The African malaria mosquito Anopheles gambiae sensu stricto continues to play an important role in malaria transmission, which is aggravated by its high degree of anthropophily, making it among the foremost vectors of this disease. In the current study we set out to unravel the strong association between this mosquito species and human beings, as it is determined by odorant cues derived from the human skin. Microbial communities on the skin play key roles in the production of human body odour. We demonstrate that the composition of the skin microbiota affects the degree of attractiveness of human beings to this mosquito species. Bacterial plate counts and 16S rRNA sequencing revealed that individuals that are highly attractive to An. gambiae s.s. have a significantly higher abundance, but lower diversity of bacteria on their skin than individuals that are poorly attractive. Bacterial genera that are correlated with the relative degree of attractiveness to mosquitoes were identified. The discovery of the connection between skin microbial populations and attractiveness to mosquitoes may lead to the development of new mosquito attractants and personalized methods for protection against vectors of malaria and other infectious diseases.
Citation: Verhulst NO, Qiu YT, Beijleveld H, Maliepaard C, Knights D, et al. (2011) Composition of Human Skin Microbiota Affects Attractiveness to Malaria Mosquitoes. PLoS ONE 6(12): e28991. doi:10.1371/journal.pone.0028991
Editor: Bradley S. Schneider, Global Viral Forecasting Initiative, United States of America
Received: September 20, 2011; Accepted: November 17, 2011; Published: December 28, 2011
Copyright: © 2011 Verhulst et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was funded by a grant from the Foundation for the National Institutes of Health through the Grand Challenges in Global Health Initiative (GCGH#121), and a grant from the Earth and Life Science Foundation of the Netherlands Organization for Scientific Research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Host location by female mosquitoes is mediated by host-derived physical and chemical cues. Physical cues include heat, moisture and visual cues, and play a role during orientation and landing , , . Chemical cues are considered most important for orientation and landing, especially for nocturnal mosquitoes ,  and humans can be ranked for attractiveness to mosquitoes by testing the emanations released from their skin , , , . The mosquito Anopheles gambiae Giles sensu stricto (hereafter An. gambiae), a nocturnal, highly anthropophilic species, is one of the most important malaria vectors in Africa . Volatiles released from human skin provide essential cues that guide this mosquito species to its host , .
Skin bacteria play an important role in the production of human body odour and without bacteria, human sweat is odourless to the human nose . Skin bacteria convert non-volatile compounds into volatile compounds having characteristic smells. The body odour of individual human beings correlates with the presence of specific microorganisms , ,  and with detailed skin microbial profiles, as assessed using denaturing gradient gel electrophoresis (DGGE) analysis . The interactions between skin microbes and the human host, however, are still poorly understood  and the effect of the skin microbial composition on disease vectors remains largely unknown .
Human eccrine sweat is more attractive to An. gambiae after incubation with skin bacteria for one or two days , and washing the feet with a bactericidal soap significantly alters the selection of biting sites by An. gambiae . Recently, it was shown that volatiles produced by human skin bacteria, grown in vitro, are attractive to female An. gambiae when tested in an olfactometer or with mosquito traps , .
Here, we examined in vivo how the composition of skin microbiota affects an individual's attractiveness to mosquitoes by assessing the attractiveness of 48 human males to An. gambiae and analysing their skin bacterial communities. We subsequently correlated the observed effects with microbial abundance and composition.
The Anopheles gambiae Giles sensu stricto colony originated from Suakoko, Liberia. Mosquitoes were reared according to the methods described previously .
