Research ArticleLIFE SCIENCES

Bacteria: A novel source for potent mosquito feeding-deterrents

See allHide authors and affiliations

Science Advances  16 Jan 2019:
Vol. 5, no. 1, eaau6141
DOI: 10.1126/sciadv.aau6141


Antibiotic and insecticidal bioactivities of the extracellular secondary metabolites produced by entomopathogenic bacteria belonging to genus Xenorhabdus have been identified; however, their novel applications such as mosquito feeding-deterrence have not been reported. Here, we show that a mixture of compounds isolated from Xenorhabdus budapestensis in vitro cultures exhibits potent feeding-deterrent activity against three deadly mosquito vectors: Aedes aegypti, Anopheles gambiae, and Culex pipiens. We demonstrate that the deterrent active fraction isolated from replicate bacterial cultures is highly enriched in two compounds consistent with the previously described fabclavines, strongly suggesting that these are the molecular species responsible for feeding-deterrence. The mosquito feeding-deterrent activity in the putative fabclavine-rich fraction is comparable to or better than that of N,N-diethyl-3-methylbenzamide (also known as DEET) or picaridin in side-by-side assays. These findings lay the groundwork for research into biologically derived, peptide-based, low–molecular weight compounds isolated from bacteria for exploitation as mosquito repellents and feeding-deterrents.


Secondary or specialized metabolites (1) are a chemically diverse group of organic compounds produced by some microbes, plants, and animals (2). Generally considered nonessential for growth and development, secondary metabolites play other roles such as conferring protection against varied environmental risks. Many secondary metabolites of microbial or plant origin have been exploited for myriad applications in the pharmaceutical industry, including antibiotics, chemotherapeutic drugs, immunosuppressants, and other medicines (2). Secondary metabolites produced by Xenorhabdus, a group of bacteria that symbiotically associate with entomopathogenic nematodes, exhibit a range of antibiotic, antifungal, and insecticidal activities (39).

Genome studies (6, 10) suggest that there is an enormous range of chemical diversity and bioactivity that still remains to be explored and may lead to the discovery of novel bioactive compounds. In Xenorhabdus, genome mining has uncovered gene clusters predicted to participate in the synthesis of several compounds of unknown biological functions (11). The products encoded by these gene clusters exhibit diverse and structurally complex chemistries. One example produced by Xenorhabdus budapestensis (Xbu) is a unique class of hybrid compounds called fabclavines. These compounds were shown to exhibit a broad range of bioactivities, including antibiotic and insecticidal activities (3, 4, 11, 12). Uniquely, culture supernatants of two insect-killing bacteria, Xenorhabdus nematophila and Photorhabdus luminescens, also deterred feeding of ants, crickets, and wasps (13, 14). However, the bioactive compounds responsible for feeding-deterrence in those studies were not identified. Together, these studies suggested that secondary metabolites produced by Xenorhabdus might act as mosquito feeding- deterrents.

Repellents (topical and area) can provide protection from mosquito (and other blood-feeding insects) bites and thus from the disease agents that they transmit while feeding (15, 16). Mosquitoes transmit pathogens that cause devastating human diseases, including dengue, Chikungunya, West Nile, and Zika viral infections, that continue to affect millions of people worldwide. Limiting the impact of mosquito-borne diseases is an important goal for global public health agencies, and the use of mosquito repellents is one important tactic. Among U.S. Environmental Protection Agency–registered topical insect repellents, a majority of more than 500 products contain the active ingredient N,N-diethyl-3-methylbenzamide (also known as DEET;, which is the most widely used and effective repellent against mosquitoes and other disease vectors. Other commercially successful (17) insect repellents include IR3535 [insect repellent 3535, ethyl butylacetylaminoproprionate (EBAAP), a derivative of β-alanine (18)], (p)icaridin (and other piperidine derivatives) (19), and para-menthane-3,8-diol (distilled from Eucalyptus citriodora). Historically, research into biologically derived DEET alternatives has primarily focused on plant metabolites. Despite the exploitation of many genera for pharmaceutical exploration, bacteria have thus far remained unexplored in the search for insect repellent chemistries.

In a screen for repellent activities produced by Xenorhabdus, we observed that Xbu extracts deterred adult female Aedes, Anopheles, and Culex mosquitoes from feeding on an artificial diet in in vitro feeding experiments. The compounds exhibiting feeding-deterrent activity from bacterial cultures were concentrated via a three-step procedure, including reversed-phase chromatography, and analyzed by mass spectrometry (MS). In addition to other molecules, the feeding-deterrent active fraction consistently contained two highly abundant molecular species that we provisionally identify as members of the previously described class of molecules called fabclavines (11). Here, we present data on the enrichment and characterization of mosquito feeding-deterrent activity produced by Xenorhabdus and present evidence that fabclavines may be potent mosquito feeding-deterrents. Our discovery adds bacteria as a notable, potential source of novel mosquito feeding-deterrents and lays the groundwork for future exploration of bacterially derived secondary metabolites as feeding-deterrents for other pest insects.


