In London and other areas of northern Europe there are populations of Culex pipiens which have drawn particular attention, not least because they will bite humans voraciously and breed in confined underground spaces. This molestus form of C. pipiens became notorious during the 1939–45 war because it attacked Londoners seeking shelter from German bombing in the underground railway tunnels (Shute, 1951) and continues to be a sporadic biting pest to maintenance workers. It is one of a group of closely related species that forms the C. pipiens species complex. It is morphologically indistinguishable from the C. pipiens that is found in surface populations; yet these are bird biting, and show other substantial differences in behaviour and ecology (Table 1 and see Service, 1993).

Table 1 A comparison of the two forms of Culex pipiens

These differences can be interpreted in a straightforward way as adaptations to a subterranean life: the ability to breed in confined spaces is a prerequisite; the greater warmth underground during winter means that there is no need for a diapause; nutrients are more readily obtained in the larval stage or from mammals rather than from birds. These differences are heritable (Pasteur, 1977), and presumably maladaptive in the above-ground environment. Interbreeding with above-ground populations is additionally hampered by the physical separation of the two habitats.

The disruptive selection for the major adaptations to the above- and below-ground environments, and the isolation of the populations are the types of factor that are thought to Table 1 help establish reproductive isolation (Rice & Hostert, 1993; Bush, 1994), possibly via pleiotropy and/or genetic hitchhiking. Indeed, where the two mosquito forms have been found swarming together, genetic studies of the offspring have established that there is partial reproductive isolation, perhaps because of differences in mating flight (Urbanelli et al., 1981).

In the light of these differences it is not surprising that some authorities have regarded the subterranean form as a distinct species C. molestus (Miles, 1977a,b; Miles & Patterson, 1979). On the other hand, many authors maintain that the pipiens and molestus forms are the same species, with any differences being purely physiological variation (e.g. Harbach et al., 1984). Between these two extremes are those who consider them as subspecies or semispecies (e.g. Urbanelli et al., 1981; Bullini, 1982). In this study no assumptions have been made about the taxonomic status of the two types of Culex. The terms pipiens and molestus form will be used to distinguish populations on the basis of the traits set out in Table 1.

In fact, the differentiation between pipiens and molestus forms seems to change from place to place within the range of the species. The ostensibly clear-cut distinction between the molestus and pipiens forms in northern Europe is not so apparent in the northern Mediterranean area and may disappear in some populations further south—close to where its range meets that of its subtropical and tropical counterpart C. quinquefasciatus. In Spain (Chevillon et al., 1995) and the south of France (Pasteur et al., 1977) there are comparable differences between the populations breeding at the surface and enclosed sites (‘epigeous’ and ‘hypogeous’), although Chevillon et al. (1995) did not find a categorical association between molestus/pipiens traits and hypogeous/epigeous breeding sites. Furthermore, intermediate forms with a mixture of molestus and pipiens traits were detected, suggesting that hybridization between the two forms was occurring.

Directing our attention further south still, Villani et al. (1986) and Nudelman et al. (1988) claimed that in the Middle East (Israel and Egypt) the populations show no differentiation into pipiens and molestus, and that populations in all types of breeding sites contain females that can breed without a blood meal (i.e. they are autogenous). Similarly there is no clear association between habitat and host preference. Reports from Israel suggest that populations showed no significant differences between preference for birds or mammals (Nudelman et al., 1988), whereas in Egypt mammals were reported as being the favoured host (Zimmerman et al., 1985).

Another intriguing similarity between the southernmost populations and the molestus form has been found in genetic surveys. Mosquito populations from the Middle East are more similar to the European molestus form, displaying both low heterozygosity and high frequency of particular alleles (Villani et al., 1986).

This study made use of the populations in the tunnels of the London Underground (the Underground) railway and surrounding regions. The advantages of this system include its known interconnectedness and age, and a history of occupation by the molestus form. An above-ground collection was made in the vicinity of each one below ground. The aim was to assess the genetic differentiation within and between the two forms and to answer the following questions.

