The invasion of southern South America by imported bumblebees and associated parasites


  • Authors, other than the first author are listed in alphabetical order.


  1. The Palaearctic Bombus ruderatus (in 1982/1983) and Bombus terrestris (1998) have both been introduced into South America (Chile) for pollination purposes. We here report on the results of sampling campaigns in 2004, and 2010–2012 showing that both species have established and massively expanded their range.
  2. Bombus terrestris, in particular, has spread by some 200 km year−1 and had reached the Atlantic coast in Argentina by the end of 2011. Both species, and especially B. terrestris, are infected by protozoan parasites that seem to spread along with the imported hosts and spillover to native species.
  3. Genetic analyses by polymorphic microsatellite loci suggest that the host population of B. terrestris is genetically diverse, as expected from a large invading founder population, and structured through isolation by distance. Genetically, the populations of the trypanosomatid parasite, Crithidia bombi, sampled in 2004 are less diverse, and distinct from the ones sampled later. Current C. bombi populations are highly heterozygous and also structured through isolation by distance correlating with the genetic distances of B. terrestris, suggesting the latter's expansion to be a main structuring factor for the parasite.
  4. Remarkably, wherever B. terrestris spreads, the native Bombus dahlbomii disappears although the reasons remain unclear. Our ecological and genetic data suggest a major invasion event that is currently unfolding in southern South America with disastrous consequences for the native bumblebee species.


Today, the invasion of species into foreign regions, that is, the spread of a species in an area where it was previously not found, and the resulting mixing of the world's biota are on the rise, being greatly facilitated by human travel and international trade (Pyšek et al. 2010). In addition, animals or plants are purposefully imported into a new area by humans because of their anticipated beneficial effects for agricultural produce or to control pests. Such biological invasions can become detrimental when the imported species escapes control and subsequently causes economic losses to crops or livestock (Vila et al. 2010) or starts threatening the native fauna and flora. However, whether or not an invasion is successful is assumed to depend on various factors, such as resource supply (Pyšek et al. 2010), introduction effort and founder diversity (Lockwood, Cassey & Blackburn 2005) or the invader's affiliation with human activities (Jeschke & Strayer 2006). A more recent emphasis is on the effects of parasites, for example, when resident parasites select against invaders, novel parasites select against residents or when invaders leave behind their parasite fauna (‘enemy release hypothesis’; Prenter et al. 2004; Crowl et al. 2008).

Cases of purposeful introductions of foreign animals and plants are often well documented, and the agenda associated with the introduction is clear. Such cases can therefore elucidate patterns and, sometimes, factors that mediate the invasion process. Here, we report on one such case – the planned introduction of European bumblebees for pollination services into southern South America. Bumblebees are important pollinators in their natural range (Bingham & Orthner 1998). As commercially produced pollinators they add substantially to the economic value of agricultural products (Goulson 2003; Velthuis & van Doorn 2006), the commercial trade is mainly in two species, Bombus terrestris L. in Europe and Bombus impatiens Cresson (Bombus occidentalis Greene to a lesser extent) in North America, both of which are bred in the hundreds of thousands of colonies mainly for greenhouse pollination of valuable crops (Velthuis & van Doorn 2006). At the same time, the natural abundance and diversity of bumblebee species is in decline in many parts of the world (Williams 1982; Biesmeijer et al. 2006; Goulson, Lye & Darvill 2008), with parasites being identified as drivers of these declines (Cameron et al. 2011). There is therefore growing concern about pathogen transport with imports and spillover to resident pollinator communities due to the fact that non-native, commercially raised bumblebees are being exported and have now established in various regions of the world (Meeus et al. 2011). These colonies commonly contain parasites (Murray et al. 2013). Cases in point are the import and subsequent naturalization of B. terrestris into Northern Japan (Hokkaido) in the 1990s (Inoue, Yokoyama & Washitani 2008) with concerns of parasite spillover (Niwa et al. 2004; Goka, Okabe & Yoneda 2006). Several Bombus species from England were introduced and established in New Zealand and have brought parasites with them (Hopkins 1914; Macfarlane & Griffin 1990).

Here, we report on the case of two European Bombus having been introduced for pollination in southern South America (Chile) in the 1980s and 1990s. We aimed at analysing the historical spread of these species, their population genetic structure and the presence of two common parasites, Crithidia (Trypanosomatidae) and Nosema (Microsporidia), to scrutinize the extent of the invasion and to elucidate the possible role of parasitic infections in this process. In fact, concerns over parasitism as a cause for the displacement of native species have recently resurfaced (Aizen, Lozada & Morales 2011). However, prior to our first campaign (2004), next to nothing was known about the parasites and diseases of Bombus in southern South America, but reports have multiplied since (Plischuk & Lange 2009; Plischuk et al. 2009, 2011; Maggi, Lucia & Abrahamovich 2011; Arbetman et al. 2013). Yet, this report represents the first systematic survey to date.

Introduction of Bombus spp. into South America

Of the 250 species of bumblebees (Bombus spp.; Cameron, Hines & Williams 2007), a total of 24 species have naturally reached South America (Cameron & Williams 2003; Hines 2008). Here, we focus on the Patagonian region (the south of Argentina and Chile) that harbours five native species. We did not encounter Bombus atratus Franklin, Bombus bellicosus Smith or Bombus funebris Smith in this study, either because their known distributional ranges do not, or only marginally, overlap with our sampling locations (Abrahamovich & Díaz 1982; Abrahamovich, Diaz & Morrone 2004; Abrahamovich, Diaz & Lucia 2007), but we did find Bombus dahlbomii Guérin-Méneville and Bombus opifex Smith.