In vitro membrane feeding system as a screening assay for determining mosquito feeding-deterrent activity in Xbu extracts

In the context of insect repellents, a deterrent is defined as “something that inhibits feeding or oviposition when present in a place where the insect would, in its absence, feed, rest or oviposit” (20). Here, we define compounds produced by Xbu as mosquito feeding-deterrents because, in the presence of Xbu compounds, female mosquitoes failed to feed on the food source described in this section.

Among methodologies to evaluate mosquito deterrents/repellents, in vitro assays offer important advantages. Because of the risk associated with accidental exposure to infectious agents during standard arm-in-cage assays (21, 22), and because of inconsistency in results among test subjects due to varying mosquito attraction to human subjects (23), researchers have developed artificial feeding systems. Blood is contained in a warming chamber and accessed by adult female mosquitoes through a natural or artificial membrane. Membrane feeders provide flexible systems that can be modified in terms of feeding solutions (2426), membrane types (24, 25, 27), and solution temperature (24, 25, 28). These systems eliminate the need to expose animals (29) or human volunteers, an especially important consideration when the compounds in the initial screening phases are of unknown toxicological or dermatological properties. Thus, we chose to modify our existing Hemotek membrane feeding system (Discovery Workshops, Accrington, UK) to screen feeding-deterrent compounds produced by Xbu bacteria. The main features of this system (Fig. 1) included (i) easy setup, (ii) minimal laboratory space requirements, (iii) a thermostat-regulated temperature source that eliminates the need for a water heater and circulation pump, and (iv) capacity for up to five screening assays at a time. The modifications used for the final bioassays included introduction of a loosely woven cotton cloth for application of the feeding-deterrent compounds, a collagen membrane, and use of a cocktail diet containing a red food dye (24) instead of blood (27).

Fig. 1 Description of the membrane feeding and feeding-deterrent screening system.

(A) Components of the feeding system, including (top/bottom panel, left to right) Hemotek temperature controller, feeder-housing assembly, metal feeder assembled with cocktail diet (red color) secured with collagen casing and O-ring, two layers of cotton cloth, metal feeder with cotton cloth secured via rubber bands (not visible), and metal feeder assembly secured to feeder housing. ID, inside diameter. Photo credit: Mayur Kumar Kajla, University of Wisconsin–Madison. (B and C) Results of the feeding assays. (B) Plot showing the number of A. aegypti mosquitoes that fed when water was applied to the cotton cloth as control, or DEET or picaridin (0.95 mg/cm2) [equivalent to 1.0% (v/v)] was applied in three replicate experiments. Each replicate consisted of 20 mosquitoes for each of the control as well as tests. (C) Representative image depicting appearance of fed (red abdomens) versus unfed Aedes mosquitoes resulting from the bioassay. Images show engorged abdomens and red dye in fed mosquitoes. Absence of both color and engorgement of abdomens indicates that no feeding occurred with 1% DEET (or picaridin) as a positive control.

We first tested this system with positive [DEET or picaridin at 0.95 mg/cm2 in water; equivalent to 1.0% (v/v)] and negative (water) controls applied to cloth covering the membrane feeder. Mosquitoes were allowed to feed for 30 min and then frozen at −20°C before scoring and counting. Mosquito feeding success was then scored by counting fed and unfed mosquitoes following bioassays in the presence or absence of the repellent compounds. The outcomes of the screening assay are shown in Fig. 1 (B and C). Using this system, we obtained reproducible and consistently high mosquito feeding success when the cotton cloth was treated with water and 0% feeding when DEET or picaridin was tested as positive repellent controls. Results of three replicate experiments (20 mosquitoes per replicate; total number of mosquitoes tested in three replicate experiments = 60) are presented in Fig. 1B and show an average feeding success of 96.67%, with water controls with Aedes aegypti, Anopheles gambiae, and Culex pipiens also fed well (75 to 80%; Fig. 2), indicating that the bioassay can be used with multiple mosquito species. On the basis of these results, we determined that the bioassay provided a robust and reproducible test arena to screen mosquito feeding-deterrent activities in the Xbu extracts at various stages of purification.

Fig. 2 Mosquito feeding-deterrent activity of Xbu Peak#3 with C. pipiens, A. gambiae, and A. aegypti.

(A) Plots show feeding-deterrent activity of Xbu compounds tested at 0.057 mg/cm2. Data from three replicate experiments are shown. Each replicate consisted of 20 mosquitoes for each of the control (water only) as well as tests (Xbu Peak#3), respectively. A. aegypti mosquitoes were included for comparison. (B) Appearance of fed (red abdomens) and unfed Anopheles and Culex mosquitoes. In this experiment, one-layer muslin cloth was used. Fisher’s exact test was used to assess differences between groups using Stata statistical software.

