Genes and odors underlying the recent evolution of mosquito preference for humans

Abstract

Mosquito species that specialize in biting humans are few but dangerous. They include the African malaria vectors Anopheles gambiae and Anopheles coluzzii, as well as the cosmopolitan dengue, chikungunya, and yellow fever vector Aedes aegypti. These mosquitoes have evolved a remarkable innate preference for human odor that helps them find and bite us. Here I review what is known about this important evolutionary adaptation from its historical documentation to its chemical and molecular basis.

Introduction

The majority of the three thousand plus mosquito species that exist worldwide are opportunistic [1]. They tend to bite the vertebrate animals that are most readily available in their environment. Limits exist, of course, but mosquitoes usually restrict themselves to broad taxonomic groups such as mammals or birds rather than narrow host species, genera, or families.

It is therefore exceptional that a handful of mosquito species have evolved to specialize in biting humans. By no coincidence, these human specialists are also some of the most efficient and deadly vectors of human disease. Anopheles gambiae and Anopheles coluzzii are the primary vectors of malaria in Africa, while Aedes aegypti is the primary vector of dengue, chikungunya, and yellow fever worldwide. Together they account for hundreds of thousands of deaths and hundreds of millions of non-lethal cases of disease each year [2].

The enormous impact of these vector species on humans has made them the subject of intense study since the early 1900s. Initial comparisons with close relatives that do not specialize highlighted the fascinating array of adaptations that help mosquitoes find and bite humans [3, 4]. More recent studies have begun to dissect the chemical and molecular basis of one such adaptation in particular – a strong innate preference for human odor.

Parallel evolution of preference for human odor in Aedes and Anopheles

Aedes and Anopheles mosquitoes are only distantly related, last sharing a common ancestor 150–200 million years ago [5]. Nevertheless, both genera include species that have independently evolved from opportunists into human-biting specialists, presumably since humans began forming dense, stable communities 10–15 thousand years ago [6]. One of the key behaviors that helps these specialists selectively target humans is their strong preference for human odor over that of other animals. This preference is displayed by females, who need the nutrients from a blood meal to synthesize eggs, and can most clearly be seen by comparing the specialists with their close opportunistic relatives.

For example, Aedes aegypti comprises two ecologically distinct subspecies (Figure 1A). The derived Ae. aegypti aegypti is the human specialist, while the ancestral Ae. aegypti formosus is ecologically variable and often still found living in forests and biting a variety of mammals and even reptiles [3, 7]. Analyses of blood found in the gut of field collected females reveal their distinct biting habits (Figure 1A). Biting, however, is heavily influenced by host availability and thus is not a reliable indicator of innate preference. Innate responses to host odor can be measured in the field using odor-baited entry traps or in the laboratory using wind tunnels and olfactometers. Initial olfactometer studies from the 1970s documented striking divergence between the subspecies where they co-occur along the coast of East Africa, with Ae. aegypti aegypti showing much stronger attraction to, and preference for, human odor (Figure 1B) [8, 9]. This was true when humans were pitted against such diverse hosts as guinea pigs, rats, and chickens [9] and persists to the present day [10].

Figure 1.

Aedes aegypti aegypti has evolved a strong preference for human body odor.

(A) Map showing approximate distribution of the human-adapted Ae. aegypti aegypti (grey) and opportunistic Ae. aegypti formosus (black). The green oval marks the area along the coast of East Africa where the two subspecies coexist. Pie charts illustrate the results of ten published blood meal analyses: field collected blood-fed females were scored as having fed on a human or a non-human animal. Locations and references for subsp. aegypti (top row): Hawaii, USA [42]; Florida, USA [43], Pate, Kenya [44], Andaman/Nicobar Islands [45]; rural Thailand [46, 47]. References for subsp. formosus and mixed populations (bottom row): Entebbe, Uganda [7]; Mombasa, Kenya [48]; Ganda, Kenya [7]. (B) An early example of host attraction and preference assays from 1973 in a two-port olfactometer [8]. Females are exposed to streams of air that have passed through a chamber that is empty or contains an odor source such as a live host. Mosquitoes flying upwind in either stream enter a trap from which they cannot escape. No-choice tests measure attraction to each host in isolation while choice tests measure preference. The two types of behaviors are related yet distinct. For example, three strains of subsp. aegypti (grey bars) showed moderate attraction to guinea pig in isolation, but strongly preferred human when given a choice. All strains originated in coastal East Africa and were tested in the early 1970s. Bars show total percent response of 400 females tested in four separate trials of 100 each. Data for replicates were not provided, precluding statistical comparison.