The attractiveness of 48 adult males aged between 20 and 64 years to An. gambiae was examined. Forty-six men were of Caucasian origin, one man was of Asian and one of Hispanic origin. The Dutch Medical Ethical Review committee (METC, Project number ABR NL16928.081.07 amended in 2007) approved the study, and written informed consent was acquired from all subjects prior to participation. Volunteers were requested to refrain from drinking alcohol , eating garlic, onions or spicy food, taking a shower, using perfumed cosmetics and were asked to wear nylon socks provided by the research team for the twenty-four hours before the sampling event. Volunteers were free from chronic illnesses and not using any medication on a regular basis. The socks provided by the research team (100% polyamide, 40 denier, Hans Textiel, The Netherlands) were washed twice with 70% ethanol and dried in a ventilated oven at 80°C before use. Volunteers were instructed not to use soap the last time they showered before the experiment.
Skin emanations from each individual were collected twice on three different days by rubbing six glass beads  (15 mm in diameter, contained in a Teflon holder, Figure S1) for 10 min. against the underside of the left foot. Feet produce volatiles that are attractive to An. gambiae and there is evidence that this body part produces volatiles that influence the selection of biting sites by this mosquito species . Beads with skin emanations were tested for attractiveness to female An. gambiae in a dual-choice olfactometer (1.60×0.66×0.43 m) against a standard ammonia concentration of 136 ppm for six times in total: two consecutive assays on each of three mornings . Release of test stimuli was alternated between left and right ports of the trapping devices of the olfactometer to rule out any positional effects. Air speed at the ports was 0.21±0.01 m/s. The experimental room was maintained at a temperature of 27.9±0.7°C and a relative humidity of 62.3±5.8%. The temperature inside the flight chamber was 27.9±1.7°C and the humidity 69.0±4.6%. The humidity of the air led into the trapping devices was maintained above 80% and its temperature was 28.0±1.5°C.After use, the trapping devices were washed in a dishwasher at 45°C with biological soap (Sonnett tabs, Sonnet OHG, Germany). The glass beads were cleaned by rinsing in a solution of 10% Helmanex® II cleaning concentrate (Hellma GmbG & Co KG, Germany) in water, subsequently in distilled water, and finally in ethanol (99.8% purity; Merck, Germany). The rinsed beads were dried in an oven at 200°C for at least one h. Between experiments, the Teflon holder was cleaned with 70% ethanol and quick-dried with a heat gun (Ferm B.V., The Netherlands).
Skin bacterial diversity
The skin bacterial composition on the feet of the individuals was determined by selective and non-selective plate counts and 16S rRNA gene sequencing. Connecting the selective and non-selective plate counts to the relative attractiveness of the skin volatiles provided a first indication of whether the skin microbiota affects the attractiveness of human skin emanations to An. gambiae. The 16S rRNA genes contain hypervariable regions, the sequences of which can provide a detailed signature of the microbiota on the human foot.
On each experimental day a bacterial sample was taken from the sole of the left foot of each individual after assessing that individual's attractiveness to mosquitoes during two successive olfactometer experiments. Bacterial samples were taken by using a sampling ring and washing buffer as described before , . Seven hundred µL of the sample was added to 300 µL glycerol (87%, Merck, Germany) and stored at −80°C for later identification by 16S rRNA sequencing. The remainder of the sample was used for plate counts on selective and non-selective media.
Selective and non-selective plate counts.
Within three hours after the bacterial sample collection, 100 µL of each sample was decimally diluted, spread on Colombia (sheep) blood agar plates (Tritium, The Netherlands) and incubated at skin temperature (34°C) to determine bacterial densities by counting colony-forming units (CFU). A range of selective media was used to determine the diversity of the human skin microbiota samples according to the method described before  (Table S1). Media were either selective for staphylococci, aerobic corynebacteria, micrococci or Propionibacteria (Tritium, The Netherlands).
Bacterial 16S rRNA genes.
In total 144 bacterial samples were collected; three samples from each volunteer on different mornings. Microbial DNA was extracted from 41 bacterial samples using a FastPrepDNA soil kit (MP Biomedicals, USA). Of these, 13 samples did not yield enough DNA and were therefore excluded from further analysis. The remaining 116 samples were extracted using the Mo-Bio Power Soil kit (MO BIO Laboratories, Inc., USA). Results of the two extraction methods were not different and the data were combined for the final analysis.