Introduction of Bombus ruderatus

For importation, European B. ruderatus Fabricius were collected in New Zealand (where it had previously been introduced in 1882) in December 1982 (199 queens) and November 1983 (192 queens) and shipped to Chile (Arretz & MacFarlane 1986). New Zealand was chosen due to the synchronicity in the breeding cycle with South America and the assumed low abundance of parasites at sampling locations. The sampling sites were believed to be free of the nematode, Sphaerularia bombi Dufour 1837 and the tracheal mite, Locustacarus buchneri Stammer 1951. They were also assumed to be free of the protozoan parasite Apicystis (Mattesia) bombi Lipa & Triggiani 1996, as this gregarine is reported to be missing in New Zealand (see Macfarlane, Lipa & Liu 1995, for a review). Several other mite species in New Zealand were also known from the native South American B. dahlbomii (Arretz & MacFarlane 1986) and thus not considered a problem. Queens of the first shipment were released on 23 December 1982 east of the city of Temuco (100 queens released near Coipue; 45 released near Cunco) in an area of red clover production. One hundred queens of the second shipment were released on 6 December 1983 near Coipue, and an additional 69 queens on 7 December 1983 near Cunco (Fig. 1; Arretz & MacFarlane 1986). Introductions were successful because workers of B. ruderatus were seen near Coipue in February and March 1983 and in the following southern spring (November/December 1983) about 7 km south of Cunco.

Figure 1.

Original distribution of the native Bombus dahlbomii in southern South America, according to Abrahamovich, Diaz and Lucia (2007) and Montalva et al. (2011); this species occurs south and west of the dashed line except for the interior of the Patagonian steppe (are marked by the dotted line). B. dahlbomii is typically associated with more or less forested areas. Release sites of imported bumblebees in Chile are indicated by arrows. The current invasion by Bombus terrestris most likely started from the research facilities at Quillota (based on Ruz 2002; Montalva et al. 2011); for Bombus ruderatus, three release sites were used – (Coipue), Malleco and Cunco, based on Arretz and MacFarlane (1986) (arrows). No records are available for Tierra del Fuego. Sampling sites for the four campaigns are indicated by their population id (see Tables 1 and S1, Supporting information).

Introduction of Bombus terrestris

The information on the release of B. terrestris in Chile is much less detailed (Ruz 2002; L. Ruz, pers. comm.). One source was from Israel, likely from the Mt. Carmel region, where B. terrestris became the dominant species (Dafni & Shmida 1996) after having been first recorded in the early 1930s in the North of Israel. A possible second source was from Belgium, but details of this shipment and its release are missing. Imports were authorized by the Servicio Agrícola y Ganadero (SAG, Chile) in 1998 (Ruz 2002). A company (Xilema S.A.) handled the actual introductions from Israel (see reports by the FIA, Fundación para la Innovación Agraria; FIA-Gobierno de Chile 2000). Xilema also ran tests, on behalf of the FIA, for pollination efficiency on tomato in the experimental station of the Pontificia Universidad Católica de Valparaíso ‘s School of Agriculture near Quillota (cf. Fig. 1) during 1997/1998 using both closed and open glasshouses (FIA-Gobierno de Chile 2000; project code FIA-PI-C-1997-2-A-009). According to FIA, Xilema had continued to import bumblebees after the termination of the project (i.e. after 1998); further companies are said to have applied to SAG for import permissions, but no further information is available. The FIA had furthermore run a project to breed B. terrestris in the years 2008–2009 (project code 08CS-1111). According to Prof. Eugenio Lopez (School of Agronomy, Valparaíso University; contacted in early 2012; L. Ruz, pers. comm.), the first shipment from Israel in 1997 was used within the FIA project in Quillota; the second shipment from Israel in 1998 was distributed to farmers across Chile in Arica, Copiapó, San Felipe, Quillota, Limache, Santiago and Los Angeles (Montalva et al. 2011). Given this historical information, we tentatively set the introduction point of B. terrestris relevant for this study to the year 1998 in Quillota. In addition to the two European species, the North American B. impatiens was also imported by farmers but no further details are known (Ruz 2002). No specimens of this species were encountered during our studies; hence, it probably did not establish.

Materials and methods


We did several explicit sampling campaigns (Fig. 1). Campaign (1): In January 2004, one of us (RS) sampled the area around Villarrica (Chile). Campaign (2): In January–February 2010, sampling was done by Paul Schmid-Hempel (PS-H) and Regula Schmid-Hempel (RS-H) in Chile from c. 100 km north of Copiapó (administrative Region III) to the island of Chiloé (Region X). Campaign (3): In January 2011, PS-H and RS-H surveyed Patagonia (along Ruta Nacional, R.N. 40) from San Carlos de Bariloche, Argentina, to Punta Arenas (Chile) and along the Atlantic coast (R.N. 3 to Comodoro Rivadavia). Campaign (4): in October–November 2011, SP checked the area around Comodoro Rivadavia for B. terrestris. Campaign (5): In January 2012, DG and JS surveyed the eastern slopes of the Argentinian Andes, from Bariloche north towards Mendoza. For details, see Table S1 (Supporting information). Generally, we aimed at around 50 specimens per site, but this could not always be achieved due to weather and local conditions.

Sampling locations were grouped into larger geographical regions: (i) ‘Chile Central’ (sampled in 2004: ‘Chile Central 2004’) representing the core Chilean mainland from Santiago south to Puerto Montt; (ii) ‘Andes East’ – the eastern slopes of the Andes, from San Martín de los Andes (where B. terrestris was first reported in Argentina) south to Gobernador Costa; (iii) ‘Argentina Atlantic’: the reach from inland, Sarmiento, to the Atlantic coast at Comodoro Rivadavia; (iv) ‘Patagonia West’: the Patagonian region on the western slope of the Andes; and (v) ‘Patagonia South’: the southern tip of South America, especially the area from around Puerto Natales to Punta Arenas. The sampling periods (January–March, except Campaign 4 to the Atlantic coast) coincided with the known phenology of Bombus in these areas (e.g. Abrahamovich & Diaz 2001; Aizen 2001).

During the sampling campaigns, sites along the major roads with abundant flowers typically visited by bumblebees were searched. These sites consisted of flowering patches either directly along the road or land that could easily be reached by foot in the vicinity. The decisions were made on the spot, such as to sample native bumblebee flowers known to be visited by the native Bombus species. Examples include amancay (Alstroemeria aurea), almost exclusively having been visited by the native B. dahlbomii (Aizen 2001) especially in forested or bushy areas, Fuchsia (magellanica), for example, abundant in many places along Lago General Carrera, various Fabaceae (e.g. Adesmia), other flowering plants (e.g. Phacelia, Boraginaceae), but also more bush-like vegetation (Luma spp.). In addition, imported plants such as clover (Trifolium) and Asteraceae (Carduus) were also prominent at many places. Note that B. dahlbomii was also found in large numbers along our ‘road sites’ in the very same vegetation as B. terrestris, such that the differences reported here are not simply due to a possible sampling bias.