In addition, this bioassay was optimal to screen minimal amounts of compounds, a major concern when working with naturally produced microbial compounds not easily obtainable in large quantities. The method also provided rapid results as well as consistent measures of feeding-deterrence that were reproducible across assay dates.

Enrichment and characterization of mosquito feeding-deterrent active compounds from Xbu

Next, we developed a procedure to isolate mosquito feeding-deterrent active compounds from Xbu cultures. Xenorhabdus bacteria are known to produce antibiotics and secondary metabolites in the late stationary phase of their growth cycle (11, 30, 31). Accordingly, we used 72-hour bacterial cultures to harvest mosquito feeding-deterrent compounds from the cell-free culture supernatants. Mosquito feeding-deterrent active compounds in the culture supernatants were concentrated via acetone precipitation. In the next step, water-soluble acetone precipitates yielded a broad peak detected at 280 nm on a reversed-phase C18 flash chromatography column, indicating that the mosquito feeding-deterrent active compounds coeluted with other compounds detectible at 280 nm. The feeding-deterrent activity concentrated and eluted as a single broad peak fraction. Representative images of C18 flash purification results are shown in fig. S1. The peak fraction was subjected to MS analysis for determination of molecular masses and structural characterization. Matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) MS revealed the presence of several abundant molecular species including compounds at m/z (mass/charge ratio) 1302.94, 1312.96, and 1430.07 in this fraction (Fig. 3A), with additional masses at m/z 44.03 units higher, leading to the idea that these were likely related. The difference of 44 atomic mass units is consistent with addition of one C2H4O moiety (11).

Fig. 3 MALDI-TOF spectrum of flash and HPLC C18 reversed-phase chromatography separated mosquito feeding-deterrent active peak fraction.

(A) MALDI-TOF analysis of the C18 flash fractionated feeding-deterrent active fraction yielded several major masses (and related molecular species) at m/z 1302.94 (1346.97), 1312.96 (1356.98), and 1430.07 (1474.10), as well as m/z 1238.953. The mass difference of m/z 44 in these pairs may be due to addition of C2H4O moieties (11). (B) MALDI-TOF analysis of HPLC-separated feeding-deterrent active Xbu Peak#3 shows enrichment of the same two abundant (and related) molecular species at m/z 1302.92 and 1346.95.

The mosquito feeding-deterrent active fraction from the C18 flash chromatography was subjected to a second fractionation step using analytical reversed-phase high-performance liquid chromatography (HPLC) after an ultrafiltration step (see Materials and Methods for details). Four major peaks were observed, with most of the repellent activity concentrated in peak number 3 (hereafter referred as Xbu Peak#3), which eluted at ~24% acetonitrile (ACN). Representative images of HPLC C18 purification are shown in fig. S2. None of the other peaks showed feeding-deterrent activity (fig. S3, A and B) except peak 4. The activity in peak 4 is likely due to cross-contamination from peak 3 as a result of incomplete resolution of these peaks. MS analysis of Xbu Peak#3 showed two highly abundant masses at m/z 1302.9 and 1346.9 (Fig. 3B), indicating that these compounds, previously observed in the active fraction from C18 flash chromatography, had been enriched in the mosquito feeding-deterrent fraction, as well as many other lower-abundance molecular species. MALDI-TOF MS on mosquito feeding-deterrent active Peak#3 was performed on three different batches of the fractionated compounds. In each replicate, these two masses were the most abundant ions detected. Spectra for these experiments are presented in fig. S4. A number of other lower-abundance molecular species are observed in each replicate fractionation of Xbu cultures (see fig. S4). Several of these compounds do not reproduce across all fractionations, eliminating them from consideration as the active component. Others, most notably at m/z 1430 and 1474, are observed in all of the active fractions. However, their relative enrichment in the active HPLC fraction is decreased compared to m/z 1302 and 1346. This can be seen when comparing the signal intensities for m/z 1302, 1346, 1430, and 1474 in the C18 flash fraction before and after HPLC (Fig. 3, A and B, and fig. S4). The decrease in m/z 1430 and 1474 relative to m/z 1302 and 1346 indicates only a partial coelution with the activity.