An. gambiae and An. coluzzii are closely related incipient species, formerly known as the ‘S’ and ‘M’ forms of An. gambiae. They belong to a larger complex of morphologically identical sibling species that have partially overlapping distributions across sub-Saharan Africa and vary widely in biting patterns [4, 11]. At one extreme An. gambiae and coluzzii specialize in biting humans, while at the other An. quadriannulatus feeds almost exclusively on animals. The remaining species, including the widespread An. arabiensis, are more or less opportunistic. Blood meal analyses suggest that both An. gambiae and An. coluzzii strongly prefer biting humans [11], but innate odor-based preferences have only been studied in a single strain of An. coluzzii originating in Liberia. This strain is strongly attracted to human odor and averse to cow odor in no-choice trials [12, 13], and clearly prefers human odor in choice trials [13]. An. quadriannulatus, in contrast, is averse to human odor but not cow odor in an olfactometer [12] and responds more strongly to pure carbon dioxide than to a live human in the field [14].

Odorants and odorant blends underlying preference for humans

What characteristics of human odor attract specialists and help them distinguish us from other animals? This question has been addressed over the past several decades by testing the response of female mosquitoes to odors in a variety of laboratory and field settings. Despite the diversity of approaches, several key principles have emerged.

First, carbon dioxide is a very important activator of mosquito host-seeking and instantly sensitizes An. gambiae/coluzzii and Ae. aegypti aegypti females to other host stimuli [15, 16]. Its ubiquitous presence in animal breath, however, provides little information to help specialists distinguish among host species, and some have suggested that it is a stronger directional attractant for opportunistic mosquitoes [14, 17]. Ae. aegypti aegypti females will certainly navigate up turbulent plumes of carbon dioxide, but not when given the option of following a human odor plume instead ([18] and references therein).

Compounds enriched in human odor, in contrast, clearly play a key role in attracting specialists (Table 1). Lactic acid is 10–100 times more abundant in the skin residues of humans than those of other animals, including other primates [19]. In a pioneering study published in 1968, Acree and colleagues identified it as the active component in a fraction of human arm washings that attracted Ae. aegypti aegypti [20]. Females responded to lactic acid in substantial numbers when carbon dioxide was present. Subsequent work replicated the original finding and further suggested that lactic acid is a signature human odorant for this mosquito. Attractive human odor extracts can be rendered unattractive by enzymatic removal of lactic acid [21], while unattractive animal odor extracts can become attractive by addition [22]. Results are more mixed for An. coluzzii, which neither responds to lactic acid in the presence of carbon dioxide nor requires it to show attraction to a human odor extract [19, 23]. Nevertheless, typically unattractive cow odor was rendered attractive to An. coluzzii by supplementation with lactic acid [19]. Ammonia is also abundant in human sweat and may be as important to An. coluzzii as lactic acid is to Ae. aegypti aegypti. Ammonia attracts the former species, but not the latter, with or without carbon dioxide across a range of doses (Table 1) [23–25].

Table 1.

Select human odorants and their behavioral effects on Ae. aegypti aegypti and An. gambiae/coluzzii. Attractants are defined as compounds that attract mosquitoes on their own or in combination with carbon dioxide. Synergists enhance attraction to odor blends. Many compounds have variable effects dependent on context, concentration, or the individual compound within the listed class. Effects listed are those most relevant in the context of attraction/preference for humans.

Compound	Receptors	Primary effect	Aedes references2	Anopheles references2
Carbon dioxide	GRs	Activator Attractant	[16, 18]	[15, 17]
Lactic acid	IRs1	Attractant/Synergist	[20–22, 25, 27]	[19, 24]
Ammonia	IRs	Attractant/Synergist	[25]	[23, 24, 50]
Carboxylic acids	IRs1	Synergist	[26]	[24, 28, 50]
Acetone	ORs	Synergist	[27]	[17]
Sulcatone	ORs	Repellent	[34, 35]	[35, 36]

¹Receptors for lactic acid and other carboxylic acids have not yet been identified, but are presumed to be IRs.

²Emphasis given to seminal and/or recent studies.