1 The subterranean and surface populations of C. pipiens in Britain are more distinct in terms of ecology and behaviour than equivalent populations in the south of the range. Is this associated with reduced gene flow and/or reproductive isolation between the molestus and pipiens forms?

2  Has there been one or several independent colonizations of the London Underground system, either through selection of molestus traits or by environmental changes triggering the expression of pre-existing characters?

3  Have molestus populations originated from long-distance migrants of surface-living molestus-type individuals from the south of the range? This idea arises from two sources. First, that C. pipiens in the Middle East are more similar to the European molestus form than pipiens (Villani et al., 1986), and, secondly, that long-distance migration on this geographical scale appears to have been demonstrated by the spread of insecticide-resistance alleles (Raymond et al., 1991).

Materials and methods

Collection sites

Samples were collected from breeding sites at 20 locations in London. Full sample details are shown in Table 2. Subterranean populations (seven) were collected during expeditions accompanying London Underground maintenance crews. Immature stages (egg rafts, larvae and pupae) were collected from pools, flooded sumps and shafts. Surface populations (11) were collected from garden ponds and water butts on allotments and were chosen as close to each Underground site as was practicable so that direct comparison could be made between them. A population of human-biting, adult females was collected from the interior of houses in Beckton, south-east London, where they had been causing a severe biting nuisance. One population (Oval) was of indeterminate status, being under open sky at the foot of a ventilation shaft. Additionally, a cultured autogenous population (Mogden) was included in the analysis. This population was originally collected by White (see Villani & Hemingway, 1987) from Mogden Sewage Works in west London. Full details of sampling techniques are given in Byrne (1996).

Table 2 Mosquito collection sites and population characteristics

Mosquito rearing

Preimaginal stages were reared in plastic containers in the laboratory at room temperature under conditions of natural light, in water collected from the breeding site. Guinea pig pellets were provided as a larval food source as required. Pupae were picked from the larval trays using a plastic pipette and transferred to small enamel dishes, which were placed in 20 cm nylon and wire frame cages. The adults were allowed to emerge in these cages which contained a 12 cm diameter Whatman filter paper soaked in 10% glucose as an energy source. Each cage was provided with a small (4 cm diameter) enamel dish for oviposition, containing tap water and a small guinea pig pellet. Populations were labelled molestus if the laboratory-reared adults produced egg rafts (i.e. they were both stenogamous and autogenous) and pipiens if they did not. Adults were frozen in liquid nitrogen within 2 weeks of emergence and stored at −70°C for later analysis.

Crossing experiments

Once breeding colonies of some populations had been successfully established, attempts were made to cross them. Virgin females of the autogenous Underground populations (Euston, Shepherd's Bush, Elephant and Castle and Mogden) were isolated by placing individual pupae in 50 mL Falcon tubes covered with netting. Reciprocal crosses were carried out between these populations by placing the females in a cage with five males taken from one of the other populations. A control cross with a male and female from the same population was also carried out. Unidirectional crosses using females from the same populations were carried out using males from some of the surface populations (Walthamstow, Leyton and the two Tufnell Park populations). The results of the crosses were scored as positive if viable egg rafts were produced or negative if they were not. Where F1 generations were produced these were raised separately and scored for the production of an F2 generation.

Laboratory methods

The electrophoretic polymorphism of 20 allozyme loci was studied by cellulose acetate gel electrophoresis as described in Byrne (1996) and full details are available at Individuals from the Shepherd's Bush population were used to standardize allele mobilities and samples from this population were run on every gel.

Analysis of data

Indices of genetic differentiation

Allele frequencies at each locus were calculated for each population and used to calculate Nei's genetic distance (D) and GST (see Nei, 1987). Principal co-ordinate analysis on a matrix of pairwise GST values was carried out using the PCO command of Genstat (Genstat 5 Committee, 1987, p. 481). A summary of allele frequencies is available with the electrophoresis conditions on the website given above. Genetic distances and GST values are given in Table 3.