Because we were interested in identifying the compounds, we subjected the mosquito feeding-deterrent active fraction to MS/MS analysis by unassisted nanospray on an Orbitrap mass spectrometer. Nanospray MS displayed the same major molecular species observed by MALDI-TOF MS, with each compound in a dominant charge state of 2, as well as other lower-abundance species. Charge states from 1 to 4 were detected (fig. S5). MS/MS was performed on the doubly charged species at m/z 651.9 and 673.9 using both collision-induced dissociation (CID) and high-energy collisional dissociation (HCD) fragmentations. The MS/MS spectra revealed that these compounds were structurally very similar, with several ions common to both compounds, and many ions showing a difference of 44 Da between them, as previously observed by MALDI-TOF MS. The MS/MS spectra and the inferred fragment ion assignments show excellent agreement with the recently described “fabclavines” isolated from a similar strain of Xbu (11). Using the naming convention described in that work and the structures for fabclavines Ib and IIb, the observed fragment ions were consistent with assignment of m/z 1302 as fabclavine IIb and m/z 1346 as fabclavine Ib (fig. S6, A and B), suggesting that the two major constituents of the active fraction are fabclavines.

Results of N-terminal Edman degradation analyses on mosquito feeding-deterrent active Peak#3 that contains two related peptides (provided in table S2) were inconclusive in determining the sequence of amino acids, possibly explained by an inaccessible N terminus of the molecule due to macrocyclization (11). Total amino acids were measured from mosquito feeding-deterrent active Peak#3 by two different strong cation ion exchange chromatography methods, sodium- and lithium-based elution systems (Molecular Structure Facility, University of California, Davis, CA, USA). Both systems resulted in significant signals for amino acids Asx and His (fig. S7, A and B). In addition, two unidentified signals were observed, which could be 2,3-diaminobutyric acid (elution profiles were consistent with this compound). Together, these data indicate that two of the amino acids Asx and His and possibly 2,3-diaminobutyric acid are likely to be present in the feeding-deterrent active fraction—amino acids that are also described in fabclavines (11), which, in addition to these, also contain phenylalanine or histidine, and a modified proline residue.

Mosquito feeding-deterrent activity comparison between Xbu compounds, DEET, and picaridin

Next, we determined the mosquito feeding-deterrence dose of the compounds produced by Xbu with Aedes aegypti mosquitoes. For this, we tested a range of concentrations of the deterrent active fraction and compared their feeding-deterrent activities to that of DEET and picaridin. Table S1 lists the mosquito fed/unfed count data used for estimating feeding-deterrence. Table 1 shows feeding-deterrence dose (expressed in mg/cm2) of the compound that resulted in 50 or 90% reduction in mosquito feeding.

Table 1 Determination of feeding-deterrence dose (50 and 90%) of DEET, picaridin, and Xbu compounds against A. aegypti.

RE90 is the relative efficacy of Xbu Peak#3 derived as a ratio of the dose causing 90% reduction in mosquito feeding of DEET or picaridin to that of Xbu Peak#3.

View this table:

A dose resulting in 50% reduction in feeding among Xbu Peak#3 fraction and DEET was similar (Xbu Peak#3 = 0.014 and DEET = 0.012 mg/cm2), whereas it was 6.14× lower than that of picaridin (0.086 mg/cm2). A dose resulting in 90% reduction in feeding among Xbu Peak#3 was much lower compared to both DEET and picaridin (Xbu Peak#3 = 0.057, DEET = 0.178, and picaridin = 0.471 mg/cm2), indicating that a much lower concentration of Xbu compounds is required to achieve 90% reduction in feeding as compared to DEET or picaridin. The relative efficacy (24) (RE90) of Xbu Peak#3 derived as a ratio of the dose resulting in 90% feeding reduction of DEET or picaridin to that of Xbu Peak#3 was 3.12 and 8.26, respectively, suggesting that mosquito feeding-deterrence dose of Xbu compounds is more potent in inhibiting mosquitoes from feeding on the cocktail diet under the conditions of the screening assay as compared to DEET or picaridin. Table 2 shows the Probit-analyzed comparison among the least-squares estimates (32) for each compound’s effect. The repellency comparison based on adjusted P values among DEET and picaridin was significant (P < 0.05); DEET and Xbu Peak#3 were not significantly different (P > 0.05); and Xbu Peak#3 was significantly different from picaridin (P < 0.05). Together, feeding-deterrent activity of the bacterial compounds is equal to or better than DEET and picaridin.

Table 2 Comparison of mosquito feeding-deterrent activity of DEET, picaridin, and Xbu Peak#3 against A. aegypti.

Table shows the Probit-analyzed comparison among the least-squares estimates for each compound’s effect (32). Feeding-deterrence comparison based on adjusted P values among DEET and picaridin was significant (adjusted P < 0.05); DEET and Xbu Peak#3 were not significant (adjusted P > 0.05); and Xbu Peak#3 was significantly different from picaridin (adjusted P < 0.05).

View this table:

While the exact mechanism by which Xbu compounds exert mosquito feeding-deterrence is a subject for future in-depth investigations, visual observations during feeding indicated that landing and probing by Aedes mosquitoes on the cloth treated with Xbu compounds were dose dependent (movie S1). At a 50% feeding-deterrence dose, a majority of mosquitoes landed and about half succeeded in feeding, while at 90% or a higher dose, few mosquitoes landed and none fed. Movie S1 shows the behavior of mosquitoes in the presence of water or Xbu compounds at 0.057 mg/cm2 (the concentration that caused 90% feeding-deterrence). The mosquitoes that landed attempted to probe through the cloth and were seen cleaning their proboscises soon after probing but did not imbibe food.