Sites were separated by at least 10–20 km, but generally separated by larger distances (coordinates, see Table S1, Supporting information). Bees were collected at random by walking slowly through a site and captured when visiting flowers, either by directly pushing them into sampling vials or by netting as appropriate. The individuals were put into highly concentrated alcohol (>90% ethanol) either immediately after sampling or first individually stored in a cooler and put into alcohol in the evenings of every sampling day. Species and sex/caste were identified at capture but verified in the laboratory later. However, all species sampled here (Bterrestris, Bruderatus, Bdahlbomii and Bopifex) are readily distinguishable from one another by visual inspection alone (Abrahamovich, Diaz & Lucia 2007).

DNA Extraction and Host Genotyping

Only genetic data from B. terrestris females are reported here due to their abundance (sample sizes), wide distribution and the availability of many polymorphic microsatellite loci. Genomic DNA (also used for parasite identification, see below) was extracted after dissection from whole guts using the DNeasy 96 Blood & Tissue Kit® (Qiagen GmbH, Hilden, Germany) following the instructions by the manufacturer. Note that no B. terrestris were present in the 2004 samples. For B. terrestris, we used seven microsatellites developed by Estoup et al. (1993) but adapted the protocol as follows. We ran two multiplex PCR: ‘BB-Msat1’ combined the PCR for primers B100, B118, B124, B126 and B132 (annealing temperature = 57 °C); ‘BB-Msat2’ combined the reaction for primers B10 and B11 (annealing temperature = 52°). BB-Msat1 and BB-Msat2 were mixed before loading the reactions onto an ABI 3730 sequencer (Applied Biosytems Inc., Norwalk, CT, USA) Fragments were further analysed with the software Peak Scanner™ (Applied Biosystems), and all results inspected visually and checked for consistency and clarity of signal.

Parasite Identification and Genotyping

Infections by the gut parasite, Crithidia (Trypanosomatidae), were genotyped from all host species. We checked for infection status (yes/no) with the Cyt b sequence (Schmid-Hempel & Tognazzo 2010), and infection status for Nosema bombi Fantham & Porter 1914 (Microsporidia) with the small subunit rRNA (using the primers 18f and 1537r from Baker et al. 1995). For this, PCR products from all bees were visualized on PCR CheckIT® gels (Elchrom Scientific, Baar, Switzerland) together with the necessary controls. Presence of a band was taken as positive for infection, absence as negative. PCR products from positive infections were purified (ExoSAP method; Hanke & Wink 1994; Werle et al. 1994) and directly sequenced using BigDye® chemistry on a 3130/3730 ABI Sequencer (Applied Biosystems). Sequences were edited and blasted for parasite identification using the software MacVector 12.5.1 (Mac Vector Inc., Cary, NC, USA). For Crithidia, all positives were of Cyt b-type ‘A1’, defining them as species C. bombi Lipa & Triggiani 1988 (Schmid-Hempel & Tognazzo 2010). Therefore, a subsequent PCR was performed for the five C. bombi microsatellites loci Cri4G9, Cri4, Cri16, Cri2F10 and Cri1B6 (Schmid-Hempel & Reber Funk 2010) to define the genotype. Two multiplex PCRs (CB-Msat1 and CB-Msat2) were done as described in Schmid-Hempel et al. (2011). The two multiplex reactions were genotyped separately on an ABI 3730 (Applied Biosystems). Fragments were analysed using the software Peak Scanner™ (Applied Biosystems) and again visually inspected. DNA samples from 2004 had been extracted during an earlier study (Salathé Zehnder 2007) using 10% Chelex®100 (Bio-Rad, Hercules CA, USA), mixed with an equal volume of Ringer solution (Merck, Darmstadt, Germany). At the time, the samples were analysed for infection status (yes/no) by one of us (RS); these infection status data were used here. For the current combined study, these extractions were retrieved from storage but, unfortunately, could no longer be used for a reanalysis. Therefore, we had to use a subsample of the infected bees from 2004 for which the guts had been stored at −20 °C (hence, not all infected bees of 2004 could be fully genotyped).

Statistical Analyses of Genetic Data

To analyse the genetic data, we used genepop (Raymond & Rousset 1995), r-packages pegas (Paradis 2010), adegenet (Jombart 2008), hierfstat (Goudet 2005) and the program structure (Pritchard, Stephens & Donnelly 2000); for geographical rendering: r package maptool (Lewin-Koh et al. 2009). For the genetic analyses of B. terrestris, populations with less than five individuals were merged with a neighbouring one whilst respecting geographical barriers (e.g. not across the Andes; cf. Table 2). This left us with a total of 28 populations. Rerunning the major analyses by simply excluding all small populations did not change any of the conclusions. Because not all individuals were infected, sample sizes for the Crithidia data are smaller. We again merged small populations into neighbouring larger ones in a similar vein as above. Given the generally smaller samples, we required at least three individuals per population for genetic analyses (cf. Table S2, Supporting information). In a number of cases, the genotyping of infections showed more than two alleles at one or several loci. Since C. bombi is strictly diploid, this indicates a multiple genotype infection. We have used the method used in Salathé and Schmid-Hempel (2011) to resolve multiple genotype infections into the single genotypes co-infecting a host, which was possible in all cases. We have then used these resolved single genotypes for the analyses reported here. An alternative method (majority rule) used the most frequent genotypes as defined by allele frequencies in the region. The corresponding details of method and the results are given in the Supporting Information for comparison. The results from both methods were qualitatively the same.


The Distribution and Spread of the Imported Species

The original distribution of the native species, especially B. dahlbomii, and prior to the 1990s is summarized in Montalva et al. (2011) and shown in Fig. 1. Furthermore, B. ruderatus had crossed the Andes and was first reported from the Río Negro province, Argentina, in 1996 (Roig Alsina & Aizen 1996; cited in Abrahamovich, Telleria & Diaz 2001). Our first campaign of 2004 was restricted to the Chilean Lake district (region IX, around Villarrica and Pucón); a total of 291 bees could be collected. Only two species were found (B. dahlbomii, B. ruderatus). In the subsequent campaigns of 2010 and later, B. terrestris was found over a wide area (Fig. 2, Table 1). In particular, B. dahlbomii common in 2004 in the Lake district had in the meantime disappeared from this area. During the 2010 campaign, the area from Santiago north to Copiapó was also visited (during January 2010), but no Bombus specimen could be found there.

