ORIGINAL ARTICLE

Temporal and Spatial Patterns of Bartonella Infectionin Black-tailed Prairie Dogs (Cynomys ludovicianus)

Ying Bai & M. Y. Kosoy & C. Ray &

R. J. Brinkerhoff & S. K. Collinge

Received: 14 February 2007 /Accepted: 14 December 2007 /Published online: 5 January 2008# Springer Science + Business Media, LLC 2007

Abstract We describe the temporal dynamics and spatialdistribution of Bartonella in black-tailed prairie dogs(Cynomys ludovicianus) based on a longitudinal studyconducted in 20 black-tailed prairie dog (BTPD) coloniesin Boulder County, CO from 2003 to 2005. Bartonellainfection was widely distributed in all colonies with anoverall prevalence of 23.1%, but varied by colony from4.8% to 42.5% and by year from 9.1 to 39.0%, with amarked increase in Bartonella activity in 2005. Levels ofbacteremia varied from 40 to 12,000 colony forming units(CFU) per milliliter of BTPD blood, but were highlyskewed with a median of 240 CFU. Bartonella infectionrates were unimodal with respect to BTPD body mass, firstincreasing among growing juveniles, then declining amongadults. Infection rates exhibited a sigmoidal response tobody mass, such that 700g may prove to be a usefulthreshold value to evaluate the likelihood of Bartonellainfection in BTPDs. Bartonella prevalence increasedthroughout the testing season for each year, as newlyemerged juveniles developed bacteremia. Data from recap-tured animals suggest that Bartonella infections did notpersist in individual BTPDs, which may explain the

relatively low prevalence of Bartonella in BTPDs com-pared to other rodent species. No association was foundbetween Bartonella prevalence and host population density.Prevalence did not differ between males and females. Thespatio-temporal pattern of Bartonella infection amongcolonies suggests epizootic spread from northern to centraland southern portions of the study area. The potentialsignificance of the BTPD-associated Bartonella for publichealth needs to be further investigated.

Introduction

The genus Bartonella includes a variety of geneticallyrelated Gram-negative bacteria that parasitize erythrocytesor endothelial cells of a range of mammalian hostsincluding man. Many new species of Bartonella have beenisolated and characterized from diverse species of rodentsand many other animals in the past decade [12, 13, 16, 23,29]. The pathogenicity of many Bartonella species has yetto be elucidated. Discovery of additional Bartonella speciesis one of the most important factors associated with therapid emergence of zoonotic Bartonella infections. As aresult of their wide distribution and human diseaseassociation, Bartonella species have recently been recog-nized as emerging pathogens [1, 5].

Several Bartonella species have been isolated from bothhumans and rodents, indicating that rodents can act aspotential reservoirs for human infections. For example, B.washoensis was first isolated from a human patient withcardiac disease, and was subsequently found in theCalifornia ground squirrel (Spermophilus beecheyi) [26].Evidence that rodent-associated Bartonella can causehuman disease led to numerous surveys that have foundhigh prevalence of Bartonella in a variety of rodents

Microb Ecol (2008) 56:373–382DOI 10.1007/s00248-007-9355-6

Y. Bai :C. Ray : R. J. Brinkerhoff : S. K. CollingeDepartment of Ecology and Evolutionary Biology,University of Colorado,Boulder, CO, USA

Y. Bai (*) :M. Y. KosoyDivision of Vector-Borne Infectious Diseases,National Center for Zoonotic, Vector-Borne, and Enteric Diseases,Centers for Disease Control and Prevention,PO Box 2087, Fort Collins, CO 80522, USAe-mail: bby5@cdc.gov

S. K. CollingeEnvironmental Studies Program, University of Colorado,Boulder, CO, USA

sampled in various parts of the world [3, 24, 36].Arthropods such as fleas have been implicated as potentialvectors for transmitting Bartonella infection among rodents[6, 35]. Vertical transmission also has been reported [25].

Most previous surveys to detect and characterizeBartonella in small mammals have been cross-sectional,limiting the potential for understanding the epidemiology ofBartonella in its hosts. Longitudinal studies allow exami-nation of the dynamics of infectious agents within theirreservoir hosts, which aids in answering many fundamentalquestions. Several longitudinal studies have explored thetemporal dynamics of Bartonella infections in wild rodents,providing valuable information. In Tunisia, Fichet-Calvet etal. [4] found that transmission of Bartonella within apopulation of fat sand rats (Psammomys obesus) occurredmostly during the summer and fall, and the infection ratedecreased as rats aged. Birtles et al. [4] found long-term orsuperseded infection of different genotypes in 12 species ofUK woodland rodents because of lack of cross-protection.Kosoy et al. [27] found persistent and extremely high levelsof bacteremia (median = 4.0 × 104 CFU) in cotton rats(Sigmodon hispidus) in the southeastern USA.

The black-tailed prairie dog (BTPD),Cynomys ludovicianus,is a member of the Sciuridae. The BTPD has been recog-nized as a keystone species in western prairie ecosystems[28], and is widely distributed on short-grass and mixed-grass prairies of North America. Because of habitat loss andintroduction of sylvatic plague, BTPDs currently exist inspatially fragmented colonies connected to different degreesby dispersal [18]. Movement of individuals between nearbycolonies can be relatively frequent [33]. As a strictlycolonial and highly social species, BTPDs may be morelikely than other rodent species to transmit disease amongneighboring individuals. Infected BTPDs potentially spreaddisease to more distant members of their colonies or toother nearby colonies. Humans living in close proximity toinfected BTPD colonies also might acquire Bartonellainfections from these animals. Describing the ecology ofzoonotic pathogens can clarify the population dynamics oftheir natural hosts and warrants further investigation forpredicting and preventing disease occurrence in humans.Bartonella species, which are widely distributed and fre-quently occur at high prevalence in rodent communities, mayprovide a very useful model for addressing those questions.