Thus, mosquito contact with the treated surface preceded feeding-deterrence at low dose. Deciphering feeding-deterrence by olfactory or gustatory mechanisms against low-volatility compounds, including DEET and picaridin (33), is difficult due to overlapping sensory responses (34). Addition of DEET to blood (35) deterred mosquitoes from feeding through direct contact with the blood meal, likely via gustatory response. External gustatory sensilla on antennae of the Chagas vector Rhodnius prolixus mediate feeding-deterrence to quinine or caffeine when applied to mesh covering feeding solutions (36). Certain fatty acids exert feeding-deterrence in A. aegypti when applied to cloth surfaces covering the artificial diet (37). Feeding-deterrence exhibited by Xbu compounds might elicit a combination of olfactory and gustatory responses upon contact by mosquitoes with the treated cloth surfaces in a dose-dependent manner.

We also observed feeding-deterrence exhibited by Xbu compounds with A. gambiae and C. pipiens. For tests with these mosquitoes, we modified our screening assay by incorporating a single layer of a muslin cloth (rather than two layers as before). These mosquitoes had difficulty feeding with the two layers of cotton cloth described for Aedes. One layer of muslin yielded 75 to 80% feeding rate (Fig. 2). Feeding assays performed with 0.057 mg/cm2 (the concentration that caused 90% feeding-deterrence with Aedes) for A. gambiae and C. pipiens are shown in Fig. 2. Results indicate that the Xbu compounds exert a potent feeding-deterrent activity against A. gambiae and C. pipiens, as was seen with Aedes. Aedes mosquitoes were also included in this trial using the single layer of cloth.


The bacteria associated with entomopathogenic nematodes offer an enormous potential for novel bioactivities especially attributable to low–molecular weight natural products. Genome mining projects continue to uncover information on biosynthetic gene clusters responsible for production of several of the bacterial secondary metabolites awaiting identification and assignment to their natural biological functions (1).

Here, we provide evidence that a potent mosquito feeding-deterrent activity could be concentrated from Xbu cultures. MS and MS/MS data support the putative identification of the two most abundant compounds enriched in the deterrent active fraction as fabclavines, strongly suggesting that the feeding-deterrent activity is likely exhibited by these compounds. The feeding-deterrent activity, as determined by membrane mosquito feeding experiments, is comparable to or higher than the gold standard repellent DEET or picaridin against A. aegypti. Furthermore, the feeding rate of Anopheles and Culex mosquitoes decreased significantly in the presence of Xbu compounds, suggesting that these compounds are effective in inhibiting a broader range of mosquito species. These results suggest that the bioactivities of the fabclavine class of natural products may be exploited in novel applications such as mosquito feeding-deterrence. The authors recognize that the mosquito feeding-deterrent active fraction is a mixture of compounds, dominated by two copurifying, closely related molecular species at m/z 1302 and 1346. Future studies should (i) explore chromatographic or otherwise resolution of all compounds in these active fractions to identify those specifically responsible for deterrence activity, (ii) determine whether the compounds can affect feeding-deterrence individually or require application as a mixture, (iii) demonstrate deterrent activity of fabclavines via de novo synthesis or knockout experiments, and (iv) determine efficacy and toxicological properties of the deterrent active compounds for use in mosquito deterrent/repellent formulations for eventual testing on human skin.


Ethics statement

No live animals or human subjects were used in mosquito-deterrent screening assays in this study.

Rearing and maintenance of mosquitoes

Colonies of A. aegypti (Liverpool strain), A. gambiae (G3), and C. pipiens (Iowa strain) were maintained at the University of Wisconsin according to standard procedures reported previously (3841). Briefly, mosquito colonies were maintained at 26.5 ± 0.5°C and 80 ± 5% relative humidity. Larvae were fed TetraMin fish food. Adult mosquitoes were maintained on a constant exposure to 10% sucrose presented through cotton balls. For egg production, adult female mosquitoes were offered defibrinated rabbit blood (HemoStat Laboratories, CA, USA) via Hemotek membrane feeding system (Discovery Workshops, Accrington, UK) described below. Edible collagen casing membrane (Nippi Edible Collagen Casing, ViskoTeepak, WI, USA) was selected as a membrane of choice for blood feeding and feeding-deterrent screening, as all three species of mosquitoes fed readily through it. For feeding-deterrent screening assays, 20 nulliparous, mated, 7- to 10-day-old adult female mosquitoes, hatched from the same batch of eggs, were separated in screened paper containers (height × width = 8.2 cm × 8.5 cm; Neptune Paper Products, NJ, USA). Separated mosquitoes were starved for 12 to 16 hours before the screening assays.