PCR amplification of the V2 region of bacterial 16S rRNA genes, amplicon quantification, pooling, and pyrosequencing were performed as previously described . Sequences were submitted to the MG-RAST database (http://metagenomics.anl.gov/) under the study number qiime:814.
Post-processing of the pyrosequencing output was performed with the QIIME software package . First appropriate denoising of the 454 pyrosequencing output was performed using the PyroNoise algorithm . Then the UCLUST software (http://www.drive5.com/usearch/usearch.pdf) was used to pick clusters of operational taxonomic units (OTUs) at the 97% similarity level. The Ribosomal Database Project (RDP) classifier software , with the default training taxonomy, assigned taxonomic labels to the resulting OTUs. Finally, the OTUs were placed in a de novo phylogenetic tree with FastTree 2 .
A GLM (Generalized Linear Model; binomial, logit link function, dispersion estimated, Genstat version 13.2) was used to investigate differences in relative attractiveness between individuals, expressed as the fraction of mosquitoes caught in the trapping device baited with the glass beads releasing odour of the individual being tested divided by the total number of mosquitoes trapped in the two trapping devices together . The GLM was followed by a t-test to calculate pairwise differences between means. Individuals were classified as highly attractive (HA) when their mean relative attractiveness was significantly higher than the mean relative attractiveness of each individual in the group classified as poorly attractive (PA) which means the standard error of the GLM parameter estimate did not overlap. Effects were considered to be significant at p<0.05 .
Skin bacterial diversity.
The effect of average bacterial densities, expressed as the logarithm of the counts of CFUs on the selective and non-selective plates, on the relative attractiveness of the individuals was analysed using a GLM (Binomial, logit link function, dispersion estimated, Genstat version 13.2).
The 16S rRNA sequence results were used to determine the diversity of skin bacterial communities of the individuals. A phylogenetic diversity (PD) test , was used to compare the branch length of the parts of the phylogenetic tree covered by the samples. To control for sequencing effort, multiple rarefaction analyses  were performed on all samples at various sequencing depths (1,000 rarefaction samples at each sequencing depth from 500 to 3,000 at intervals of 500). A t-test was used (R programming environment, http://www.R-project.org) to test for a significant difference in microbial diversity between the PA and HA group at different sampling depths. No statistics were performed for the diversity at sequence depths above 1500 because the samples of some individuals did not contain more than 1500 sequences.
Based on previous results of in vitro experiments, an ANOVA (R programming environment) test was performed to determine whether the observed variance in the relative abundances of five specific genera  was partitioned according to the PA and HA group. The data were controlled for sequencing effort. A thousand stochastic rarefactions of the data were used at a simulated sequencing depth of 1500. An ANOVA was performed on each of those rarefied OTU-tables and the median P-values determined.
The relative abundances of the genera identified in the skin bacterial samples of the PA and HA individuals were analysed by multivariate partial least squares discriminant analysis (PLS-DA; SIMCA-P 12.0, Umetrics, Sweden) , . The data were log-transformed and scaled to unit variance. The number of significant PLS components was determined by cross-validation and the model validated by permutation testing , . To avoid over-parameterization of the model, genera with a minor contribution to the discrimination of the two groups in a first PLS-DA were excluded from the final model (variable influence on projection value<1 , ).
The relative attractiveness of the individuals to An. gambiae was significantly different (P<0.001; GLM). Nine out of the 48 individuals were significantly more attractive (Highly Attractive, HA group, Figure 1) than seven other individuals (Poorly Attractive, PA group, p<0.05; GLM, Figure 1).