Table 1. Number of individuals sampled per host species and population
IdaPopulation, regionDate Bombus dahlbomii Bombus opifex Bombus ruderatus Bombus terrestris Sample N
  1. a

    Population id (running number) used in the maps (cf. Fig. 1).

 Chile Central 2004      
05Aguas Grandes16.1.2004850380123
02Campo Anders18.1.200450018068
07Pucón I10.1.20041400014
 Chile Central      
11Los Angeles21.1.20100003030
13Nahal 123.1.201000145
14Los Sauces23.1.201000322557
16Puerto Montt27.1.20101004849
 Andes East      
23El Manso14.1.20110016162
24El Rincón14.1.20110006161
25Trevelin I16.1.20110003333
26Trevelin II17.1.20110034245
43Gobernador Costa7.2.20110002727
45Luján de Cuyo5.1.201209009
46San Rafael6.1.201209009
49Chos Malal8.1.20120002525
53S. Martín de los Andes10.1.20120005252
 Argentina Atlantic      
21Comodoro Rivadavia4.11.20110009696
42Rio Senguerr7.2.201100011
 Patagonia West      
29LGral Carrera I19.1.20119004554
30LGral Carrera II19.1.20110002121
31Rio Baker I19.1.201160006
32Rio Baker II20.1.20110002727
33LGral Carrera III20.1.201130003
34LGral Carrera IV20.1.201120002
35Los Antiguos21.1.20110001414
 Patagonia South      
36El Chaltén23.1.201110001
37El Calafate26.1.20111200012
38Puerta Natales29.1.20116300063
39Punta Arenas I31.1.201170007
40Punta Arenas II31.1.20115700057
 Total 3782819610181620
Figure 2.

Presence of the native species, Bombus dahlbomii, and the two imported, Bombus terrestris and Bombus ruderatus. Data for ruderatus and dahlbomii are from the 2004 sampling campaign; pie size proportional to sample size (see Table 1). Approximate data for the range of terrestris in 1998/1999 are taken from Montalva et al. (2011). Asterisks indicate release sites for the two successfully imported species (open symbol: B. ruderatus; closed symbol: B. terrestris).

In the campaign of 2011, B. terrestris was found on the Chilean side of the Andes as far south as Villa O'Higgins (Aysén region; Fig. 2). In January/February 2011, the invasion front of B. terrestris was likely spreading along the river Baker and the shores of Lake General Carrera, as suggested by the many sexuals found there (see Table S1, Supporting information; males in B. terrestris, young queens in B. dahlbomii). Moreover, whereas only B. dahlbomii were found at one sampling point along Baker (site Río Baker I), only B. terrestris were found 10–20 km further downstream (Río Baker II), thus forming a relatively sharp boundary in an apparently homogeneous habitat. Bombus terrestris seemed to be spreading eastward along the northern shore of lake General Carrera and had just reached Los Antiguos (Argentina) at its eastern end. Along the northern lake shore, only B. terrestris were found (sites Lago Gral Carrera I, II; 70–90% being males), and only B. dahlbomii along the southern shore (connecting to the Río Baker valley; sites Lago Gral Carrera III, IV). Sampling in a corridor crossing the Patagonian steppe suggested that the spread of B. terrestris in early 2011 had reached a point between Sarmiento and the Atlantic port of Comodoro Rivadavia. No specimens of the invader were found at this time on the coast itself despite intensive searching. Yet, sampling in October–November 2011 showed that B. terrestris was now present in large numbers around Comodoro Rivadavia (Table 1, Fig. 2). In the campaign of 2012, B. terrestris was found on the eastern slopes of the Andes at a considerable distance north of its first sighting near San Martín de los Andes (Neuquén province) in 2006 (Torretta, Medan & Arahamovich 2006; Table 1, Fig. 2). The native species B. opifex was only met around San Rafael where B. terrestris was not present at that time.

Our campaigns of 2004 and 2010 confirmed the distributional status of B. ruderatus as described by Montalva et al. (2011); specimens were collected in an area reaching from Temuco to Puerto Montt, Chile. In 2011, we collected B. ruderatus in small numbers also on the eastern slopes of the Andes, from Bariloche south to Corcovado (Table 1, Fig. 2).

Parasite Infections

Crithidia bombi

Among all species, the average C. bombi infection prevalence was 14·3%, with a peak value of 80% for population Los Angeles, and substantial infections in Central Chile and the region of Puerto Montt – Bariloche (Fig. 3; Table S2, Supporting information). A total of 183 hosts infected by C. bombi could be fully genotyped with microsatellites. Of those, 41·5% contained multiple infections; this proportion was not different among host species (B. dahlbomii: 44%, B. ruderatus: 36%, B. terrestris: 41%; Table S3, Supporting information).

Figure 3.

Prevalence of Crithidia bombi infections in all hosts for the 2004 and later campaigns; pie size proportional to sample size (see Table S2, Supporting information). Infections in the 2004 campaign (Lake district, Chile) are in Bombus ruderatus and Bombus dahlbomii (no Bombus terrestris present at the time); infections of later campaigns in all species but mostly in B. terrestris.

Host species differed in infection prevalence (χ2 = 17·22, < 0·001, d.f. = 3, = 1620) with 18·5% for B. dahlbomii (= 378), 7·7% for B. ruderatus (= 196), 14·4% for B. terrestris (= 1018) and none in B. opifex (n = 28). Thus, 70 of the total of 232 infections (30·2%) were carried by the native species B. dahlbomii. Significant differences were also found among populations (χ2 = 351·04, d.f. = 49, < 0·0001; = 50 populations) and regions (χ2 = 93·77, d.f. = 5, < 0·0001; = 6 regions; see Table S2, Supporting information for additional GLM analyses), but these effects are partly confounded with species.