Table 3 Range of D and GST values between different mosquito populations. Figures in parentheses are means

Mantel tests

The statistical significance of the effects of geographical distance, division into subterranean and surface populations and division by Underground line on genetic distance were tested by a series of Mantel tests. The tests were performed using the R Package (Legendre & Vaudor, 1991) and are shown in Table 4.

Table 4 Mantel tests for mosquito populations. The test statistic t is the standardized Mantel statistic. Probabilities refer to estimates of the probability of obtaining the results by chance under the null hypothesis of no relationship


Population types

The populations collected from the London Underground all produced egg rafts under laboratory conditions, with the exception of that found at the Oval. As the adults had been reared from larval stages, mating and oviposition all occurred within the cages. The Oval population failed to produce egg rafts and was labelled pipiens type, along with the surface populations. The Mogden population was also molestus. The Beckton population could not be assigned any status as no F1 was produced from the adult females collected. The other characters typical of the molestus type could not be scored as easily. Homodynamy, the ability to breed continuously without a winter diapause, did occur in laboratory conditions, even under the natural light regime employed, and the presence of mosquito populations in the Underground system throughout the winter months suggested that this character was present in at least some individuals. Host preference was not tested directly. Only the Beckton population, in which adults were collected after having bitten humans, was known to be using a mammal host, although it is not known whether they would also have used birds.


All crosses between molestus populations resulted in a viable F1 generation. All of these F1 generations went on to produce a viable F2 generation. Conversely, none of the crosses between surface males and autogenous females resulted in any offspring, demonstrating a lack of stenogamy in surface-population males The females used in these crosses were allowed 1 week to produce egg rafts and were then isolated from the males for a further week. After this time they were placed with males from their own population. All but two females went on to produce egg rafts following exposure to males from their own population. The allele frequencies of the offspring showed that the molestus males had fathered the offspring. One intermediate result was obtained. When Oval males were crossed with one female from the Euston population a viable egg raft was produced; however, the F1 population failed to produce an F2 generation.

Allele frequencies

For each locus 20 individuals were scored and the allele frequencies calculated. Of the 20 loci scored, all but two (Me-1 and Gpd) were polymorphic in some of the populations. Surface populations were considerably more polymorphic than those collected from the Underground. Only two loci (Hk and Mpi) were polymorphic in all of the Underground populations. Preliminary inspection of the allele frequency data shows some of the fundamental differences between the pipiens and molestus populations. The molestus populations do not have any unique alleles but a subset of those found in the surface C. pipiens populations. There is a striking difference between the two types in levels of polymorphism. The molestus populations show a consistent reduction in polymorphism to about a quarter of the heterozygosity when compared to the surface-dwelling populations (mean Hsurface=0.27, HUnderground=0.07) The Mogden population has an even lower heterozygosity than those from the Underground (HMogden=0.05), which is consistent with the fact that they have been in long-term culture.

Principal co-ordinate analysis

An immediate picture of variation among the populations can be gained from Fig. 1, which shows the principal co-ordinate analysis of GST. There are three clusters: (i) the populations from the London Underground (▪) and Beckton (□); (ii) the surface C. pipiens () and Oval Underground (); and (iii) the Mogden molestus population ().

Fig. 1
figure 1

Principal co-ordinate analysis of GST showing the first two principal co-ordinates used to analyse the difference in GST between 20 populations of Culex pipiens collected from above and below ground in London. It demonstrates that the populations cluster into two major groups, which correspond to: (i) molestus populations from the London Underground (▪) and Beckton (□); and (ii) the surface Culex pipiens populations () and the Oval population (). Mogden () is an outlier.

Mantel tests

The results of the Mantel tests are shown in Table 4. There is no clear association between geographical and genetic distances in the total population (test 1, P=0.08), the surface populations (test 2,P=0.27) or the London Underground populations (test 3, P=0.46). The Mogden population is responsible for the borderline significance in the total population, as not only is it genetically distinct but it is also geographically the most distant. In contrast the Underground populations were clearly distinct from the surface populations (test 4, P<0.001). The Beckton population is not distinguishable from the London Underground populations (test 6, P=0.25), whereas Mogden, a laboratory population, could be distinguished (test 7, P<0.005). There was significant differentiation between the three Underground lines, (test 5, P=0.016).