Bars show the attractiveness parameter estimate results from the Generalized Linear Model (GLM) used to investigate the relative attractiveness  of each individual to An. gambiae. Individuals were classified as highly attractive (HA, blue bars) when their mean relative attractiveness was significantly higher than the mean relative attractiveness of each individual in the group classified as poorly attractive (PA, red bars) (GLM, p<0.05). Error bars represent the standard error of the mean from six replications.
Non-selective plate counts showed that, on average, 5.8×105 culturable bacteria were present per cm2 on the sole of a human foot, as determined by counting colony-forming units (CFUs) on blood agar plates. The average number of bacteria per cm2 on the sole of a foot positively correlated with the relative attractiveness of the individuals to An. gambiae (p = 0.003; GLM; Figure 2). The abundance of Staphylococcus spp. was also positively correlated with the relative attractiveness of the individuals to An. gambiae (p = 0.01; GLM; Figure S2). The number of bacteria per cm2 as determined by CFU counts on blood agar plates and the number of colonies per cm2 on plates selective for Staphylococcus spp. strongly correlated, suggesting that many of the bacterial colonies found on the blood agar plates were Staphylococcus spp. The abundance of Corynebacteria spp., Micrococcus spp. and Propionibacteria spp. did not show a correlation with the relative attractiveness of the individuals (p = 0.085, p = 0.28 and p = 0.41, respectively; GLM).
Correlation between the number of bacteria (log), determined by counts of colony forming units (CFUs) on non-selective plates and the relative attractiveness of the individuals. The relative attractiveness is expressed as the number of mosquitoes caught in the trapping device releasing the odour of the tested individual divided by the total number of mosquitoes trapped in both trapping devices . The red line indicates the fitted relationship according to the Generalized Linear Model (GLM).
The 16S rRNA sequence results showed that the phylogenetic diversity (PD) scores  of the HA individuals and the PA individuals were significantly different. PD-scores of communities on the PA individuals were significantly higher than those on the HA individuals at a simulated depth of 500 and 1000 sequences and marginally significant at a sequencing depth of 1500 sequences (p = 0.032, P = 0.043 and p = 0.057, respectively; t-test; Figure 3). At a sequence depth of 1000, the PD-scores of bacterial communities on the skin of PA individuals were 38% higher than the PD-scores of bacterial communities on the skin of HA individuals.
P-value for the difference in diversity score between PA (dashed red line) and HA (solid blue line) individuals is given at three sampling depths (not calculated for higher numbers of sequences, because the samples from some individuals did not yield more than 1500 sequences).
The relative abundances of OTUs identified by the pyrosequencing procedure and classified within the Staphylococcus and Pseudomonas genera were significantly different between HA and PA individuals (p = 0.024 and p = 0.005, respectively; t-test). The abundance of Staphylococcus spp. was 2.62 times higher in the HA group than in the PA group and the abundance of Pseudomonas spp. 3.11 times higher in the PA group than in the HA group. The abundance of Brevibacterium spp. and Corynebacterium spp. was not significantly different between the PA and HA group (p = 0.52 and p = 0.26, respectively; t-test). The effect of the abundance of Bacillus spp. on the attractiveness was not tested because they were present in only a limited number of samples.
Additional genera that correlated with human attractiveness to An. gambiae were identified by PLS-DA , , . The model differentiated the PA and HA groups based on the relative abundances of the bacterial genera (three significant latent variables, R2Xcum = 0.544, R2Ycum = 0.993, Q2cum = 0.823). PLS regression coefficients were determined to identify genera that were most characteristic for either group (Figure S3). Variovorax spp. and Pseudomonas spp. were significantly correlated with PA individuals based on their high PLS regression coefficients (95% confidence interval, Figure 4 and Figure S3). Leptotrichia spp., Delftia spp. and Actinobacteria Gp3 spp. were significantly correlated with HA individuals (Figure 4 and Figure S3).