Mosquito feeding-deterrent screening assay

Mosquitoes were exposed to Xbu feeding-deterrent compounds or test repellents for 30 min at room temperature (incandescent light, 25 to 26°C) between 10:00 a.m. and 4:00 p.m. (U.S. Central Time) using a modified membrane-feeding apparatus described in Fig. 1.

Briefly, the Hemotek temperature controller was set to a constant temperature of 37°C 5 to 10 min ahead of the assay. A 2.5 cm × 2.5 cm piece of a fresh, pre-cut, and thoroughly water-washed collagen casing membrane was secured to the metal feeder using an O-ring. Approximately 2.5 ml of the cocktail diet containing 2% (v/v) red food dye (McCormick & Company Inc., MD, USA) was introduced from the opening at the back of the metal feeder. A cotton cloth was placed on top of the collagen casing and secured by rubber bands. Test compound or water was then applied by immersing the feeder assembly in 1 ml of the respective test solution, and the metal feeder assembly was then secured to the feeder housing and exposed to mosquitoes housed in a screened container. Several types of cloth materials varying in thickness and texture/thread count were evaluated with an objective of finding the one that provided optimal mosquito feeding success. Eventually, a double-layer cotton cloth obtained from JOANN Fabrics (Madison, WI) was used with A. aegypti. The features and dimensions of the chosen cotton cloth were as follows: thread count = 14.9 × 17.0 threads in 1 cm2 (average of 10 measurements), diameter of the circular cloth applied to metal feeder = 3.67 cm (see Fig. 1A), and total area of the circular cloth area exposed to mosquitoes = ~10.57 cm2. However, for A. gambiae and C. pipiens, one layer of muslin cloth (thread count = 24.3 × 33.9 threads in 1 cm2; average of 10 measurements) was used, as this cloth worked well with these mosquitoes. Both types of cloth materials are shown in fig. S8.

Because the Xbu compounds were dissolved in water for testing, DEET and picaridin were also prepared in 0.2-μm filtered double-distilled water. DEET was diluted from SC Johnson’s OFF! Deep Woods containing 25% DEET and picaridin from Clean Insect Repellent (purchased from Walgreens, USA) containing 7% picaridin. One percent stock solutions of DEET and picaridin (equivalent to 10 mg/ml) were dissolved in water from which a range of dilutions was prepared for determination of feeding-deterrence dose. Both of these repellents were tested at a concentration range of 1 to 0.01% (v/v) corresponding to 0.95 to 0.0095 mg/cm2. Both compounds dissolved well in water at the tested concentrations. DEET and picaridin were tested on the same day with a 4-hour gap in between tests. The HPLC-separated, lyophilized bacterial sample was dissolved in water and filtered through a 0.2-μm filter. For an accurate assessment, protein content in this sample was determined via two methods: (i) bicinchoninic acid (BCA) assay according to the manufacturer’s instructions (Pierce BCA Protein Assay Kit, Thermo Fisher Scientific, USA) and (ii) total amino acid analysis (Molecular Structure Facility, University of California, Davis, CA). The difference in protein content determined by these two methods was negligible. Thus, for routine measurements, the BCA method was used. A dosage range from 0.73 to 0.048 mg/ml (v/v) corresponding to 69.5 to 4.57 μg/cm2 of the mosquito feeding-deterrent active compounds was tested with A. aegypti. Testing with A. gambiae and C. pipiens was done only with one concentration of Xbu compounds that yielded a feeding-deterrence dose of 90% with A. aegypti [i.e., 0.060% (v/v), corresponding to 0.057 mg/cm2; Table 1]. Three replicate feeding experiments were conducted for each compound with mosquitoes hatched from different egg batches (to account for cohort bias) over a period of 3 weeks. One replicate consisted of 20 mosquitoes tested on the same day with each concentration of the dilution range. For example, one replication of samples with a concentration range of 69.5 to 4.57 μg/cm2 was tested on the same day. Extensive washing of the metal feeders, replenishing of collagen membrane, cotton cloths, and feeding solution for each exposure test and time gaps between assays assured that there was no carryover of the compounds or contamination of the feeders in between assays. As an extra precaution, feeding-deterrence assays with bacterial compounds were purposely performed on a different day than assays with DEET and picaridin. After the completion of the screening assay (30 min), mosquitoes were killed by freezing at −20°C. Fed versus unfed mosquitoes were counted under a dissecting microscope and verified by at least two people. Fed mosquitoes could be easily distinguished as they were engorged and had red abdomens, which were clearly visible even to an unaided eye (fully fed or partially fed), while unfed mosquitos were lean and did not have red abdomens. Mosquitoes that were not engorged and had no red color were considered as unfed and hence deterred (Figs. 1C and 2B). This distinction provided a reliable, quantitative means of assessing the feeding success of mosquitoes with different compounds and across assays and allowed a robust comparison. Count data (proportion of unfed mosquitoes from a total) were used for statistical analysis. Several tests with Xbu Peak#3 from different batches exhibited consistent, concentration-dependent feeding-deterrent activity. Here, results of three replications are presented. Figures 1B and 2A and Tables 1 and 2 present results of all of the feeding-deterrence screening experiments.