Our research goals are to understand the ecologicalfactors that influence disease distribution and dynamics inhuman-altered landscapes. We studied Bartonella in BTPDsfrom 20 colonies based on a longitudinal design. Weestimated potential dispersal of BTPD among coloniesusing Bartonella as an ecological marker. Here we describethe mass-, sex-, and density-specific demographics ofBartonella infections in BTPD, and the spatial and temporaldistribution of these parasites among BTPD colonies.

Materials and Methods

Study Sites and Trapping

We studied 20 BTPD colonies (1A–20A) located inBoulder County, CO, in the midst of an urban corridor thatruns along the Front Range east of the Rocky Mountains(Fig. 1). Two major roads (Highways 119 and 36) runthrough the study area. For spatial analysis, we divided ourstudy area into three regions separated by the two high-ways: seven colonies (1A, 2A, 4A, 5A, 8A, 17A, 19A)comprise the “northern” study area, separated from othercolonies by Highway 119; six colonies (6A, 7A, 11A, 12A,14A, 20A) comprise the “southern” study area, separatedfrom other colonies by Highway 36; and seven colonies(3A, 9A, 10A, 13A, 15A, 16A, 18A) located between thetwo highways comprise the “central” study area. Prospec-tive sampling using mark-release-recapture was conductedfrom June to August in 2003–2005. Within each colony,BTPDs were trapped during one, 4-day session per year,using a 5 × 10, 6 × 8 or 7 × 7 trapping grid, depending oncolony shape. Each trapping grid consisted of 50 (5 × 10)or 48 (6 × 8) or 49 (7 × 7) stations (grid nodes). Tomahawklive traps (Tomahawk Live Trap, Inc., Tomahawk, WI),

Figure 1 Study area, showing 20 prairie dog colonies (1A–20A) inurban contexts and short-grass prairie in Boulder County, CO. Forspatial analyses, the study area was divided into northern, central, andsouthern regions, with reference to two major highways crossing thestudy area

374 Y. Bai et al.

baited with a mixture of corn, oats, and barley were set ateach station at either 20m (2003) or 25m (2004, 2005)intervals and were checked within 3h of being set each day.Captured animals were weighed, sexed, bled from thefemoral vein, ear tagged, and released at the sites of capture.Blood samples were stored at −70°C until culturing.

Population Density Estimates

A relative density index for the BTPD population in eachcolony was estimated based on visual counts as describedin Menkens et al. [30]. Timed visual counts of prairie dogswithin colonies were conducted using an observer withbinoculars stationed 100m from the colony perimeter [11,20]. Counts then were standardized for visual area, time ofday, weather, and season [32].

Bartonella Culturing

Culturing was performed for samples collected in 2003 and2005, following methods in Kosoy et al. [24]. Whole bloodwas diluted to 1:4 in brain heart infusion (BHI) medium(BBL, Becton DickinsonMicrobiology System, Cockeysville,MD) that was supplemented with 5% amphotericin B. Dilutedblood (0.1mL) was plated on heart infusion agar supplementedwith 5% rabbit blood (BBL, Becton Dickinson MicrobiologySystems, Cockeysville, MD). Agar plates were incubated at35°C for up to 30days in an aerobic atmosphere of 5% carbondioxide. Plates were monitored for bacterial growth at leastonce per week after initial plating. Bartonella was tentativelyidentified based on colony morphology, Gram-stainingcharacteristics, and bacterial microscopy after subculturingto confirm purity. The number of colonies on the originalplates was counted to allow calculation of colony formingunits (CFU) per milliliter of blood. Cultures were harvestedin 5mL of BHI medium plus 10% glycerol.

Verification of Bartonella

Cultured Bartonella was verified by polymerase chainreaction (PCR) amplification of a region in the citratesynthase (gltA) gene using the specific oligonucleotideprimers BhCS781.p 5′-GGGGACCAGCT CATGGTGG-3′and BhCS1137.n 5′-AATGCAAAAAGAA CAGTAAACA-3′ to generate a 379-bp amplicon of the Bartonella citratesynthase gene (gltA) [31]. Crude DNA extracts wereobtained by beating a heavy suspension of the organisms at95°C for 10min. PCR amplifications were performed in a50-μL reaction mixture containing 5μL of 10× PCR buffer,5pmol of each primer, 200μM of each diethylnitrophenylthiophosphate (dNTP) (Invitrogen, Cergy-Pontoise, France)2.5U Taq DNA polymerase (EuroblueTaq, Eurobio, LesUlis, France), and 2.5μL DNA. Distilled water and DNA of

B. doshiae were included as negative and positive controls ineach PCR run. PCR was carried out in a PTC 200 Peltierthermal cycler (MJ Research, Inc., MA) using the followingprogram parameters: a 3-min denaturation at 95°C, followedby 35 cycles of 1-min denaturation at 95°C, 1-min annealingat 56°C, and 1-min elongation at 72°C. Amplification wascompleted by holding the reaction mixture at 72°C for 10min.PCR products were visualized for the presence of ampliconsof the correct size by electrophoresis in a 1.5% agarose gelwith bromide staining.