Partial least squares discriminant analysis (PLS-DA) loading plot based on the relative abundance of bacterial genera in the microbiota profiles of poorly attractive (PA) and highly attractive (HA) individuals. Bacterial genera closer to HA or PA in the plot are more closely correlated to either HA or PA individuals. PLS 1 (R2X = 0.370, R2Y = 0.682, Q2 = 0.553) and PLS 2 (R2X = 0.102, R2Y = 0.260, Q2 = 0.424) are given. Genera that significantly contribute to the prediction of the model are indicated in blue (HA individuals) and red (PA individuals) (based on 95% confidence intervals, Figure S3). Some sequences could only be identified to division (D), class (C), order (O) or family (F).
In this study we demonstrated that the composition and abundance of the human skin microbiota influences the relative degree of attractiveness of a human to the malaria mosquito An. gambiae. Individuals with a higher microbial diversity were less attractive to the mosquitoes and several bacterial genera were identified that correlated with HA or PA individuals. Identification of the volatiles produced by these genera is likely to lead to the development of new mosquito attractants or repellents .
Only a small part of the bacteria found on the human skin are culturable  and therefore it was an important confirmative finding that the results from our in vivo study corroborated previous in vitro studies in which volatiles released by Staphylococcus epidermidis were attractive to An. gambiae females ,  and volatiles from Pseudomonas aeruginosa unattractive . The undirected search by PLS-DA in this study resulted in the identification of several new genera that are correlated with the relative degree of attractiveness of human beings to An. gambiae.
The correlation between Pseudomonas spp. and PA individuals is in accordance with in vitro experiments showing that the blend of compounds produced by P. aeruginosa is attractive to An. gambiae (in contrast to the volatiles produced by four other bacterial species, all commonly found on human skin) . Our results suggest that Pseudomonas spp. and possibly Variovorax spp. a) convert some of the attractive compounds produced by other bacteria, b) signal to other bacteria in ways that prevent them from emitting these attractive compounds, c) produce compounds that repel An. gambiae, or d) mask the effect of the attractive volatiles emanating from the human skin. More heterogeneous microbiotas may include more bacterial species that produce volatiles attenuating the attractiveness of PA individuals to mosquitoes, and may explain the interference effect described for the yellow fever mosquito Aedes aegypti (L). : higher levels of specific volatile compounds were found to be responsible for decreased attractiveness of individuals to Ae. aegypti. We hypothesize that lower attractiveness to mosquitoes is caused by a selective group of skin microbiota that emanates compounds that interfere with the attraction of mosquitoes to their human hosts and thus function as an in-built defence system . Genes of the Major Histocompatibility Complex (MHC) have been shown to influence body odour , ,  and may exert this influence by changing the skin microbiota composition and hence the volatiles produced by these bacteria and/or the human host , .
The current study shows that the skin microbiota could play an important role in this built-in defence system and may, therefore, affect transmission of malaria parasites , . Individuals with a higher microbial diversity and a higher abundance of Pseudomonas spp. or Variovorax spp. are less attractive to mosquitoes and may therefore receive fewer bites. Future studies should confirm if individuals with a specific microbiota composition run a lower risk of becoming infected with parasites, and consequently have a higher survival probability.
Compounds that inhibit microbial production of human odour , or manipulation of the composition of the skin microbiota may reduce a person's attractiveness to mosquitoes. Analysis of the bacterial volatiles attractive to mosquitoes produced by the Leptotrichia spp., Delftia spp. and Actinobacteria Gp3 spp. bacteria identified in this study  will also contribute to the development of attractants to be used in traps for monitoring malaria mosquito populations or lure-and-kill strategies .
The results presented in this study contribute to our fundamental understanding of the behavioural ecology of mosquitoes. The specialization of An. gambiae s.s. on human odours mediated by the composition of the human bacterial community may account for the high degree of anthropophily of An. gambiae s.s. Interestingly, the two closely-related sibling species An. quadriannulatus and An. arabiensis, have a wider host range that is more zoophilic , ,  or opportunistic , , respectively. Bacterial communities of other vertebrate species are likely to differ from those on human beings and may play an important role in determining the host range of mosquitoes .
