were spread on Colombia (sheep) blood agar plates (Tritium, The Netherlands; http://www.tritium-microbiologie.nl/) before use in the behavioural experiments.
The number of colony-forming units (cfu) in the samples taken from the human feet was determined using selective plates for the five microbiota genera most abundant on human skin. According to the method described by Taylor et al , media were selective for staphylococci, aerobic corynebacteria, micrococci, propionibacteria and Pityrosporum species (Tritium, The Netherlands).
Experiments were prepared and performed according to the methods described by Smallegange et al . For each test 30 (mated) female mosquitoes of 5-8 d old, which had never received a blood meal, were selected 14-18 h before the experiment and placed in a cylindrical release cage (d = 8, h = 10 cm) with access to tap water from damp cotton wool. The experiments were performed during the last 4 h of the scotophase, when An. gambiae females are known to be highly responsive to host odours [23, 24]. In each trial, test odours were released in the air stream before a group of mosquitoes was set free from a cage which was placed at the downwind end of the flight chamber, 1.60 m from the two ports. Mosquitoes were left in the flight chamber for 15 min. Specimens that entered each of the two trapping devices were counted at the end of the experiments. Mosquitoes remaining in the flight chamber were removed with a vacuum cleaner. Each trial started with a fresh batch of mosquitoes, clean trapping devices, and new stimuli. A randomized complete block design was used which included five test treatments and one control treatment (so six treatments total) over six days. For each treatment, 30 female mosquitoes were released in the olfactometer. The sequence of test odours was randomized on the same day and between days. Test stimuli were alternated between right and left ports in different replicates to rule out any positional effects. Surgical gloves were worn by the researcher at all times to avoid contamination of equipment with human volatiles.
Excised blocks of blood agar (1.5 × 1.5 × 0.3 cm) with or without microbiota were placed on a glass slide (1.5 × 1.5 cm) and then heated on a brass block (Figure 1). One blood agar block was placed in each trapping device. The temperature of each block (34.0 ± 0.2°C) was regulated with a universal thermostat with external sensor (UT 100, Conrad electronic, The Netherlands).
In the first series of experiments diluted microbiota samples (100 μl) of one of the volunteers (28 years old) were spread on Colombia (sheep) blood agar plates (Tritium, The Netherlands) and incubated for 0, 6, 12, 24, 36, 72 hours at skin temperature (34°C)  before testing. Small blocks (1.5 × 1.5 × 0.3 cm) of blood agar incubated with skin microbiota were tested against small blocks of sterile blood agar.
In the second series of experiments all samples, originating from one of the volunteers (28 years old), were incubated for 12 h, after it had been decimally diluted (1:1; 1:10; 1:100; 1:1,000 or 1:10,000). Sterile control blood agar plates were incubated together with the plates with the skin microbiota. Excised blocks of blood agar incubated with skin microbiota were tested against blocks of sterile blood agar.
For a third series of experiments a bacterial sample was taken as described above from five volunteers (male, Caucasian) to test whether the main result obtained by using bacteria of the volunteer in the first series of experiments could be repeated with samples from other volunteers. The volunteer from the first series of experiments was one of the five volunteers in this experiment (# 4). Bacterial samples of each volunteer were diluted to a concentration of 2.63·102 cfu per cm2 (similar as found in the first series of experiments), spread on blood agar plates and incubated for 12 hours (based on the result of the first series of experiments). Blocks of blood agar with microbiota from the volunteers were tested against a control of sterile blood agar in the olfactometer.
To verify that An. gambiae is attracted to VOCs released by microorganisms common on human skin , an olfactometer experiment was conducted with a reference strain of Staphylococcus epidermidis (DSMZ 11047). Small blocks (1.5 × 1.5 × 0.3 cm) of blood agar grown with S. epidermidis at a concentration of 2.63·102 cfu per cm2 for 12 hours were tested against a control of sterile blood agar in the olfactometer.
In the first series a negative control of blood agar with microbiota that was not incubated was randomized with the treatments. In the second and third series two excised blocks of blood agar (1.5 × 1.5 × 0.3 cm) without microorganisms were tested against each other and randomized with the treatments as a control.
Experiments in which only clean moist air was released from both ports of the olfactometer were conducted to test the symmetry of the trapping system. Six tests were performed on one day which showed that the system was symmetrical (χ2-test, d.f. = 1, P = 1.00, trap entry response 14.5%).
For each test 50 (mated) female mosquitoes, five to eight days old, which had never received a blood meal, were selected 14-18 h prior to the experiment and placed in a cylindrical release cage (d = 8, h = 17.5 cm) with access to tap water from damp cotton wool. The experiments were performed during the last 4 h of the scotophase. The mosquitoes were released from the release cage in the centre of the large cage. After 4 h, the MM-X traps were closed and transferred to a freezer to kill the mosquitoes. Experiments were repeated for six days, altering the side of each treatment every day. Surgical gloves were worn to avoid contamination of equipment with human volatiles.