Bartonella Detection in 2004 Samples

Most samples from 2004 could not be cultured becausemost Bartonella were killed when a freezer failed. Instead,nested PCR was applied directly on DNA samples extractedfrom the blood samples for detection of Bartonella in thesesamples. For comparison with culturing results, we tested agroup of surviving samples (n = 64) using both methods.Eighty-six percent (55/64) tested samples had concordantresults by the two methods (34 samples were negative and21 samples were positive by both culturing and nestedPCR); two samples were positive by culturing but negativeby nested PCR; seven samples were negative by culturingbut positive by nested PCR. Chi-square test indicated anagreement between culturing and nested PCR results (χ2 =0.8149, p = 0.37). DNA was extracted from animal bloodusing a QIAamp Kit (Qiagen, Chatsworth, CA), followingprocedures in the kit manual. We followed a nested PCRprocedure using two pairs of primers, designed in theGamalyia Institute of Epidemiology and Microbiology inMoscow [21]. The first amplification used the pair of outerprimers G003o_F 5′-TATTACAGATCCGCAACAGAGAATGA-3′ and G003o_R 5′- ACTCGATGACCAAAACCCATAA-3′ to generate a 427-bp amplicon of the BartonellagltA gene using the following program parameters: a 5-mindenaturation at 95°C, followed by four cycles of 1-mindenaturation at 95°C, 1-min annealing at 56°C, 1-min elon-gation at 72°C, 35 cycles of 55-s denaturation at 95°C, 55-sannealing at 56.3°C, 55-s elongation at 72°C, 1-min finaldenaturation at 95°C, and 1-min annealing at 56.3°C.Amplification was completed by holding the reaction mixtureat 72°C for 10min. The PCR products were subjected to thesecond step of amplification using the pair of inner primersG004i_F: TTCACAGGTCCCAACTCTT GCCGC TAT andG004i_R: TGCCAGACGTACAGTG GATGTAGAAGC togenerate a 324-bp amplicon of the Bartonella gltA genefollowing the procedure outlined above for single PCR.

Statistical Analyses

Fisher’s exact test (when the sample size of any testedgroup is less than 60) or chi-square analysis (n ≥ 60) was

Bartonella in Black-tailed Prairie Dogs 375375

used to determine whether Bartonella prevalence differedsignificantly among years, months, colonies, and sexes. Anodds ratio was computed to compare Bartonella infectionprevalence in different mass classes of BTPDs. Simple linearcorrelation was performed to determine the association ofBartonella prevalence with animal mass and populationparameters. All analyses were performed using related pro-cedures in SAS (Statistical Analysis System, version 9.1).Any p < 0.05 was considered statistically significant.

Results

Prairie Dog Captures

A total of 1,362 BTPDs were captured within the 20colonies during the entire study period, with 325, 513, and524 individuals captured in 2003, 2004, and 2005,respectively. The total number captured in each colonyvaried from 20 (4A) to 108 (2A, 19A) (Table 1).

Temporal Dynamics of Bartonella Infection

A total of 1,302 BTPDs were tested for Bartonellainfection, 301 of which were positive either by culturing(232) or by nested PCR (69). Overall prevalence of

Bartonella in BTPD was 23.1%, with 9.1%, 14.9%, and39.0% in 2003, 2004, and 2005, respectively (Table 2), andwith significant difference among years (χ2 = 128, p <<0.01). In the first year, 2003, the prevalence of Bartonellainfection was lowest. In fact, no infection was found inJune, the very first month of the study. Bartonella was firstdetected in July, with a prevalence of 8.9%. Prevalenceincreased to 13.3% in August, but this increase was notsignificant (Fisher’s exact test, p = 0.41). In 2004, theprevalence of Bartonella in June was 5.9%, then signifi-cantly increased to 17.9% in July (χ2 = 9.99, p << 0.01). Adrastic increase in Bartonella activity was observed in2005: starting at a much higher prevalence of 17.8% inJune, Bartonella prevalence reached an extremely highlevel of 46.5% in July (χ2 = 34.63, p << 0.01) (Fig. 2).

Mass-dependence of Bartonella Prevalence in Prairie Dogs

The association between Bartonella prevalence and animalbody mass considered data obtained from 2003 and 2005samples. Data from 2004 were excluded because of theapplication of a different method for detecting Bartonellainfections in the BTPD samples. Prairie dogs wereclassified into nine mass classes, from <300g to ≥1,000gby 100-g increments. Bartonella prevalence exhibited highvariation among mass classes, being highest in class 300–

Table 1 Monthly captures of black-tailed prairie dogs within 20 study colonies, 2003–2005

Colony 2003 2004 2005 Total

June July August Subtotal June July Subtotal June July Subtotal June July August Total

1A 16 16 16 22 38 16 9 25 32 47 792A 37 37 40 6 46 22 3 25 62 46 1083A 17 17 31 31 38 38 69 17 864A 5 5 9 9 6 6 20 205A 21 21 22 22 26 26 69 696A 16 16 17 17 6 6 39 397A 12 18 30 16 16 29 29 12 63 758A 5 11 16 11 8 19 21 6 27 32 19 11 629A 29 29 39 39 27 27 27 68 9510A 12 12 15 15 13 13 28 12 4011A 4 7 11 17 17 16 16 37 7 4412A 4 6 10 14 14 22 22 14 26 6 4613A 8 8 42 42 40 40 90 9014A 6 6 25 25 29 29 60 6015A 24 24 14 14 28 28 66 6616A 16 16 34 34 47 47 97 9717A 1 1 2 10 10 28 28 1 39 4018A 1 7 8 22 22 29 29 1 58 5919A 3 21 24 49 49 35 35 87 21 10820A 17 17 34 34 28 28 79 79Total 14 266 45 325 127 386 513 137 387 524 278 1039 45 1362