Isolation, enrichment, and characterization of mosquito feeding-deterrent active compounds produced by Xbu

Xbu bacteria (42) obtained from H. Goodrich-Blair (Department of Bacteriology, University of Wisconsin–Madison, USA) were maintained as glycerol stocks at −80°C. For culturing, the bacteria were freshly streaked onto nutrient bromothymol blue triphenyltetrazolium chloride agar (NBTA) plates (42). A single, isolated blue colony (image of bacteria streaked on LB plate is seen in fig. S9) was taken for further growth in modified minimal medium containing 0.05 M Na2HPO4, 0.05 M KH2PO4, 0.02 M (NH4)2SO4, 0.001 M MgSO4, 0.25% yeast extract, and 0.1 M glucose supplemented with 1% yeast extract. A single bacterial colony was resuspended in 1.5-ml medium from which 200 μl was inoculated in a flask containing 400-ml medium. Bacterial cultures (typically six flasks of 400 ml) were grown at 30°C at 120 rpm in a rotary shaker to stationary phase and harvested at 72 hours after inoculation via centrifugation (3913g, Beckman rotor JA14 at 4°C).

The chilled cell-free culture supernatant was mixed with two volumes of ice-cold acetone and incubated cold for 12 to 16 hours while stirring. After precipitation, spent medium/acetone was discarded and precipitates containing the mosquito feeding-deterrent active compounds were dissolved in ultrapure water so as to concentrate these to about 10× of the original culture volume (i.e., 240 ml of water). This solution was first centrifuged at 3578g at 4°C (Beckman rotor JA20) and then filtered through 0.45-μm filter and loaded on a C18 flash reversed-phase column (Reveleris Flash Cartridge, 12 g, BUCHI Corporation, DE, USA) using a peristaltic pump. The column was then connected to a fast protein liquid chromatography (FPLC; Akta Prime Plus, GE Healthcare Bio-Sciences, PA, USA). Solvents for reversed-phase columns were (i) 0.2-μm filtered double-distilled water and (ii) HPLC-grade, 0.2-μm filtered ACN (Fisher Scientific, USA). Both solvents contained 0.1% (v/v) trifluoroacetic acid (TFA).

An ACN gradient of 50 to 100% was used to collect the mosquito feeding-deterrent active peak fraction on C18 flash column via FPLC. The elution pattern of the mosquito feeding-deterrent active peak was consistent and reproducible between different batches of cultures/purifications. Representative results of this step are shown in fig. S1. The active fraction from the C18 flash chromatography was filtered through a 5-kDa molecular weight cutoff ultrafiltration device (Sartorius, Fisher Scientific, USA), and the flow through was subjected to further purification using HPLC on an analytical reversed-phase C18 column [Vydac C18 (5 μm, 4.6 mm inside diameter × 150 mm), catalog no. 218TP5415, Fisher Scientific, USA]. A gradient of 13 to 40% ACN over 40 min yielded four major peaks detectable at 214 nm, with most of the repellent activity concentrated in Peak#3, which eluted at an ACN concentration of ~24% and a retention time of ~14:00 min. Representative images of HPLC C18 purification are shown in fig. S2, with mosquito feeding-deterrent active Xbu Peak#3 indicated. Other peaks (except peak number 4) had no feeding-deterrent activity (fig. S3). Several HPLC runs were conducted to generate material for feeding-deterrent activity analysis.

MS was performed in the Mass Spectrometry Facility at the University of Wisconsin–Madison Biotechnology Center. MALDI-TOF spectra of the active, C18 flash chromatography fraction and the subsequent HPLC-separated active fractions were acquired on a SCIEX 4800 TOF-TOF mass spectrometer (SCIEX, MA, USA). Samples (0.5 μl) were spotted onto a 384-well Opti-TOF insert and 0.5-μl α-cyano-4-hydroxycinnamic acid (Fluka, Switzerland) at 6 mg/ml in 75% ACN, 0.1% TFA was added, and the sample and matrix were mixed on a plate shaker. The spot was air-dried, and data were acquired in positive ion reflector mode over the m/z range of 700 to 4000 using a laser intensity of 2800 and averaging 1000 shots per spectrum. Mass assignment was performed using external calibration, with calibration taking place immediately before data collection. Electrospray MS and MS/MS were performed using unassisted nanospray by loading the mosquito feeding-deterrent active fraction into pulled, coated borosilicate tips (New Objective, MA, USA) and acquiring positive ion, profile mode spectra on an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, CA, USA). MS and MS/MS data were collected in the Orbitrap analyzer at a resolving power setting of 120,000. MS data were acquired over the m/z range of 150 to 1800, while HCD MS/MS data were collected from m/z 100 to 1500 and CID MS/MS spectra were collected from m/z 180 to 1500. CID and HCD fragmentations used 35% normalized collision energy.