On six mornings, experiments with unbaited traps in the MM-X setup were done to test the symmetry of the trapping system. Sterile blood agar was tested against sterile agar without blood to test the effect of the blood in the agar. As a control, sterile blood agar (1.5 × 1.5 × 0.3 cm) was tested against sterile blood agar.
Compounds present in the odour blend which was tested in experimental room trapping experiment.
Compound (dilutions in H2 O)
Release rates μg/h
2-methylbutanoic acid (1:1000)
3-methylbutanoic acid (1:1000)
distilled water 0.10 mm
distilled water 0.05 mm
distilled water 0.03 mm
One hundred μL of each compound (Fluka, Sigma, ≥ 95%; Table 1), either pure or diluted in distilled water, was dispensed from sealed sachets (25 × 25 mm) of Low Density PolyEthylene (LDPE; Audion Elektro, The Netherlands). The thickness of the polyethylene material was varied to adjust release rates for each compound  (Table 1). LDPE sachets were suspended from a hook and placed inside the black tube of the MM-X trap (Figure 2). The control consisted of an equal number of sachets with the same size and thickness as the sachets containing experimental compounds, but filled with distilled water only. Release rates (μg/h) were measured by weighing the sachets before and after the experiments .
Samples were analysed by thermal desorption from the cartridge onto a cold trap and subsequent thermal desorption for introduction into the Trace GC Ultra coupled to a Trace DSQ (both from Thermo Scientific, The Netherlands). The thermal desorption programme consisted of a 3 min He dry purge and 1 min He prepurge both at 30°C. Cartridge desorption was performed at 250°C for 3 min and the volatiles were focused on a general purpose hydrophobic cold trap (Markes) at 0°C. Injection onto the analytical column was achieved by heating of the cold trap at the maximum speed (>60°C/s) to 250°C and splitting of the carrier gas (He) resulting in an injection of 1/6 of the total volatile amount. The transfer line between the cold trap and the GC was kept at 160°C.
A 30 m × 0.25 mm ID × 1.0 μm Rtx-5 MS with He carrier gas (1.0 mL/min) was used. The GC oven temperature programme was: 3 min at 45°C, ramping of 8°C/min to 280°C and a 2 minute hold at 280°C. The transfer line between the GC and MS was kept at 275°C. Mass-spectra were recorded by electron impact ionization at 70 eV, scanning in positive mode from 35-300 m/z with a scan speed of five scans/s and an ion source temperature of 250°C. The filament was switched off from 13.6-13.8 min because of a high background peak.
Peak analysis was performed using Xcalibur software and peak deconvolution by AMDIS, http://chemdata.nist.gov/mass-spc/amdis/. The obtained spectra were compared to the NIST-library. Calculated and reported retention indices and injection of authentic synthetic reference compounds (Table 1) provided additional information for identification.
Carbon dioxide (CO2) is assumed to play an important role in mosquito host-seeking behaviour . Therefore, CO2 levels in the olfactometer were measured on two days. Simultaneously, An. gambiae females were released to investigate whether their response to skin microbiota is (partly) due to CO2 emission from the skin microbiota. The concentration of CO2 inside the olfactometer was measured at two different positions (Figure 1) using a Xentra 4100 CO2 analyzer (Servomex, The Netherlands), at intervals of 3 min, according to the method described by Spitzen et al . Carbon dioxide concentrations were measured over a range of 0 - 1030 ppm with an accuracy of 0.1 ppm. The data were downloaded to a PC using Das Wizard© 2.0 software (Measurement Computing Corporation, USA).
For this purpose, blood agar plates with a concentration of skin microorganisms of 2.63·102 cfu per cm2, incubated for 12 h at 34°C, were prepared, and blocks of 1.5 × 1.5 × 0.3 cm, heated on a brass block (34°C), were tested in the olfactometer.
For each two-choice test in the olfactometer and MM-X setup a χ2-test was used to analyze whether the total (i.e. sum of all replicates) number of mosquitoes that was trapped in the treatment trapping device and the total number that was trapped in the control trapping device differed from a 1:1 distribution (P < 0.05). A Generalized Linear Model (GLM, P < 0.05; Genstat for Windows, release 9.2) with binomial function, linked in logit, dispersion estimated, was used to investigate the effect of treatments on the trap entry response in the olfactometer experiments. The trap entry response is defined as the number of female mosquitoes caught in both trapping devices as the percentage of mosquitoes that flew out of the release cage .
Differences in CO2 concentrations at the different time intervals and between blood agar with or without skin microbiota were tested using a t-test for each time-interval (P < 0.05; Genstat for Windows, release 9.2).