Nosema bombi

Microsporidian infections were identified by rRNA sequences. The closest matches were to N. bombi that was prevalent in the Chile samples, matches to Nosema thomsoni and Nosema portugal were mainly found in Argentina, and a few matches to Vairimorpha sp. at several locations. Overall, 2% of the host bees carried a microsporidian infection (= 1296 bees; Table S2, Supporting information); none was found in B. dahlbomii (= 198) or B. opifex (= 28), but 6·5% (= 77) in B. ruderatus and 2·0% (= 993) in B. terrestris (among species χ2 = 12·95, d.f. = 3, = 0·005). All Nosema infections were thus contained within the imported species. Although N. bombi was too rare for reliable analyses, nominally, populations and regions showed significant differences (Table S2, Supporting information). Region ‘Patagonia West’ (prevalence 10·9%, = 128 hosts) stood out due to high infection levels in B. terrestris that were sampled along lake General Carrera. In all, prevalences of C. bombi and N. bombi were not associated (χ2 = 0·034, d.f. = 1, = 0·85; for population means: Spearman's r = 0·249, = 37, > 0·1).

Genetic Population Structure of Bombus terrestris

A total of 814 females could be successfully genotyped out of 1018 collected B. terrestris, representing 28 populations with at least five individuals (Table 2). Some specimen could not be typed because of technical non-amplification (not null alleles); in some cases, locus B118 did not amplify. All of the seven loci were polymorphic and, in any population, had an average of around five alleles, with a conspicuously low value (3·2 ± 0·77 alleles) for the Atlantic coast population of Comodoro Rivadavia. Similarly, observed heterozygosity value, Hobs, and genic diversity, Hs, were relatively high (Tables 2 and S4, Supporting information). Almost all loci, in all populations, were in HW equilibrium; only 11 out of the 231 tested values (4·8%) were significantly deviating at the < 0·0001 level (Bonferroni-corrected threshold), presumably chance significances. Thus, B. terrestris populations in South America are genetically variable and nominally in HW equilibrium.

Table 2. Genetic data for Bombus terrestris females for the 2010–2012 campaign. Only populations with ≥5 individuals included. No B. terrestris were found at the sampling locations during the 2004 campaign or in region South Patagonia
IdaPopulation, regionaSampledbTypedcn Allelsd (SD)Hobse (SD)Hsf (SD)
  1. a

    Population id as in map (Fig. 1). Note that populations with an asterisk were merged to contain ≥5 individuals: ‘El Manso’ with ‘Bariloche’, ‘Sarmiento’ with ‘Rio Saghuer’, ‘Zapala’ with ‘Picunches’, ‘Nahal’ with ‘Los Angeles’; population ‘Coyhaique’ removed (only one individual). Other populations with no B. terrestris not listed.

  2. b

    Number of B. terrestris (all individuals) sampled in population.

  3. c

    Number of B. terrestris females genotyped for this study.

  4. d

    Average allelic richness (SD) over all loci; rarefaction for minimum sample sizes.

  5. e

    Hobs = fraction of heterozygotes observed over all loci, according to (Nei 1987).

  6. f

    Hs = genic diversity (expected heterozygosity).

  7. g

    In these samples, locus B118 missing (non-amplifying) in population.

 Chile Central (2010)     
09Docas765·27 (1·21)0·762 (0·131)0·798 (0·088)
17Ensenada555·43 (1·40)0·828 (0·214)0·818 (0·107)
14Los Sauces25205·13 (0·58)0·805 (0·078)0·803 (0·058)
12Mininco48455·38 (0·84)0·785 (0·083)0·822 (0·046)
11, 13Nahal*3475·51 (1·45)0·762 (0·189)0·838 (0·066)
18Octay21215·21 (0·85)0·801 (0·095)0·808 (0·053)
15Pucón41415·33 (0·58)0·831 (0·053)0·823 (0·036)
16Puerto Montt48314·99 (0·81)0·743 (0·111)0·799 (0·059)
20Cobquecura18164·98 (0·72)0·830 (0·129)0·791 (9·066)
19Talhuaca44405·46 (0·84)0·791 (0·101)0·825 (0·053)
 Andes East     
49Chos Malal2594·61 (0·82)0·873 (0·149)0·744 (0·108)
27Corcovado62574·86 (0·89)0·723 (0·103)0·772 (0·093)
22, 23El Manso*63545·47 (0·72)0·803 (0·049)0·833 (0·039)
24El Rincón61585·10 (0·61)0·836 (0·057)0·805 (0·039)
43Gobernador Costa27275·12 (0·87)0·787 (0·109)0·794 (0·086)
52Lacar68335·78 (0·70)0·801 (0·071)0·853 (0·038)
47Malargue773·49 (0·77)0·857 (0·184)0·702 (0·092)
48Pehuenches42415·23 (0·71)0·815 (0·067)0·817 (0·050)
53S Martín Andes52475·34 (0·66)0·814 (0·093)0·829 (0·037)
25Trevelin I3384·48 (1·06)0·724 (0·142)0·710 (0·145)
26Trevelin II42424·72 (0·66)0·735 (0·123)0·754 (0·075)
50, 51Zapala*30175·15 (0·93)0·781 (0·116)0·814 (0·056)
 Argentina Atlantic     
21Comodoro Rivadavia96963·20 (0·77)0·615 (0·187)0·595 (0·156)
41, 42Sarmiento*11115·07 (0·97)0·584 (0·165)0·782 (0·096)
 Patagonia West     
29LGral Carrera I45394·54 (0·52)0·777 (0·072)0·742 (0077)
30LGral Carrera II21146·10 (3·10)0·687 (0·193)g0·728 (0·073)g
35Los Antiguos1494·07 (0·81)0·816 (0·115)0·722 (0·083)
31Rio Baker II27135·53 (3·40)0·731 (0·187)g0·786 (0·082)g

The populations of B. terrestris show some differentiation among one another (Gst = 1233·85, χ2 = 1528·6, < 0·001; Goudet et al. 1996; Goudet 2005). Analysis by the program structure (Pritchard, Stephens & Donnelly 2000) using the method of Evanno, Regnaut & Goudet (2005) generated K = 3 subpopulations, primarily with bees from the Atlantic coast (Comodore Rivadavia) and, to some degree, from central Chile being separated from the remaining ones (Appendix S1, Supporting information). No structuring was visible when using the ratio of pairwise Fst/(1−Fst) as suggested by Francois et al. (2010) for a principal component analysis of genetic distance (Jombart, Devillard & Balloux 2010; Fig. S1, Supporting information). Similarly, genetical correspondence analysis (Jombart, Devillard & Balloux 2010) and the Monmonier algorithm (r package hierfstat) also failed to identify clear boundaries between populations. A nearest neighbour-joining tree suggested that the geographically far spread populations were indeed also genetically far away (Fig. S2, Supporting information). Instead of separated populations, differences among the 28 populations seem to result from an effect of isolation by distance (Fig. 4a). We also checked for a relationship of genetic distance, the Fst/(1−Fst) ratio, with the physical distance from the likely source of introduction (here: population Docas). But neither populations of B. terrestris (Spearman's = 0·04, = 0·83, = 25 populations) nor those of C. bombi (Spearman's = −0·32, = 0·13 = 23 populations) showed any such correlation.