Levels of genetic differentiation

The initial assignment of two population types based on laboratory mating behaviour suggested that there were two distinct groups, which correlated to hypogeous (enclosed) and epigeous (open) breeding sites. The genetic data also show significant support for the division of the London population of C. pipiens into two subgroups, pipiens and molestus, which correspond to the different breeding behaviour in the laboratory. This result is readily apparent from the ordination of GST values, which clusters the populations into two groups. The separation into two distinct clusters indicates that gene flow between the two is absent or limited. Each subterranean population was paired with a surface population close by, so appreciable gene flow over a few hundred metres would have been sufficient to produce a thorough intermingling of the two categories of population in the ordination.

Two observations suggest that even where molestus and pipiens populations are in contact gene flow is limited or, conceivably, absent. First, the Oval population (, Fig. 1), clusters with the surface populations on the criteria of genetic distance and postmating and premating reproductive isolation, despite being underground (although open to the sky). Similarly, a previous study by Chevillon et al. (1995) found that mosquitoes collected from what they termed ‘open hypogeous sites’ were not differentiated from epigeous populations. Secondly, the Beckton population (□, Fig. 1) is clearly flying on the surface to reach its human victims. Although the breeding sites beneath houses have been blocked, other sites have not yet been located by the pest control authorities; one possibility is the large containers on an adjacent but inaccessible derelict industrial site. The genetic data only give a hint of gene flow: Beckton clearly clusters with the Underground populations, but it does show higher heterozygosity than the Underground populations, which might be a consequence of limited gene exchange with the local surface populations.

The Underground populations are much less tightly clustered than those overground, suggesting that gene flow is more limited or genetic drift relatively stronger. The significant genetic differentiation between lines (see Table 4) can be explained as a consequence of the shaping of gene flow by the tunnel system. Whereas the lines run through tunnels connected principally at the interchange stations, the piston-like action of trains within a tunnel creates a strong draught that may move the adults between breeding sites on the same line.

This picture of isolation contrasts with genetic studies from more southerly populations (France, Spain and Italy), where there is more evidence of extensive gene flow between pipiens and molestus populations. The penetration of insecticide-resistance genes into hypogeous populations (Rivet & Pasteur, 1993; Chevillon et al., 1995) provides the most straightforward case. This transmission of resistance alleles is not necessarily an indication of frequent gene exchange, as the strong selection pressure imposed by insecticide use could have established resistance in the hypogeous populations from a very small number of successful migrants. More straightforward evidence comes from Italy (Urbanelli et al., 1981). Three types of population were found on the basis of genetic surveys of allozyme loci: those consisting solely of C. pipiens pipiens, others consisting solely of C. p. molestus and also populations which were intermediate and were presumed to have arisen by hybridization of the other populations.

Reproductive isolation

The lack of demonstrable gene flow between the two forms in London may, in part, be explained by the implications of the breeding experiments, in which no offspring were recorded from crosses between Underground and surface populations. The ability to mate in confined spaces (stenogamy) may be a major contribution to premating reproductive isolation. Intriguingly, the Oval population, which seems to be an essentially overground population living underground (but below open sky) produced one (F1) egg raft. The F2 eggs, however, failed to hatch.

The genetic distances measured by the present study between the Underground and surface populations (0.07–0.24) would suggest that the relationship between the two population types is between that of local populations and subspecies (see Ayala, 1975). However, genetic distances may not be useful in populations where demographic effects are the dominant factor in the population history; for example, they do not reliably separate domestic livestock breeds (see, e.g. MacHugh et al., 1997). Using Nei's formula (see Nei, 1987), the estimated time since separation of the surface and Underground populations is ≈100 000 years, even using the correction for drift in one population. This time-scale is, of course, inconsistent with the known history of the tunnels, most of which are less than 100 years old. A second illustration of the imprecise nature of such estimates of separation time is provided by the Mogden population, which has only existed as a caged population for ≈10 years and yet is as distinct as molestus and pipiens. These differences can be explained as a combination of severe genetic drift and the effects of selection in a caged environment. This laboratory history may be analogous to that of populations that adapt to a subterranean life in the wild: they may also be presumed to have experienced strong selection and substantial genetic drift.