No human odorants apart from lactic acid and ammonia have proven consistently attractive to human-preferring mosquitoes when presented singly or with carbon dioxide. Olfaction is highly contextual, however, and a compound that is neutral or repellent when presented alone may be attractive when mixed with other compounds. This type of synergism is critical for specialists. Ammonia enhances the attractiveness of lactic acid to Ae. aegypti aegypti [25], and vice versa lactic acid enhances the attractiveness of ammonia to An. coluzzii [24]. Acetone and an array of carboxylic acids also attract the two species when added to blends [26–28]. Remarkably, An. coluzzii responds to the mix of carboxylic acids given off by Limburger cheese – a food item that shares both its pungent aroma and characteristic bacteria with human feet [28]. Recognizing the importance of synergism among human odorants, several groups have worked to develop synthetic blends containing 3–15 compounds mixed in precise proportions that attract as many, or sometimes even more, mosquitoes than human odor itself [29–31].

Finally, research to date has focused almost exclusively on attraction in no-choice assays rather than preference in a choice context. There are conceptual and empirical reasons to believe that the chemical bases of these behavioral responses are not identical. A mosquito may be attracted to an odor blend, yet still discriminate against it when presented with an alternative option. Indeed, the synthetic blends that attract as many or more mosquitoes than human odor in no-choice trials, perform poorly when pitted directly against human odor in a choice setting [29–31]. Likewise, adding lactic acid to an animal odor blend can rescue attraction by human-preferring mosquitoes [22], but can it fool them into choosing the animal? This test has not been done, but the answer is likely no. There is clearly still work to be done to fully understand the nature of preference for humans at the chemical level.

Genetic and neural changes underlying preference for humans

What genetic and neural changes underlie the evolution of preference for human odor? Potential answers to this question fall into two broad categories corresponding to changes in odor detection by peripheral sensory neurons (Figure 2A) versus changes in integration by central circuits in the brain. Research to date has focused solely on odor detection at the periphery. The results suggest, at minimum, that evolution has ‘tuned’ olfactory sensory neurons to important human odorants via changes in the expression and sensitivity of receptors.

Figure 2.

Changes in peripheral olfactory genes associated with preference for humans.

(A) Schematic of peripheral olfactory system. The mosquito antenna, maxillary palp, and tip of proboscis (not shown) are covered with sensory hairs called sensilla (insets) that house olfactory sensory neurons. The tuning of individual neurons is determined by the binding specificity of several different types of proteins [32]. Odorant-binding proteins (OBPs) carry odorants through the aqueous lymph within each hair to the neuronal membrane where odorant receptors (ORs), ionotropic receptors (IRs), or gustatory receptors (GRs), recognize them and cause the neuron to fire. Individual receptors and OBPs recognize only a subset of odorants and therefore confer specificity to the neurons or sensory hairs in which they are found. Other accessory proteins may also contribute. There are three major morphological types of olfactory sensilla (insets). Capitate pegs on the maxillary palp house one neuron expressing carbon-dioxide sensitive GRs and two neurons expressing select ORs ([32] and references therein). Trichoid and grooved-peg sensilla on the antenna house neurons that likely express ORs and IRs, respectively, based on odor-response profiling [49]. Evolutionary changes in the sequence or expression of peripheral olfactory genes may contribute to host preference by making the mosquito more or less sensitive to specific human and animal odorants. (B) The number of all antennal genes or differentially expressed antennal genes in Ae. aegypti that belong to the OR, IR, and OBP families [10]. ORs, but not OBPs and to a lesser extent IRs, were enriched among genes differentially expressed in human- vs. animal-preferring subspecies and F2 hybrids. Asterisks indicate significant enrichment at the P<0.0001 level. (C) The number of human vs. other odorants for which cognate An. gambiae/coluzzii ORs were upregulated or downregulated by at least 10%, on average, relative to those of the animal-preferring An. quadriannulatus [38]. Fisher’s Exact Test P=0.04.

Insect olfactory receptors include members of three large gene families, the odorant receptors (ORs), ionotropic receptors (IRs), and gustatory receptors (GRs) [32]. These receptors, along with other olfactory genes such as odorant-binding proteins (OBPs), determine the ‘tuning’ of the sensory neurons in which they are expressed (Figure 2A). ORs recognize diverse compounds such as esters, alcohols, and ketones, while antennal IRs are more narrowly focused on the recognition of amines and organic acids, and three highly conserved mosquito GRs recognize carbon dioxide ([32] and references therein). Most of the key attractants and synergists for specialists, including ammonia, lactic acid, and carboxylic acids, are known or presumed IR ligands (Table 1). One might therefore predict that the genetic changes important for preference will have occurred at IR loci or genes affecting the IR pathway. This possibility has yet to be thoroughly explored. Interestingly, however, ORs are clearly important.