Figure 4.

Isolation by distance in specimens collected during the campaigns of 2010–2012 in Chile and Argentina. Geographical distance is given by angular degrees (1° c. 111 km) along the shortest distance; genetic distance is given by the ratio of pairwise Fst/(1−Fst) values for polymorphic loci (= 7 in Bombus terrestris,= 4 Crithidia bombi). (a) Pairwise isolation by distance in B. terrestris (Spearman's = 0·609, < 0·001, = 799; = 28 populations). (b) Pairwise isolation by distance in C. bombi (Spearman's = 0·482, < 0·001, = 171; = 19 populations).

Genetic Population Structure of Crithidia bombi

The population genetic structure of C. bombi infections collected in the 2004 campaign showed lower genetic variation as compared to infections from the later campaigns (Table 3). For example, allelic richness for 2004 (average 1·95 ± 1·00 SD alleles, = 5 populations, rarefaction minimum size) was lower than in the populations sampled later (campaigns 2010–2012: 2·63 ± 0·81 SD, = 19; Mann–Whitney = 435, = 0·003): differences were visible even in the same location (e.g. the Pucón sample in 2004 vs. 2010, Table 3). An exception was locus Cri16 that was completely monomorphic with only one allele (size 119) throughout all samples (Tables S5 and S6, Supporting information). All further genetic analyses were therefore run with locus Cri16 excluded. Apart from two out of 70 tests, all loci and populations were in HW equilibrium (at a Bonferroni-corrected value of P < 0·0001); three significant values likely are chance positives.

Table 3. Genetic data for Crithidia bombi, only populations with ≥3 infections included, based on multiple infections (for single infections, see Supporting Information)
IdaPopulation, regionInfectedbTypedcGenotypesdRichnesse (SD)Hobsf (S.D.)Hsg (SD)
  1. a

    Population id as in map (Fig. 1). Note that populations with an asterisks were merged to contain ≥ 3 infections: ‘Pehuenches’ merged to ‘San Martín de los Andes’, ‘Los Antiguos’ to ‘LGral Carrera I’.

  2. b

    Number of hosts confirmed infected by Cyt b or microsatellites.

  3. c

    Number of hosts with a microsatellite genotype for parasites.

  4. d

    Number of C. bombi genotypes analysed (with multiple infections).

  5. e

    Average allelic richness (SD) over all loci; rarefaction for n = 10 (five diploid individuals).

  6. f

    Hobs = fraction of heterozygotes observed over all loci, according to (Nei 1987).

  7. g

    Hs = genic diversity (expected heterozygosity).

 Chile Central 2004      
05Aguas grandes1515221·86 (1·00)0·443 (0·520)0·319 (0·368)
02Campo Anders66101·67 (0·83)0·500 (0·577)0·283 (0·331)
06Cunco4451·94 (1·30)0·450 (0·526)0·319 (0·384)
07Pucón I6691·67 (0·82)0·500 (0·577)0·281 (0·329)
01Valdivia99121·82 (0·67)0·542 (0·529)0·328 (0·283)
 Chile Central      
09Docas5572·69 (1·13)0·750 (0·500)0·539 (0·359)
17Ensenada6682·73 (0·93)0·750 (0·319)0·568 (0·218)
14Los Sauces3342·49 (0·77)0·812 (0·375)0·521 (0·200)
11Los Angeles2423402·50 (0·57)0·756 (0·311)0·551 (0·177)
12Mininco6352·71 (0·75)0·700 (0·346)0·562 (0·164)
13Nahal 13232·25 (0·50)0·833 (0·333)0·500 (0·136)
18Octay87122·82 (1·04)0·625 (0·369)0·570 (0·292)
15Pucón3332·75 (1·26)0·750 (0·500)0·542 (0·369)
16Puerto Montt1010132·61 (0·68)0·827 (0·346)0·574 (0·208)
20Cobquecura99132·61 (0·57)0·769 (0·294)0·580 (0·174)
19Talhuaca4442·30 (0·50)0·750 (0·353)0·489 (0·168)
 Andes East      
27Corcovado4443·00 (1·08)0·687 (0·314)0·635 (0·269)
23El Manso1716202·83 (0·91)0·750 (0·373)0·597 (0·272)
24El Rincón2625372·87 (0·85)0·824 (0·333)0·615 (0·231)
48, 53S Martín Andes*3343·01 (0·92)0·875 (0·250)0·646 (0·172)
25Trevelin I7792·78 (0·66)0·833 (0·213)0·621 (0·144)
26Trevelin II3342·91 (1·01)0·812 (0·375)0·594 (0·246)
 Argentina Atlantic      
21Comodoro Rivadavia6551·75 (0·50)0·750 (0·500)0·375 (0·250)
 Patagonia West      
29, 35LGral Carrera I*5562·36 (1·17)0·542 (0·459)0·446 (0. 363)
 Total1921832592·45 (0·88)0·701 (0·383)0·503 (0·260)

Overall, the population of C. bombi in the area is structured [Gst = 60·64 (χ2 = 73·1, < 0·001)], and an analysis by the program structure resulted in three populations. However, only the population of the 2004 campaign was clearly different from the remaining ones (see Appendix S1, Supporting information). Therefore, rather than having clearly distinct subpopulations, isolation by distance seems the most defining feature of C. bombi populations; this is true for the later campaigns (see Fig. 4b) but also when tested for all samples (Spearman's r = 0·138, = 0·022, = 276; = 24 populations). A neighbour-joining tree suggested that the samples collected in 2004 are set apart from those collected later and that infections from the eastern slopes of the Andes were close to those from southern Central Chile (Fig. 5). In explicit tests, C. bombi infections were genetically differentiated between the 2004 campaign (all hosts, = 40 infections) and later (2010–2012) campaigns, but (for the later campaigns) not different among host species: B. terrestris (= 136), B. dahlbomii (= 3) or B. ruderatus (= 4). A G-test also separated early and late (in genepop) at every locus and overall (< 0·0001 in all cases).