Each colony (1A–20A) was trapped during a single, 4-day session each year. Session dates for each colony varied among years. Most colonieswere trapped during July of each year, so total trapping effort varied among months

376 Y. Bai et al.

399g (71.4%) and lowest in class 900–999g (7.1%).Prevalence in the smallest juveniles (class <300g; 42.9%)was relatively low, but increased quickly (class 300–399g;71.4%) with marginal significance (Fisher’s exact test, p =0.07). Prevalence then declined with body mass, but did notfall below 44% until mass reached 700g, after whichprevalence declined dramatically (Fig. 3). The odds ofBartonella infection were over four times higher for BTPDsweighing <700g than for those weighing ≥700g (odds ratio

4.1, χ2 = 16.78, p << 0.01). Correlation analysis demon-strated a highly significant negative association betweenBartonella prevalence and host mass for BTPDs weighing>300g (all sampled classes; r = −0.95).

Bartonella in Recaptured Prairie Dogs

A total of 155 BTPDs were captured more than once during2003–2005. Bartonella infection was examined in 136 of

Table 2 Prevalence (%) of Bartonella infection in black-tailed prairie dogs in 20 study colonies, 2003–2005

Region Colony 2003a 2004b 2005a Totalc

No.tested

No.positive

Percent No.tested

No.positive

Percent No.tested

No.positive

Percent No.tested

No.positive

Percent

Northern 1A 16 3 18.8 38 2 5.3 25 3 12 79 8 10.12A 35 7 20 41 1 2.4 25 4 16 101 12 11.94A 5 1 20 9 0 0 6 2 33.3 20 3 155A 20 2 10 22 8 36.4 26 12 46.2 68 22 32.48A 16 1 6.3 19 0 0 27 2 7.4 62 3 4.817A 2 0 0 10 1 10 28 16 57.1 40 17 42.519A 24 8 33.3 49 11 22.5 35 11 31.4 108 30 27.8Subtotal 118 22 18.6 188 23 12.2 172 50 29.1 478 95 19.9

Central 3A 17 2 11.8 30 1 3.3 36 14 38.9 83 17 20.59A 27 1 3.7 39 8 20.5 27 5 18.5 93 14 15.110A 12 0 0 13 1 7.7 13 1 7.7 38 2 5.313A 8 0 0 42 7 16.7 40 24 60 90 31 34.415A 24 2 8.3 14 3 21.4 28 19 67.9 66 24 36.416A 16 0 0 34 5 14.7 46 35 76.1 96 40 41.718A 8 0 0 22 2 9.1 29 9 31 59 11 18.6Subtotal 112 5 4.5 194 27 13.9 219 107 48.9 525 139 26.5

Southern 6A 16 2 12.5 8 1 12.5 6 3 50 30 6 207A 30 0 0 16 3 18.8 29 15 51.7 75 18 2411A 11 0 0 7 2 28.6 15 6 40 33 8 24.212A 10 0 0 13 4 30.8 22 8 36.4 45 12 26.714A 6 0 0 16 5 31.3 29 8 27.6 51 13 25.520A 15 0 0 22 4 18.2 28 6 21.4 65 10 15.4Subtotal 88 2 2.3 82 19 23.2 129 46 35.7 299 67 22.4Total 318 29 9.1 464 69 14.9 520 203 39.0 1302 301 23.1

Infection detected by culturing (a ), by nested PCR (b ), or by either method (c )

Figure 2 Bartonella prevalencein black-tailed prairie dogs(BTPDs) by month-year,Boulder, CO, 2003–2005. Thefraction reported at each pointrepresents the number ofinfected BTPDs over the num-ber tested. Intra-year prevalencevalues are linked by solidlines; inter-year by dashed lines

Bartonella in Black-tailed Prairie Dogs 377377

these individuals, among which 113 were captured twice,and 23 were captured three times (n = 295 samplesanalyzed). The body mass of examined recaptures rangedfrom 290 to 1,260g. Forty-six (15.6%) of the examinedrecaptures weighed <700g, and 249 (84.4%) weighed ≥700g.Bartonella infection was found in 23 recaptured individuals,including six captured three times and 17 captured twice.Infection was detected only once in each of the 23 individuals.

Bacteremia

Levels of bacteremia were determined for the 232 samplesfrom 2003 and 2005, all of which were confirmed Bartonellapositive by culturing. Levels of bacteremia varied from 40 to12,000 CFU/mL. The distribution of CFU was highlyskewed, with 50% of samples exhibiting ≤240 CFU andonly 18.3% samples exhibiting >1,000 CFU. Animals withlarger mass were less likely to have high loads of bacteria.Although 40 of 173 individuals (23%) weighing <700 greached a level of >1000 CFU, only seven of 59 individuals(12%) weighing ≥700 g reached the same level (Fisher’sexact test, p = 0.04) (Fig. 4).