The abundance of compounds in the chromatograms of the GC-MS analysis were fitted to an exponential regression line (Genstat for Windows, release 9.2) to test whether the concentration of microbiota present on blood agar plates had a significant effect on the abundance of each compound in the headspace samples of these plates (P < 0.01).
Differences between the abundance of compounds in the chromatograms of S. epidermidis and sterile blood agar plate headspace samples were tested using ANOVA. When a treatment effect was found (ANOVA, P < 0.05) a t-test was used to compare pairwise differences of the mean (Genstat for Windows, release 9.2).
Based on colony counts on blood agar plates, the bacterial density on the sole of the volunteer's foot was estimated to be 1.90·105 cfu per cm2. Selective plates showed that staphylococci were most abundant (1.86·105 cfu per cm2); corynebacteria (5.22·104cfu per cm2) and propionibacteria (4.54·103 cfu per cm2) were present in lower numbers. Micrococci and Pityrosporum were not found during any of the experiments.
In the second series of experiments the 1:10 dilution was chosen such that the concentration of bacteria before incubation was the same as in the first experiment (2.63·102 cfu per cm2), and higher and lower concentrations could be tested (1:1; 1:100; 1:1,000 and 1:10,000). The total bacterial density on the sole of the foot was estimated to be 1.14·107 cfu per cm2, the staphylococci density 9.27·106, corynebacteria 2.16·106 and propionibacteria 5.73·105 cfu per cm2. Traps with blood agar with skin microbiota dilutions of 1:1 or 1:10 caught significantly more An. gambiae than traps with sterile blood agar (Figure 4B; χ2-test, d.f. = 1, P < 0.01). The trap entry response was not significantly different between treatments (Figure 4B; GLM, d.f. = 5, P > 0.05). The control experiments with sterile blood agar on both sides showed no positional bias (Figure 4B; χ2-test, d.f. = 1, P = 0.157).
In the third series of experiments the blood agar with bacteria from each volunteer (2.63·102 cfu per cm2, 12 h incubation) caught significantly more An. gambiae than the sterile blood agar (χ2-test, d.f. = 1, P < 0.05).
Mean carbon dioxide concentrations during behavioural experiments in the olfactometer.
Mean CO2 concentration (ppm) ± SE
Sterile blood agar
Skin microbiota on blood agar
439.90 ± 1.63
441.18 ± 1.72
438.02 ± 1.58
438.59 ± 1.63
435.20 ± 1.52
435.60 ± 1.58
432.74 ± 1.48
434.09 ± 1.52
430.71 ± 1.43
432.17 ± 1.47
Bacterial headspace compounds.
Compounds found in lower amounts in diluted samples and control
Exponential regression parameters
A + B*(R**X)
Studies on human odour
Compounds found in higher amounts in diluted samples and control
Sterile blood agar tested in the MM-X setup against sterile agar without blood showed no significant differences in mosquito numbers caught (Figure 8; χ2-test, d.f. = 1, P = 0.74). On average, 52.5% of the mosquitoes that left the release cage were caught.
Two unbaited MM-X traps caught equal numbers of mosquitoes showing that the system was symmetrical (Figure 8; χ2-test, d.f. = 1, P = 0.24). Together the unbaited traps caught on average 16% of the released mosquitoes.
The MM-X trap containing the ten synthetic odours caught significantly more mosquitoes than the control trap (Figure 8; χ2-test, d.f. = 1, P = 0.011). Thirty-three percent of the mosquitoes that left the release cage were caught by the two traps (Figure 8).
Good understanding of olfactory-mediated behaviour is crucial for the development of mosquito control strategies using host-derived semiochemicals . Through the analysis of organic volatiles produced by microbiota isolated from the human skin, compounds were identified that affect the host-seeking behaviour of An. gambiae. This finding underlines the important role of microbiology in the elucidation of mosquito-host interactions. Using the compounds identified a synthetic mixture attractive to this mosquito species was developed.
Although blood agar alone was attractive to the mosquitoes, the microbiota mixture used in this study caught significantly more mosquitoes than sterile blood agar, and it is therefore concluded that the volatiles produced by the microbiota themselves are chiefly responsible for the observed effects. The blood agar used in these experiments is a medium rich in organic substances and many of the volatiles identified in the headspace analysis of the blood agar plates incubated with bacteria were also found in the sterile blood agar plates (Figure 5, 6, 7). This could explain why blood agar itself was attractive to the mosquitoes. However, the abundance and composition of volatiles emanated by incubated blood agar plates was much different from those of sterile blood agar plates, causing the mosquito's choice for the former. Volatiles associated with blood can be attractive to mosquitoes . The indoor trapping experiments with agar with and without blood however, showed that blood itself was not a source of attractiveness for the mosquitoes (Figure 8).