Statistical analysis

To estimate feeding-deterrence dose, we used Probit analysis on the log10-transformed feeding counts data using Statistical Analysis Software, version 9.4 (SAS Institute Inc., NC, USA). The feeding response variable obtained from mosquito counts data followed a binomial rather than a normal distribution. For this kind of nonnormal dataset, a generalized linear model (GLM) is appropriate. The Probit model is one kind of GLM that is used for analyses related to a dose-dependent variable. Data were log10-transformed because of the wide range of concentration values (0.01 to 1) investigated. Feeding-deterrent activity (Tables 1 and 2) was expressed as a dose (in mg/cm2) of the compound (bacterial or DEET or picaridin) applied to cloth that resulted in a 50 or 90% inhibition in mosquito feeding rate similarly as described previously (24). Relative efficacy (RE90) of the deterrent activity was derived by dividing a dose that results in 90% feeding-deterrence of DEET to that of the picaridin and Xbu Peak#3, respectively. Least-squares estimation (32) was conducted to compare differences among feeding-deterrent activity between the three compounds based on adjusted P values (Table 2). P values in Fig. 2 were calculated using Fisher’s exact test using Stata statistical software (Stata Statistical Software: Release 15, StataCorp LLC, College Station, TX).


Supplementary material for this article is available at

Fig. S1. Enrichment of the mosquito feeding-deterrent active compounds via flash chromatography on reversed-phase C18 column.

Fig. S2. Enrichment of the mosquito feeding-deterrent active compounds via HPLC on an analytical reversed-phase C18 column.

Fig. S3. Elution profile of feeding-deterrent active fraction.

Fig. S4. MALDI-TOF MS analysis of HPLC-enriched feeding-deterrent Peak#3.

Fig. S5. Electrospray ionization–Orbitrap MS analysis of the HPLC-enriched deterrent active Xbu Peak#3 showing observed charged states.

Fig. S6. MS/MS spectra of doubly-charged molecular species at m/z 651.9 and 673.9.

Fig. S7. Amino acid analyses of purified feeding-deterrent fraction.

Fig. S8. Appearance of cotton cloths used in the Hemotek-based membrane-feeding assays.

Fig. S9. Typical example of Xbu streaked on LBTA (Luria broth agar supplemented with indicator stains bromothymol blue and triphenyltetrazolium chloride) plate.

Table S1. Mosquito feeding data used for Probit analysis to generate Tables 1 and 2.

Table S2. Results of the N-terminal Edman sequencing trial.

Movie S1. Video clip showing feeding response of mosquitoes with water control and at a 90% feeding-deterrence dose of Xbu Peak#3.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgements: We thank H. Goodrich-Blair and A. Torres for sharing the strain of Xbu; Il-H. Kim for discussions on bacterial cultures; L. C. Bartholomay for providing mosquitoes; Bartholomay and R. Weyker for suggestions on collagen casing membrane; E. Norris for suggestions on repellent assays; undergraduates N. Wong, S. Muehlfeld, and M. Dyhr for help with mosquito colony maintenance; W. Goodman and R. Frederick for consultation on compound purification; and Y. Lee, J. Zhu, and C. P. Yadav (NIMR, Delhi, India) for support with statistical analyses. M.K.K. thanks DBT, Government of India, for Ramalingaswami Re-entry Fellowship and for allowing editing of manuscript during final stages, and N. Valecha and R. Dixit (NIMR, Delhi, India) for logistical help. Funding: This work was supported by NIH R21 grant no. AI123719 to S.M.P. Author contributions: S.M.P. conceived the study, provided guidance on study design, and edited the manuscript. M.K.K. designed and performed the experiments, collected and analyzed the data, and wrote the manuscript. G.A.B.-W. collected all MS data and annotated the MS/MS spectra. All authors edited, reviewed, and approved the manuscript. Competing interests: M.K.K. and S.M.P. are inventors on a patent application related to this work filed by the Wisconsin Alumni Research Foundation (no. 62/579,275, filed on 31 October 2017). The authors declare that they have no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. Raw MS data referenced in this publication have been deposited in the MassIVE data repository ( under accession MSV000083089, with digital object identifier doi:10.25345/C5VS3B. These data will be made public and available for download upon publication of this work.

Stay Connected to Science Advances

Navigate This Article