Host Population Density and Bartonella Prevalence

Host density, estimated by visual counts, varied by year andby colony. The average density among all colonies during2003–2005 was 23.3 animals per ha, ranging from 11.7/ha(12A and 18A) to 37.8/ha (39A), and average density variedfrom 16.8/ha to 28.9/ha among years (Table 3). Nosignificant correlation was found between Bartonella prev-alence and population density in any year (all p > 0.05).

Host Sex and Bartonella Prevalence

The tested animals consisted of 682 (52.4%) female and620 (47.6%) male prairie dogs. Although prevalence inboth sexes varied by year (9.9%–36.7% for females, 8.3%–

40.6% for males), there was no difference in Bartonellaprevalence between sexes either within or across years (allχ2 ≤ 0.86, p ≥ 0.35).

Spatial Dynamics of Bartonella

Infected BTPDs were found in all 20 colonies, but with ahigh variation in prevalence: from 4.8% (3/62, 8A) to41.7% (40/96, 16A) (Table 2). Average prevalence was sig-nificantly different among colonies (χ2 = 85.69, p << 0.01).

In 2003, Bartonella-infected BTPDs were found in only10 of the 20 colonies, and prevalence varied from 0 to33.3% (19A) per colony. The majority of infected BTPDs(22/29) were distributed in six of the seven northerncolonies, with the highest number of Bartonella-positiveanimals located in colonies close to the foothills (eight in19A and seven in 2A). Only one southern colony and three

Figure 4 The distribution of levels of Bartonella bacteremia(measured as number of colony forming units or CFU) was skewedtoward smaller mass classes. Analysis based on data from 2003 and2005. The numbers shown in or above bars represent the number ofBartonella-infected BTPD in mass class <700 g or ≥700 g accordingly

Figure 3 The pattern ofBartonella prevalence in BTPDsversus BTPD mass. Analysisbased on data from 2003 and2005. The fraction reported ateach point represents the num-ber of infected BTPDs over thenumber tested in each massclass

378 Y. Bai et al.

central colonies produced Bartonella-positive samples (twoin 6A and five total in 3A, 9A and 15A; Table 2). Prevalencewas significantly lower among southern and central thannorthern colonies (p < 0.01), but there was no differencebetween southern and central colonies (p = 0.40, Fig. 5).

In 2004, Bartonella infection was detected in 18colonies. The prevalence varied from 0 to 36.4% (5A) percolony. There was a trend toward higher prevalence incentral and southern colonies. In contrast, prevalencedeclined somewhat in the north: no infection was detectedin two colonies (4A, 8A) that were previously infected.Infected BTPDs were distributed in all regions (Table 2)with similar prevalence (p=0.87, Fig. 5).

In 2005, Bartonella infections were again widelydistributed. Average prevalence rose, varying from 7.4%(8A) to 76.1% (16A) per colony. The majority (107/207) ofinfected animals were found within central colonies, but theprevalence did not differ among regions (p=0.60, Fig. 5).Colonies with the highest number of infected BTPDs were16A, 13A, and 15A, all in the central region, plus 17A, thesouthernmost of the northern colonies (Table 2).

Discussion

Bartonella infections are usually highly prevalent in wildrodents, with a rate exceeding 50% in some rodent commu-nities [3, 7, 19, 24, 36]. We report a lower overall prevalence

of Bartonella in BTPDs (23.1%). Moreover, we found lowerlevels of Bartonella bacteremia in BTPDs (median=240CFU) compared to cotton rats (median=4.0×104 CFU) [27].

Animal mass was strongly related to the prevalence ofBartonella in BTPDs. Juvenile BTPDs quickly becameinfected with Bartonella, either before or shortly after firstemerging from their natal burrows. But, as they grew past300 g, infection prevalence declined, in strong correlationwith body mass (r=−0.95). Juvenile animals were signif-icantly more likely to be infected with Bartonella than wereadult animals, and were more likely to have higher levels ofbacteremia. These patterns can be explained by bacterialclearance through acquired immunity [27]. However,Bartonella-specific antibodies have been found in only afew animals, and majority of animals with antibodies hadvery low titers [27]. This might suggest that humoralimmunity is not a major feature of Bartonella infections inrodents, but play a certain role in cats [8]. Cellularimmunity associated with an increase in CD4 Th1 cellsmight be important [2]. Alternatively, mortality or loss offitness induced by the parasite, either alone or in synergywith other factors, can regulate host populations [17, 34].Older, infected adults might be more vulnerable topredation or starvation than their uninfected counterparts.We found that BTPDs became significantly less susceptibleto Bartonella infection after reaching a mass level of 700 g(λ≥4.1). This result suggests at least two hypotheses.BTPDs may experience a dramatic adjustment in immuneresponses when they reach this size. In this case, 700 gmight be used as a mass threshold to evaluate the likelihoodof Bartonella infection in BTPDs. Alternatively, theobserved threshold may reflect BTPD demography. Wesuspect that the majority of animals weighing <700 g wereborn within the year of capture, whereas the majority ofanimals captured weighing >700 g were born in theprevious year. Under this alternative, the utility of 700 gas a predictive threshold may depend on the timing of the

Table 3 Relative estimates of black-tailed prairie dog density (animals/ha)in 20 study colonies, 2003–2005