Skin emanation collection. Teflon holder with six glass beads for collecting skin emanations from human feet to be used for mosquito attractiveness tests in the olfactometer. Distances are given in mm.
Staphylococcus spp.-selective plate counts and relative attractiveness to An. gambiae. Correlation between the number of Staphylococcus spp. bacteria (log), determined by counts of colony forming units (CFUs) on Staphylococcus spp. selective plates and the mean relative attractiveness of the individuals. The relative attractiveness is expressed as the number of mosquitoes caught in the trapping device releasing the odour of the tested individual divided by the total number of mosquitoes trapped in both trapping devices. The red line indicates the fitted relationship according to the Generalized Linear Model (GLM).
Coefficient plot of the bacterial profiles of poorly attractive (PA) and highly attractive (HA) individuals. Partial least squares-discriminant analysis (PLS-DA) coefficient plot based on the relative abundance of bacterial genera in the microbiota profiles of PA and HA individuals. Genera with significantly positive (>0) or negative (<0) PLS regression coefficients (i.e. no overlap between the 95% confidence interval indicated and the horizontal axis) contribute significantly to the prediction of the HA individuals (blue bars) or PA individuals (red bars), respectively. Coefficients were scaled and centred. Some sequences could only be identified to division (D), class (C), order (O) or family (F).
Ingredients of selective media used to determine the diversity of the human skin microbiota samples.
We thank all volunteers for their participation and F. K. M. van Aggelen, A. J. Gidding and L. Koopman for mosquito rearing. We acknowledge G. Leenders for the construction of the device for sampling skin emanations, H. Heilig for help with DNA extraction and S. Nakielny, M. Kos and J. Hageman for assistance with the analysis of the sequencing data.
Conceived and designed the experiments: NOV YTQ HB CM HJB JJAvL WT RCS. Performed the experiments: NOV YTQ HB DB-L CLL WV RCS. Analyzed the data: NOV YTQ HB CM DK SS DB-L CLL WV GWH RM FHJC RK RCS. Wrote the paper: NOV CM DK DB-L MD JJAvL WT RK RCS.
- 1. Khan AA, Maibach HI, Strauss WG (1968) The role of convection currents in mosquito attraction to human skin. Mosquito News 28: 462–464.
- 2. Cardé RT, Gibson G (2010) Host finding by female mosquitoes: mechanisms of orientation to host odours and other cues. In: Takken W, Knols BGJ, editors. Olfaction in Vector-Host Interactions. Wageningen: Wageningen Academic Publishers. pp. 115–140.
- 3. Allan SA, Day JF, Edman JD (1987) Visual ecology of biting flies. Annual Review of Entomology 32: 297–316.
- 4. Olanga E, Okal M, Mbadi P, Kokwaro E, Mukabana W (2010) Attraction of Anopheles gambiae to odour baits augmented with heat and moisture. Malaria Journal 9: 6.
- 5. Takken W, Knols BGJ (1999) Odor-mediated behavior of afrotropical malaria mosquitoes. Annual Review of Entomology 44: 131–157.
- 6. Knols BGJ, De Jong R, Takken W (1995) Differential attractiveness of isolated humans to mosquitoes in Tanzania. Transactions of the Royal Society of Tropical Medicine and Hygiene 89: 604–606.
- 7. Logan JG, Birkett MA, Clark SJ, Powers S, Seal NJ, et al. (2008) Identification of human-derived volatile chemicals that interfere with attraction of Aedes aegypti mosquitoes. Journal of Chemical Ecology 34: 308–322.
- 8. Bernier UR, Booth MM, Yost RA (1999) Analysis of human skin emanations by gas chromatography/mass spectrometry. 1.Thermal desorption of attractants for the yellow fever mosquito (Aedes aegypti) from handled glass beads. Analytical Chemistry 71: 1–7.