Figure 5.

Neighbour-joining tree for pairwise genetic distances (Fst/(1−Fst)) of Crithidia bombi infections. Infections from the 2004 campaign (dashed rectangle) separate. No Bombus terrestris were sampled in this campaign. Location ‘Docas’ (rectangle) is close to the introduction site of B. terrestris. Locations on the eastern slope of the Andes (Argentina) are in italics. Location ‘C Rivadavia’ is on the Argentinian Atlantic coast (cf. Fig. 1).

Finally, we looked at the association of hosts and parasite genetic distances, which was possible for a subset of = 106 infected host individuals from 22 populations. No evidence for any association between genotypes per se was found – as is to be expected for neutral markers. However, the pairwise genetic distances of hosts or parasites from the same populations correlated rather well (Fig. S3, Supporting information).


The introduction of two Palaearctic bumblebee species into South America – B. ruderatus in 1982/1983 and B. terrestris in 1998 – has led to a dramatic invasion process as both are now found far from their introduction sites. B. ruderatus is now well-established in mainland Chile as well as on the eastern slopes of the Andes (Table 1, Fig. 2). It crossed into Argentina around the Bariloche region in 1994 (Roig Alsina & Aizen 1996) and coexisted with the native species until B. terrestris arrived there in 2006 (Torretta, Medan & Arahamovich 2006). The situation for B. terrestris is much more dramatic. By the beginning of 2010, it was by far the dominant species in the agricultural mainland of Chile, from Santiago to Puerto Montt and, possibly, Chiloé Island (where we spotted a few B. dahlbomii and B. terrestris in the 2010 campaign). By 2011/2012, it had spread in Patagonia south to a point near Villa O'Higgins in Chile and had reached the northern and eastern shores of Lago General Carrera (named Lago Buenos Aires in Argentina). Furthermore, by February 2011 – despite intensive monitoring – no B. terrestris was found along the R.N. 26 between Comodoro Rivadavia and Sarmiento in Argentina. Yet, by October 2011, it had reached the Atlantic coast (Fig. 2). Bombus terrestris was first seen east of the Andes in early 2006 in the region of San Martín de los Andes (Torretta, Medan & Arahamovich 2006), presumably having crossed the mountains there as the passes to the north, reaching over 4000 m, are presumably impossible to cross. By 2012, B. terrestris had spread further north and south along the eastern slopes of the Andes (Fig. 2).

The fate of the native B. dahlbomii is especially remarkable as our distributional data suggest a rapid displacement by B. terrestris. A few examples illustrate this. In 2004, the native B. dahlbomii and the introduced B. ruderatus were abundant around the Chilean Lake district of Villarrica – Pucón (Table 1, Fig. 2). But by 2010, B. terrestris had become the dominant species whereas B. dahlbomii was no longer found. During the 2010/2011 campaigns, we noticed clear boundaries between the advancing B. terrestris and the presence of the native B. dahlbomii in Southern Patagonia and around Lake General Carrera. Furthermore, B. ruderatus was present around Bariloche and San Martín de los Andes but had become much less abundant after B. terrestris had arrived (Torretta, Medan & Arahamovich 2006). It seems therefore that wherever B. terrestris appears, the native B. dahlbomii disappears whilst, interestingly, B. ruderatus manages to persist at low abundance. Most spectacularly, the advance of B. terrestris has been in the order of 200 km per year – an astonishing figure even given the vast spatial scale of Patagonia. Since not only sexuals but also workers are found in all locations (Table S1, Supporting information), viable colonies must have established in the advancing front as workers are assumed not to venture farther than 1–2 km from their nest (Osborne et al. 1999; Darvill, Knight & Goulson 2004; Knight et al. 2005).

On the western (Chilean) side of the Andes studied here, the habitats are generally temperate in climate, covered with vegetation, forests and agricultural lands. Hence, B. terrestris will find suitable conditions there and had established in most of this area around 2008 at the latest (Fig. 2). Along the eastern slopes of the Andes, B. terrestris is found where similar – albeit generally drier than in the West – habitat conditions prevail. By contrast, virtually all of Patagonia between the foothills of the Andes and the Atlantic coast is dry steppe with scarce floral resources. Given where we found the bees during the 2011 campaign, B. terrestris seems to spread across the steppe along the major river systems draining towards the Atlantic. These river systems offer moist conditions and more abundant floral resources. The prevailing (and sometimes very strong) westerly winds across the Andes no doubt aid this dispersal. Note that in its European native range, B. terrestris does not extend into the far North (this is, only up to c. 60°N in Scandinavia) but is widespread throughout the Mediterranean, a region with dry summers and abundant floral resources in a moist winter. Furthermore, our genetic analyses suggest that virtually all populations of B. terrestris are as genetically diverse as the core populations of mainland Europe (Estoup et al. 1996; Schmid-Hempel et al. 2007; Table 2). As of 2012, the overall population of B. terrestris is not separated into distinct subpopulations but shows isolation by distance (Fig. 4).

Based on our data, we can therefore sketch a likely expansion scenario for B. terrestris (Fig. 6). After an initial expansion across central Chile, which it dominated <10 years after its release, multiple crossings across the Andes are likely to have occurred. Likely, a first crossing around 2006 in the region of San Martín de los Andes, was followed by others in the South, notably around L Gral Carrera. At the same time, B. terrestris spread farther south along the western side of the Andes, and later north and south along the eastern side. Finally, B. terrestris now rapidly expands east across the Patagonian steppe. Given the current speed of dispersal, it will probably reach the Strait of Magellan in a few years, perhaps only slowed down by the subantarctic climate of Southern Patagonia, as one might conclude by its absence from subarctic Scandinavia.

Figure 6.