Colony 2003 2004 2005 Average

1A N/A 64.4 9.8 37.12A 31.1 41.3 31.3 34.63A 19.6 42.2 41.8 34.54A 26.7 6.2 7.6 13.55A 11.6 9.8 24.9 15.46A 13.8 14.2 46.2 24.77A 16 11.6 14.7 14.18A 20.4 41.8 4.9 22.49A 23.6 49.3 40.4 37.810A 8 33.8 20 20.611A 20.9 22.7 22.7 22.112A 16.4 6.7 12 11.713A 8.9 17.8 30 18.914A 13.3 24 16 17.815A 17.3 66.7 25.8 36.616A 4 17.3 24 15.117A 37.3 36 28.4 33.918A 4 11.1 20 11.719A 20.4 33.3 39.6 31.120A 6.7 27.1 24.9 19.6Average 16.8 28.9 24.2 23.3

Estimates determined from repeated visual counts (see text)

Figure 5 Spatial dynamics of Bartonella, showing the prevalence ineach region by year. Prevalence was higher in northern colonies (blackbars) only in 2003 (p<0.01) and never differed between central (graybars) and southern (white bars) within years

Bartonella in Black-tailed Prairie Dogs 379379

epidemiologic study. It may be useful for studies conductedin the summer, when most adults weigh >700 g, but lessuseful in the fall, when first-year animals are exceeding700 g while still in the early stages of infection.

Consistent with early reports [14, 27], our longitudinalstudy of Bartonella infection in BTPDs also exhibited apattern of seasonal dynamics with increasing prevalencethroughout the summer (Fig. 2). These seasonal dynamicsmight be explained by changes in host population structure.Although BTPDs do not typically hibernate, juveniles startemerging from their natal burrows in late April/May. Oncejuveniles emerged, the prevalence of Bartonella increasedfor the remainder of the summer, with the majority ofjuvenile animals developing Bartonella bacteremia causedby their naïve immune systems. There might be interactionsbetween animal mass and season on Bartonella prevalence.However, as our trappings were done only during a relativelyshort period of time each year, the design of our field studydoes not allow us to analyze such interactions, if any.

Bartonella prevalence in 2005 was extremely highcompared to other years of this study. This might be relatedto local activity of plague, caused by the bacterium Yersiniapestis. After plague occurred in several colonies in 2005, themajority of the BTPD population in this region was comprisedof newly emerged young animals (based on mass measure-ment). If the fleas of dead hosts were seeking new hosts, thiscould cause heavy infestation and increase the prevalence ofother flea-borne infections. Other ecological factors may alsohave contributed to the change. However, epidemiologicalfactors may also have played a strong role. Noting thatBartonella occurred at very low prevalence in a spatiallyrestricted region of the study site in 2003, our results mightindicate the initiation and spread of an epizootic. Given theusually high prevalence of Bartonella in rodents, our datasetmay constitute a rare opportunity to investigate the dynamicsof early spread in this type of host–pathogen system. Froman ecological standpoint, the study of Bartonella infection inBTPDs can potentially provide much-needed data to testtheoretical models of infectious disease dynamics withinnaturally susceptible populations and the role of suchinfections in the regulation of the host population [15].

In contrast with previous studies [4, 27], only a smallportion of recaptures were infected with Bartonella, and allof the Bartonella-positive recaptures were infected onlyonce in our study. This might be because most of ourrecaptures (84.3%) were relatively old (≥700 g) and hadalready passed through the active bacteremic phase of theirinfection, in accordance with our observation of low oddsof infection in this group. On another hand, the observedlack of persistent bacteremia suggests that Bartonella doesnot persist for months in BTPDs, and may explain whyBartonella prevalence is lower in this species than in otherrodents. A possible explanation for the differences in

Bartonella epidemiology between rodents is based on theobservation that BTPDs carried only one type/species ofBartonella (Bai et al. 2007, in press), whereas cotton rats[27] or rodents in the UK [4] carried up to three differentgenotypes/species that can replace each other because oflow cross-immunity.

Fichet-Calvert et al. [14] found an inverse relationshipbetween prevalence of infection with Bartonella and hostabundance. We did not observe an association betweenBartonella prevalence and host population density. How-ever, the density estimated by visual counts, as in this study,may not be very accurate compared to the measurement bythe number of captures/100 m trap line used by Fichet-Calvert. Bartonella infection was also not related to hostsex in BTPDs, as other study has reported [27].

As keystone species in western prairie ecosystems [28],prairie dogs have been studied with regard to patterns ofextinction, recolonization, and genetic structure [18, 31].The spatial distribution of Bartonella in BTPDs suggestspathogen spread from northern to central and southern partsof study area over 3 years. Because the type/species ofBartonella found in BTPDs is not found in other rodents inthis region (Bai et al., unpublished), the pattern of spreadmay have resulted as follows: juvenile BTPDs contractedinfection from infectious fleas within their natal coloniesand spread the pathogen by dispersing among colonies. Thepathogen may have been spread directly by dispersal ofinfectious hosts, especially if juveniles dispersed beforeshedding infection, or by the movement of infectious fleascarried by dispersing hosts. The pattern of pathogen spreadmight indicate directions of BTPD dispersal. Our observa-tions support the hypothesis that BTPDs are capable ofdispersing among colonies despite fragmentation of thelandscape. However, there appear to be specific restrictionson the movement of BTPDs. For example, in the center ofour study area, we observed more infected BTPDs incolonies that were relatively close to each other (13A, 15A,16A, and 18A), and fewer infected individuals in moreisolated colonies (9A, 10A), suggesting that BTPDs morelikely dispersed between nearby colonies. Our observationsupports the hypothesis that geographic distance betweencolonies has a strong effect on dispersal compared withother landscape features (Martin et al., unpublished). Roadsmay also facilitate dispersal of BTPDs by acting as primarydispersal corridors [22]. In the northern region, the patternof pathogen spread suggests that BTPDs moved along theroadway from colony 19A to colony 17A, but not to colony8A, which is closer but lacks road access. Colonies 19Aand 8A are also separated by streams and ponds, whichmay be barriers to movement of animals that traffic disease[9]. As a result, 8A seems quite isolated from othercolonies, with few infected animals being detected eachyear. Although long-distance dispersal of BTPDs may