- 9. Qiu YT, Smallegange RC, van Loon JJA, Ter Braak CJF, Takken W (2006) Interindividual variation in the attractiveness of human odours to the malaria mosquito Anopheles gambiae s.s. Medical and Veterinary Entomology 20: 280–287.
- 10. Shelley WWB, Hurley HHJ Jr, Nichols AAC (1953) Axillary odor; experimental study of the role of bacteria, apocrine sweat, and deodorants. A M A Archives of Dermatology and Syphilology 68: 430–446.
- 11. Rennie PJ, Gower DB, Holland KT (1991) In-vitro and in-vivo studies of human axillary odor and the cutaneous microflora. British Journal of Dermatology 124: 596–602.
- 12. Taylor D, Daulby A, Grimshaw S, James G, Mercer J, et al. (2003) Characterization of the microflora of the human axilla. International Journal of Cosmetic Science 25: 137–145.
- 13. Ara K, Hama M, Akiba S, Koike K, Okisaka K, et al. (2006) Foot odor due to microbial metabolism and its control. Canadian Journal of Microbiology 52: 357–364.
- 14. Xu Y, Dixon S, Brereton R, Soini H, Novotny M, et al. (2007) Comparison of human axillary odour profiles obtained by gas chromatography/mass spectrometry and skin microbial profiles obtained by denaturing gradient gel electrophoresis using multivariate pattern recognition. Metabolomics 3: 427–437.
- 15. Tims S, van Wamel W, Endtz HP, van Belkum A, Kayser M (2010) Microbial DNA fingerprinting of human fingerprints: dynamic colonization of fingertip microflora challenges human host inferences for forensic purposes. International Journal of Legal Medicine 124: 477–481.
- 16. Verhulst NO, Takken W, Dicke M, Schraa G, Smallegange RC (2010) Chemical ecology of interactions between human skin microbiota and mosquitoes. FEMS Microbiology Ecology 74: 1–9.
- 17. Braks MAH, Takken W (1999) Incubated human sweat but not fresh sweat attracts the malaria mosquito Anopheles gambiae sensu stricto. Journal of Chemical Ecology 25: 663–672.
- 18. de Jong R, Knols BGJ (1995) Selection of biting sites on man by two malaria mosquito species. Experientia 51: 80–84.
- 19. Verhulst NO, Beijleveld H, Knols BGJ, Takken W, Schraa G, et al. (2009) Cultured skin microbiota attracts malaria mosquitoes. Malaria Journal 8: 302.
- 20. Verhulst NO, Andriessen R, Groenhagen U, Bukovinszkiné Kiss G, Schulz S, et al. (2010) Differential attraction of malaria mosquitoes to volatile blends produced by human skin bacteria. PLoS ONE 5: e15829.
- 21. Lefèvre T, Gouagna L-C, Dabiré KR, Elguero E, Fontenille D, et al. (2010) Beer consumption increases human attractiveness to malaria mosquitoes. PLoS ONE 5: e9546.
- 22. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, et al. (2009) Bacterial community variation in human body habitats across space and time. Science 326: 1694–1697.
- 23. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, et al. (2010) QIIME allows analysis of high-throughput community sequencing data. Nature Methods 7: 335–336.
- 24. Quince C, Lanzen A, Curtis TP, Davenport RJ, Hall N, et al. (2009) Accurate determination of microbial diversity from 454 pyrosequencing data. Nature Methods 6: 639–641.
- 25. Wang Q, Garrity GM, Tiedje JM, Cole JR (2007) Naive bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and Environmental Microbiology 73: 5261–5267.
- 26. Price MN, Dehal PS, Arkin AP (2010) FastTree 2 – approximately maximum-likelihood trees for large alignments. PLoS ONE 5: e9490.
- 27. Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biological Conservation 61: 1–10.
- 28. Gotelli NJ, Colwell RK (2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4: 379–391.