DeGennaro and colleagues [33] examined the behavior of Ae. aegypti aegypti females in which the function of all ORs had been eliminated via a mutation in their obligate coreceptor orco. As long as carbon dioxide was present, these females responded as strongly as wild type females to human odor in no-choice trials, yet had significantly reduced preference for humans over guinea pigs in choice trials. The authors suggested that IR and GR ligands are sufficient to drive attraction, while the OR pathway specializes in host discrimination. Further work by McBride and colleagues [10] identified important evolutionary changes in the OR family as a whole (Figure 2B) and a specific OR name AaegOr4. In the two Ae. aegypti subspecies and their hybrids, preference for humans was tightly correlated with naturally occurring AaegOr4 variants that were more highly expressed and more sensitive to a component of human odor called sulcatone. The pattern suggests that an increase in sensitivity to this human-enriched compound contributes to preference for humans.

At face value, this result is at odds with previous work identifying sulcatone as a repellent that may help steer mosquitoes away from individual humans whose odor contains naturally high levels (Table 1) [34–36]. Why would increased sensitivity to a repellent compound in human odor confer preference for that odor? For one, the behavioral effects of sulcatone may be context/concentration dependent. Perhaps sulcatone would enhance attraction if added in small quantities to a blend that does not already contain the compound rather than to real human odor that has a significant amount. Alternatively, the Ae. aegypti genome may contain multiple sulcatone-sensitive receptors mediating distinct behavioral effects. An increase in the sensitivity of AaegOr4 could help counteract repellent effects mediated by other ORs. This interesting controversy highlights the potential complexity of the push-pull context from which preference emerges.

OR evolution may also contribute to preference in Anopheles, despite independent specialization on humans in this genus and near complete lack of orthology at OR loci [37]. Rinker and colleagues [38] compared the OR, IR, and OBP families in An. gambiae/coluzzii and the zoophilic An. quadriannulatus, revealing widespread sequence and transcriptional divergence. Such divergence, in and of itself, does not prove a link to preference – many of the differences could have accumulated by chance or mediate other behaviors. Nevertheless, further analysis of OR expression in female antennae revealed a suggestive pattern. Odorants whose cognate receptors were upregulated, on average, in An. coluzzii relative to An. quadriannulatus were more likely to be human-associated than odorants whose receptors were downregulated (Figure 2C). It is important to note that this analysis assumes conservation of odor-binding profiles in OR orthologs from the two species.

Finally, an electrophysiological study also suggests that specialization in Anopheles may have involved changes in the peripheral olfactory system. In recordings from 20–50 antennal sensory neurons, more cells were excited by human-associated carboxylic acids in An. coluzzii than in the animal-preferring An. quadriannulatus or even the opportunistic An. arabiensis [39]. The underlying receptors and/or accessory proteins are not known. Recent large scale deorphanization studies have identified ligands for many An. gambiae/coluzzii ORs [40, 41], but we known much less about IRs and OBPs, and no genes in any family have been mapped to specific sensory neurons or hairs on antennae. We also lack this basic information for Ae. aegypti. Filling these gaps would greatly facilitate further study of the peripheral changes underlying preference for human odor.

Conclusions

In 2004, Besansky and colleagues argued that the time was right for unraveling the molecular genetic basis of preference for humans in disease-vectoring mosquitoes, with comparisons between closely related species providing powerful aides in this process [6]. Ten years later, this approach is starting to bear fruit. Several recent studies suggest that changes in the tuning and expression of peripheral receptors have sensitized the antennae of Ae. aegypti and An. gambiae/coluzzii to human odorants. The odorant receptor family clearly plays a role in this process, despite the fact that most known attractants in human odor are not OR ligands. This finding suggests that many of the compounds mosquitoes use to discriminate humans from other animals have yet to be appreciated as such. It also highlights the distinction between attraction and preference – two properties of mosquito host seeking behavior that likely have different, if overlapping, genetic and chemical bases. Future studies may also reveal important changes in IRs or other peripheral olfactory genes.

Beyond the peripheral olfactory system, the critical role of synergism in mosquito host-seeking behavior makes central olfactory circuits a potentially fruitful and fascinating area for future work. Responsible for integrating signals mediated by different receptors, these circuits may have experienced changes conferring preference for the specific blend of compounds that define the way we smell.