380 Y. Bai et al.

occur, it is less likely that BTPDs cross highways naturallyin this urbanized landscape and perhaps more likely thatpathogen movement across highways occurs through theactive management and relocation of BTPDs in BoulderCounty (Martin et al., unpublished). However, where thespatial distribution of Bartonella among BTPD coloniescan reflect connectivity of colonies and fragmentation ofgrasslands, Bartonella infection may serve as a goodecological marker to estimate the potential dispersal ofBTPD and to measure geographical and ecological factorsresponsible for the connectivity of colonies.

Recently, increased attention has been focused on therole of Bartonella as threats to human health [1, 6, 10].Many Bartonella species are potentially pathogenic tohumans, but if contact between humans and the naturalhost of the species is minimal, this potential is seldomfulfilled [4]. In our study, some colonies that were heavilyinfected (17A, 13A, and 15A) are very close to residentialareas, which may increase the risk of BTPDs transmittingdiseases to humans. BTPD-associated Bartonella has beenproposed as a Candidatus subspecies of human pathogenBartonella washoensis (Bai et al. 2007, in press). Furtherinvestigation is needed to determine the potential signifi-cance of the subspecies for public health.

Acknowledgments We gratefully acknowledge the City of BoulderOpen Space Department, and Boulder County Parks and Open SpaceDepartment for access to black-tailed prairie dog colonies. We thankall field crews for trapping and collecting blood samples. Thisresearch was supported by grants from the NSF/NIH joint programin Ecology of Infectious Diseases (DEB-0224328) and the NationalCenter for Environmental Research (NCER) STAR program of theUS-EPA (R-82909101-0).

References

1. Anderson, BE, Neuman, MA (1997) Bartonella spp. as emerginghuman pathogens. Clin Microbiol Rev 10:203–219

2. Arvand, M, Ignatius, R, Regnath, T, Hahn, H, Mielke, ME (2001)Bartonella henselae-specific cell-mediated immune responsesdisplay a predominantly Th1 phenotype in experimentallyinfected C57BL/6 mice. Infect Immun 69:6427–6433

3. Birtles, RJ, Harrison, TG, Molyneux, DH (1994) Grahamella insmall woodland mammals in the UK—Isolation, prevalence andhost-specificity. Ann Trop Med Parasitol 88:317–327

4. Birtles, RJ, Hazel, SM, Bennett, M, Bown, K (2001) Longitudinalmonitoring of the dynamics of infections due to Bartonellaspecies in UK woodland rodents. Epidemiol Infect 126:323–329

5. Boulouis, HJ, Chang, CC, Henn, JB, Kasten, RW, Chomel, BB(2005) Factors associated with the rapid emergence of zoonoticBartonella infections. Vet Res 36:383–410

6. Breitschwerdt, EB, Kordick, DL (2000) Bartonella infection inanimals: carriership, reservoir potential, pathogenicity, and zoonoticpotential for human infection. Clin Microbiol Rev 13:428–438

7. Castle, KT, Kosoy, M, Lerdthusnee, K, Phelan, L, Bai, Y, Gage, KL,Leepitakrat,W,Monkanna, T, Khlaimanee, N, Chandranoi, K, Jones,JW, Coleman, RE (2004) Prevalence and diversity of Bartonella in

rodents of northern Thailand: a comparison with Bartonella inrodents from southern China. Am J Trop Med Hyg 70:429–33

8. Childs, JE, Olson, JG, Wolf, A, Cohen, N, Fakile, Y, Rooney, JA,Bacellar, F, Regnery, R (1995) Prevalence of antibodies toRochalimaea species (cat-scratch disease agent) in cats. Vet Rec136:519–520

9. Collinge, SK, Johnson, WC, Ray, C, Matchett, R, Grensten, J,Cully, JF, Gage, KL, Kosoy, MY, Loye, JE, Martin, AP (2005)Landscape structure and plague occurrence in black-tailed prairiedogs. Landsc Ecol 20:941–955

10. Comer, JA, Paddock, CD, Childs, JE (2001) Urban zoonosescaused by Bartonella, Coxiella, Ehrlichia, and Rickettsia species.Vector Borne Zoonotic Dis 1:91–118

11. Cully, JF Jr, Barnes, AM, Quan, TJ, Maupin, G (1997) Dynamicsof plague in a Gunnison’s prairie dog colony complex from NewMexico. J Wildl Dis 33:706–719

12. Dehio, C, Lanz, C, Pohl, R, Behrens, P, Bermond, D, Piemont, Y,Pelz, K, Sander, A (2001) Bartonella schoenbuchii sp. nov.,isolated from the blood of wild roe deer. Int J Syst Evol Microbiol51:1557–1565

13. Droz, S, Chi, B, Horn, E, Steigerwalt, AG, Whitney, AM,Brenner, DJ (1999) Bartonella koehlerae sp. nov., isolated fromcats. J Clin Microbiol 37:1117–1122