The results obtained with skin microbiota are likely representative for humans in general, as the skin microbiota from five men caused attractiveness to An. gambiae. To determine the possible correlation between skin microbiota composition and the attractiveness of humans to malaria mosquitoes, the attractiveness of the volunteers needs to be investigated. A research like this would require a higher number of volunteers .
Classification of the microbiota in the foot samples used in the first and second olfactometer experiments showed that staphylococci were most abundant and corynebacteria occurred in low numbers only. Micrococci and Pityrosporum were not found on the foot of the volunteer, although these are reportedly present on the feet of 30-58% of healthy humans [26, 33]. Staphylococcus species have been reported to produce 3-methylbutanoic acid , a compound that was also detected in the headspace analysis in this study and is associated with foot malodour . Corynebacteria and propionibacteria are capable of catabolizing skin lipids to Long Chain Fatty Acids (LCFAs; C14-C30) and LCFAs to Volatile Fatty Acids (VFAs; C2-C12) [36, 37], which were also present in our headspace samples (2-methylbutanoic acid, 3-methylbutanoic acid). Previous studies have shown that aliphatic carboxylic acids play a role in the host-seeking behaviour of An. gambiae [11, 22, 38].
The human odour profile consists of more than 350 compounds [15, 16]. Eight of the 14 putative attractive compounds found in the present study have been reported previously from studies on human odour or human sweat [15, 16, 39–41] (Table 3), which links our results on odour production by in vitro cultured skin microbiota to these previous studies.
With the new approach presented here it is possible to identify compounds that affect the host-seeking behaviour of An. gambiae. Although there is no information on the correlation between the release rate of the compounds present in the synthetic odour blend and the actual concentration of the odorant chemicals in the headspace of the microbiota, the behavioural response to a blend of 10 compounds that were abundantly present among the bacteria-derived VOCs suggests that this strategy of kairomone identification is an effective means of kairomone discovery. Testing the blend of ten compounds in indoor trapping experiments represents an intermediate research step between laboratory-based olfactometer studies and (semi-)field studies, as the MM-X traps are currently also used in semi-field and field experiments in Africa [42–44].
Although the synthetic blend was more attractive than the control, the trap catches were lower than when skin microbiota or agar alone, were tested. Both quantity and quality of constituents present in synthetic blends are known to have an effect on trapping efficacy [11, 22, 45, 46]. As the release rates of the chemicals from the sachets depended on the volatility of the compounds and the size and thickness of the LDPE sachets  (Table 2), one would expect that a stronger positive effect of the blend can be achieved by influencing the release rates of the individual components in the blend by variation of these characteristics. In addition, some compounds that were present in the blend may have had a negative effect on the attractiveness, depending on concentration, and other compounds that may increase the attractiveness of the blend may be missing .
Inhibition of the metabolism of certain skin bacteria may reduce a person's attractiveness to malaria mosquitoes. Indeed, some compounds like citral, citronellal and geraniol block foot odour-producing enzymes in bacteria [35, 48]. This knowledge can lead to the development of a new class of odour-masking or inhibitory compounds, which can be exploited in the protection from mosquito bites, aiming at compounds that reduce the production of attractive volatiles on the human skin.
Skin microorganisms are known to determine the human odour profile  and with the results presented here it is plausible that the composition of the skin microbiota determines an individual's attractiveness to malaria mosquitoes. The discovery that human skin microorganisms mediate malaria mosquito behaviour provides new opportunities for the control of this disease, for example biotechnological approaches using bacteria for mass production of mosquito attractants or modification of the composition of the microbial flora on the human skin to reduce attractiveness.
Skin microorganisms attract malaria mosquitoes when grown on blood agar in both olfactometer and indoor trapping experiments. A study with five volunteers showed that this effect is probably representative for humans in general. Analysis of the headspace of the skin bacteria in a dilution series resulted in fourteen putative kairomones. A synthetic blend of 10 of these was attractive to An. gambiae. This approach to identify semiochemicals could potentially be a novel means of vector-borne disease control through the deployment of semiochemical-baited trapping systems [31, 49]. Further knowledge of the effect of skin bacteria on mosquito attraction could support the development of new repellents by blocking skin bacteria or the mass production of skin bacteria as mosquito attractant.
We would like to thank all volunteers for their participation and are grateful to F. K. M. van Aggelen, A. J. Gidding and L. Koopman for assistance with mosquito rearing. We thank Prof. Dr. M. Dicke, Dr. J. J. A. van Loon and Dr. Y. T. Qiu for commenting on an earlier version of this manuscript, R. de Vrijer-Albers for assistance with the microbiology work, Dr. F. C. Griepink for providing LDPE material and Dr. H. M. Smid for making the photographs. This study was funded by a grant from the Foundation for the National Institutes of Health (NIH) through the Grand Challenges in Global Health Initiative (GCGH#121). BGJK is supported by a VIDI grant (#864.03.004) from the Netherlands Organization for Scientific Research (NWO).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.