- 29. Eriksson L, Johansson E, Kettaneh-Wold N, Trygg J, Wikström C, et al. (2006) Multi- and megavariate data analysis. Umeå, Sweden: Umetrics AB. 425 p.
- 30. Bruinsma M, van Broekhoven S, Poelman E, Posthumus M, Müller M, et al. (2010) Inhibition of lipoxygenase affects induction of both direct and indirect plant defences against herbivorous insects. Oecologia 162: 393–404.
- 31. Pérez-Enciso M, Tenenhaus M (2003) Prediction of clinical outcome with microarray data: a partial least squares discriminant analysis (PLS-DA) approach. Human Genetics 112: 581–592.
- 32. Gao Z, Tseng C-h, Pei Z, Blaser MJ (2007) Molecular analysis of human forearm superficial skin bacterial biota. Proceedings of the National Academy of Sciences of the United States of America 104: 2927–2932.
- 33. Logan JG (2008) Why do mosquitoes “choose” to bite some people more than others? Outlooks on Pest Management 19: 280–283.
- 34. Yamazaki K, Boyse EA, Thaler MHT, Mathieson BJ, Abbott J, et al. (1976) Control of mating preferences in mice by genes in the major histocompatibility complex. Journal of Experimental Medicine 144: 1324–1335.
- 35. Wedekind C, Furi S (1997) Body odour preferences in men and women: do they aim for specific MHC combinations or simply heterozygosity? Proceedings of the Royal Society Biological Sciences Series B 264: 1471–1479.
- 36. Savelev SU, Antony-Babu S, Roberts SC, Wang H, Clare AS, et al. (2008) Individual variation in 3-methylbutanal: a putative link between human leukocyte antigen and skin microflora. Journal of Chemical Ecology 34: 1253–1257.
- 37. Penn D, Potts WK (1998) How do major histocompatibility complex genes influence odor and mating preferences? Advances In Immunology 69: 411–436.
- 38. Braks MAH, Anderson RA, Knols BGJ (1999) Infochemicals in mosquito host selection: Human skin microflora and Plasmodium parasites. Parasitology Today 15: 409–413.
- 39. Okumu FO, Govella NJ, Moore SJ, Chitnis N, Killeen GF (2010) Potential benefits, limitations and target product-profiles of odor-baited mosquito traps for malaria control in Africa. PLoS ONE 5: e11573.
- 40. White GB, Tessfaye F, Boreham PFL, Lemma G (1980) Malaria vector capacity of Anopheles arabiensis and An. quadriannulatus in Ethiopia: chromosomal interpretation after 6 years storage of field preparations. Transactions of the Royal Society of Tropical Medicine and Hygiene 74: 683–684.
- 41. Hunt RH, Coetzee M, Fettene M (1998) The Anopheles gambiae complex: a new species from Ethiopia. Transactions of the Royal Society of Tropical Medicine and Hygiene 92: 231–235.
- 42. Torr SJ, della Torre A, Calzetta M, Costantini C, Vale GA (2008) Towards a fuller understanding of mosquito behaviour: use of electrocuting grids to compare the odour-orientated responses of Anopheles arabiensis and An. quadriannulatus in the field. Medical and Veterinary Entomology 22: 93–108.
- 43. Costantini C, Sagnon N, della Torre A, Diallo M, Brady J, et al. (1998) Odor-mediated host preferences of West-African mosquitoes, with particular reference to malaria vectors. American Journal of Tropical Medicine and Hygiene 58: 56–63.
- 44. Costantini C, Gibson G, Sagnon N, della Torre A, Brady J, et al. (1996) Mosquito responses to carbon dioxide in a West African Sudan savanna village. Medical and Veterinary Entomology 10: 220–227.
- 45. Smallegange RC, Verhulst NO, Takken W (2011) Sweaty skin: an invitation to bite? Trends in Parasitology 27: 143–148.