Origin of the Underground populations

Two possible sources were considered for the colonization of the London Underground—local London surface populations and long-distance migrants from Mediterranean populations.

Long-distance migrants

Populations of C. pipiens from the Mediterranean region, especially those from North Africa, show similarities to molestus populations from Britain. In particular, they have reduced polymorphism and higher frequency of some alleles associated with the Underground populations (Villani et al., 1986). Little evidence has been found to support the proposal that the London Underground was colonized from this source. The alleles found in the Underground do not include any that are otherwise found only outside Britain. Only Got-1 is fixed for an allele that is rare on the surface but that is frequent in molestus-type individuals from other areas (Chevillon et al., 1998). The identity of alleles was confirmed by cross-checking with control strains provided by M. Raymond.

Local colonization

The data are more consistent with local populations being the source of molestus populations in the Underground. However, Fig. 1 shows that there is no association between individual Underground populations and local surface populations, which suggests that multiple independent colonizations are unlikely or, at least, that such events are rare. The most parsimonious explanation is a single colonization. This would explain the fixation of a single allele at multiple allele loci such as the esterases. The similarity of the Underground populations to the more distant Beckton population (nearest Underground line 4 km) raises the possibility that adaptation to the underground environment may have occurred only once, and that the established subterranean populations are the source for subsequent colonizations.

Adaptation, population fragmentation and reproductive isolation

The distinction between pipiens and molestus forms appears to increase with latitude. This pattern can be interpreted in terms of the interplay of deterministic and stochastic processes that can lead to reproductive isolation. The requirement for a winter diapause at high latitudes may play a pivotal role. In North Africa (below 38°N) the above-ground populations are diapause-incompetent, a pattern that may be related to day length (Zimmerman et al., 1985). At the other extreme of the range, northern European subterranean populations become isolated from the surface populations with the onset of diapause in the surface form in the Autumn. Gene flow is therefore restricted for two reasons. First, there is a shorter period when breeding in the two forms overlaps. Secondly, if subterranean populations set up satellite populations above ground, they will not survive the winter and will only have a limited life during which they could act as stepping-stones for gene exchange.

These trends vary in a predictable way from the south to the north of the range. The intervening northern Mediterranean populations show the expected evidence of intermediate gene flow and this would explain the diminished differentiation into molestus and pipiens forms. Two stochastic processes also come into play in the north of the range. Genetic drift will tend to fix molestus traits in subterranean populations. The consequent linkage disequilibrium with eurygamy above ground and lack of diapause below ground will further hamper gene flow throughout the genome. In addition, the establishment of different clones of the Rickettsia-like bacterium Wolbachia pipientis can provide the coup de grâce to gene flow by leading to incompatibility of matings between differently infected populations. The crossing types are sufficiently labile to change from place to place and through time in the same population (Raymond et al., 1986; Magnin et al., 1987). Indeed C. pipiens shows a high number of incompatible cytotypes relative to other insects, despite the unexceptional degree of genetic divergence between their strains of Wolbachia, which may indicate interactions between Wolbachia and Culex genomes (Guillemaud et al., 1997).

This combination of factors can explain why gene flow is negligible or absent between the two forms in the populations that were studied. As it appears plausible that colonization of the London Underground has occurred only once, the reproductive isolation could have accumulated in the few hundred generations since then. Conversely, in the south of the range, where climate promotes more frequent gene flow, this may be sufficient to prevent the establishment of reproductive isolation, even though the pleiotropic effects of selection in the underground environment are the same.