14. Fichet-Calvet, E, Jomaa, I, Ben Ismail, R, Ashford, RW (2000)Pattern of infection of haemoparasites in the fat sand rat,Rsammomys obesus, in Tunisia, and effect on the host. Ann TropMed Parasitol 94:55–68

15. Grenfell, BT, Dobson, AP (1993) The dynamics of diseases ofwildlife. Cambridge University Press, Cambridge

16. Gundi, VA, Davoust, B, Khamis, A, Boni, M, Raoult, D, LaScola, B (2004) Isolation of Bartonella rattimassiliensis sp. nov.and Bartonella phoceensis sp. nov. from European Rattusnorvegicus. J Clin Microbiol 42:3816–3818

17. Holmes, JC (1995) Population regulation: a dynamic complex ofinteractions. Wildl Res 22:11–19

18. Hoogland, JL (1995) The black-tailed prairie dog: social life ofa burrowing mammal. The University of Chicago Press,Chicago, IL

19. Jardine, C, Appleyard, G, Kosoy, MY, McColl, D (2005) Rodent-associated Bartonella in Saskatchewan, Canada. Vector BorneZoonotic Diseases 5:402–409

20. Johnson, WC, Collinge, SK (2004) Landscape effects on black-tailed prairie dog colonies. Biol Conserv 115:487–497

21. Kirillov, M, Markov, A, Lopyrev, I, Pankratova, V, Levitskii, S,Bashkirov, V, Smirnov, G, Kruglov, A, Osadchaya, V, Frolova, G,Barmina, G, Morozova, O, Kosoy, M (2007) Molecular geneticmethods of bartonellae typing. Mol Genet Microbiol Virol(Moscow) 1:8–15

22. Knowles, CJ (1985) Observations on prairie dog dispersal inMontana. Prairie Nat 17:33–40

23. Kordick, DL, Swaminathan, B, Greene, CE, Wilson, KH,Whitney, AM, O'Connor, S, Hollis, DG, Matar, GM, Steigerwalt,AG, Malcolm, GB, Hayes, PS, Hadfield, TL, Breitschwerdt, EB,Brenner, DJ (1996) Bartonella vinsonii subsp. berkhoffii subsp.nov., isolated from dogs; Bartonella vinsonii subsp. vinsonii; andemended description of Bartonella vinsonii. Int J Syst Bacteriol46:704–709

24. Kosoy, MY, Regnery, RL, Tzianabos, T, Marston, EL, Jones, DC,Green, D, Maupin, GO, Olson, JG, Childs, JE (1997) Distribution,diversity, and host specificity of Bartonella in rodents from theSoutheastern United States. Am J Trop Med Hyg 57:578–588

25. Kosoy, MY, Regnery, RL, Kosaya, O, Jones, DC (1998) Isolationof Bartonella spp. from embryos and neonates of naturallyinfected rodents. J Wildl Dis 34:305–309

26. Kosoy, M, Murray, M, Gilmore, RD, Bai, Y, Gage, K (2003)Bartonella strains from ground squirrels are identical to Bartonella

Bartonella in Black-tailed Prairie Dogs 381381

washoensis isolated from a human patient. J Clin Microbiol41:645–550

27. Kosoy, M, Mandel, E, Green, D, Marston, E, Childs, J (2004)Prospective studies of Bartonella of rodents. Part I. Demographicand temporal patterns in population dynamics. Vector BorneZoonotic Diseases 4:285–295

28. Kotliar, NB, Baker, BW, Whicker, AD, Plumb, G (1999) A criticalreview of assumptions about the prairie dog as a keystone species.Environ Manage 24:177–192

29. Maillard, R, Riegel, P, Barrat, F, Bouillin, C, Thibault, D, Gandoin C,Halos, L, Demanche, C, Alliot, A, Guillot, J, Piemont, Y, Boulouis,HJ, Vayssier-Taussat, M (2004) Bartonella chomelii sp. nov.,isolated from French domestic cattle (Bos taurus). Int J Syst EvolMicrobiol 54:215–220

30. Menkens, GE, Biggins, DE, Anderson, SH (1990) Visual counts as anindex of white-tailed prairie dog density.Wildl Soc Bull 118:290–296

31. Norman, AF, Regnery, R, Jameson, P, Greene, C, Krause, DC(1995) Differentiation of Bartonella-like isolates at the species

level by PCR-restriction fragment length polymorphism in thecitrate synthase gene. J Clin Microbiol 33:1797–1803

32. Powell, KL, Robel, RJ, Kemp, KE (1994) Aboveground counts ofblack-tailed prairie dogs—temporal nature and relationship toburrow entrance density. J Wildl Manage 58:351–355

33. Roach, JL, Stapp, P, Horne, BV, Antolin, MF (2001) Geneticstructure of a metapopulation of black-tailed prairie dogs. JMammal 82:946–959

34. Scott, ME, Dobson, A (1989) The role of parasites in regulatinghost abundance. Parasitol Today 5:176–183

35. Stevenson, HL, Bai, Y, Kosoy, MY, Montenieri, JA (2003)Detection of novel Bartonella strains and Yersinia pestis in prairiedogs and their fleas (Siphonaptera: ceratophyllidae and pulicidae)using multiplex polymerase chain reaction. J Med Entomol 40:329–337

36. Ying, B, Kosoy, MY, Maupin, GO, Tsuchiya, KR, Gage, K (2002)Genetic and ecologic characteristics of Bartonella communities inrodents in southern China. Am J Trop Med Hyg 66:622–627

382 Y. Bai et al.