The putative invasion of Bombus terrestris into South America. The species was imported into Chile (asterisk) in 1998/1999 and escaped soon afterwards. The lines and yearly progress depict a likely scenario according to the results of the sampling campaigns reported here. Note that passes of the Andes between Chile and Argentina can probably not be crossed north of approximately the Temuco-San Martín de los Andes line.

Elucidating a possible role of parasites for the invasion process, we found that parasites, notably C. bombi, are present at high prevalence. Furthermore, populations of C. bombi infections are genetically similarly structured as its main host, B. terrestris, and a historical spread is suggested because infections were considerably less diverse in 2004 than later (Table 3) and can also be separated on a phylogeographic tree (Fig. 5). Also, B. terrestris is known to have carried the infection in the 2009 samples from around Bariloche, Argentina, at high levels (21·6%; Plischuk & Lange 2009), whereas it was not present in four native species collected in several parts of Argentina a few years earlier. Based on Fig. 5, perhaps, C. bombi infections could have been brought in first by B. ruderatus, and then enriched in diversity by the later introduction of B. terrestris, with which it now spreads rapidly across vast scales. Genetic distances among populations of C. bombi do indeed covary with the genetic distances of its now main host, B. terrestris (Fig. S3, Supporting information).

The apparently quick and irreversible suppression of B. dahlbomii by the advancing B. terrestris is an enigma and has been commented upon earlier by other observers (Morales 2007; Arbetman et al. 2013). What possible processes could account for this? Ecological competition between the native species and the invaders occurs (e.g. Morales & Aizen 2006; Madjidian, Morales & Smith 2008 for B. ruderatus; Montalva et al. 2011 for B. terrestris). For example, B. terrestris could ecologically out-compete the native B. dahlbomii through its more generalized flower use (Montalva et al. 2011) or by being more adapted to fragmented habitats (Aizen & Feinsinger 2003). It remains unclear though whether ecological competition can account for such a rapid displacement, and, at least for B. ruderatus, this has been considered unlikely (Aizen, Lozada & Morales 2011). Nevertheless, the role of ecological effects for the displacement of B. dahlbomii remains to be studied.

As outlined above, a factor for displacement could be novel, introduced parasites. Is this likely in this case? Experimental evidence is lacking and will be difficult to establish for such large scales. However, our data together with literature reports can sketch some possibilities. For example, we found that N. bombi only occurs in the introduced species. Plischuk et al. (2009) reported that Nosema ceranae Fries 1996 – an emerging parasite of honeybees in Europe (Klee et al. 2007; Paxton et al. 2007) – is present in native bumblebee species in Argentina (B. atratus, Bombus morio, B. bellicosus). So far, only the pathology of N. bombi is confirmed, but the generally very low prevalence of Nosema in our samples (Table S2, Supporting information) should speak against a major role for this pathogen. It is conceivable that the low prevalence of N. bombi is due to non-sampling of heavily affected (and non-flying) bees. Note, however, we detected Nosema by molecular markers, which identify the presence of the parasite long before any spores can be seen or any pathological effects are noticeable. No infections by the gregarine, Apicystis (Mattesia) bombi (which we have not investigated here), were found in native species in Argentina sampled between 2005 and 2009 outside Patagonia (= 441, B. atratus, B. morio, B. bellicosus, B. opifex, B. tucumanus; (Plischuk & Lange 2009), nor in samples of B. dahlbomii (= 52) and B. ruderatus (= 30; Arbetman et al. 2013) collected between 1994 and 2012 in the area around Bariloche and San Martín de los Andes but before B. terrestris arrived. After the arrival of B. terrestris (before 2009), the prevalence of A. bombi for B. terrestris in January 2009 was 3·6% (= 111 bees; C. bombi: 21·6%; Plischuk & Lange 2009). By 2010, in the same area, 12·1% (= 107) of B. terrestris were infected (Plischuk et al. 2011). By 2012 (Arbetman et al. 2013), prevalence had increased to 47% in B. terrestris (= 30), 56% in B. ruderatus (= 9) and 11% in B. dahlbomii (= 9). By 2009, the parasite also occurred in honeybees in the Bariloche region (10·1%, = 138 for 2009/2010) whereas it appears to be absent in the Pampas region further north-east (Plischuk et al. 2011). The pathology of A. bombi is not well researched (e.g. Liu, Macfarland & Pengelly 1974), but it is considered virulent (Schmid-Hempel 1998).

We now show that C. bombi is very widespread and abundant wherever B. terrestris is found (Fig. 3, Table S2, Supporting information). Furthermore, its population genetic structure seems to match that of B. terrestris. According to Plischuk & Lange (2009), C. bombi appears not to be present in native species outside Patagonia. The primary pathogenic effect of C. bombi is to sterilize founding queens (Brown, Schmid-Hempel & Schmid-Hempel 2003 for B. terrestris), which leads to failure during colony founding and severely compromised reproductive success. This effect might be stronger in a species like B. dahlbomii that likely have never encountered this parasite before (Otterstatter & Thomson 2008); yet, this remains to be tested. Whatever further studies will find, only a novel, rather abundant, sufficiently virulent parasite should have those rapid effects that could indeed reduce the native population in a short matter of time. Both A. bombi and C. bombi would qualify for this.


The introduction into South America of B. ruderatus and, especially of B. terrestris, for pollination purposes is having massive consequences. Bombus terrestris forms a healthy, genetically diverse population that is currently spreading at high rates and already has settled in a vast area, along with parasites, such as C. bombi, N. bombi and Apicystis bombi. In its wake, native B. dahlbomii populations diminished dramatically. Given its biology, B. terrestris has the potential to spread further north and south along the Andes and likely will reach Tierra del Fuego (where currently nothing is known about the native bumblebees) in a few years. Currently, B. terrestris spreads eastwards in Argentina along major river systems towards the Atlantic. B. ruderatus spreads much less aggressively and manages to coexist with B. terrestris in places. The current changes in South America clearly are a stunning example of a rapid invasion event with potentially devastating consequences for the only remaining native Bombus species in Patagonia and a lesson for possible future introductions.


We thank José Montalva for help in bee collecting. Financially supported by ETH Zurich, a grant by the SNF (no. 3100-066733 to PSH), the Carnegie Trust for the Universities of Scotland and the Percy Sladen Memorial Fund to DG. Genetic data shown here were produced at the Genetic Diversity Centre of ETH.