Infection behavior, life history, and host parasitism rates of Emblemasoma erro (Diptera: Sarcophagidae), an acoustically hunting parasitoid of the cicada Tibicen dorsatus (Hemiptera: Cicadidae)

Abstract

Background: ‘Eavesdropping’ parasitoids find their hosts by homing in on the communication signals of other insects. These parasitoids often exploit chemical communication, but at least some species of the sarcophagid genusEmblemasomaeavesdropon the acoustic communications of cicadas. Despite considerable scientific interest in acoustic parasitoids, we know remarkably little about most species of Emblemasoma. To better understand the ecology and behavioral diversity of these flies, I used a combination of field and laboratory techniques to elucidate theinfection behavior and life history of E.erro,which uses the cicada Tibicen dorsatusasa host, and I also investigated parasitoid loads and parasitism rates of T.dorsatus inmultiple host populations in the central United States.

Results: Female E. erro used the acoustic signals of male T. dorsatus as the primary means of locating hosts, but they also required physical movement by the host, usually either walking or flight, to provide visual cues for the final larviposition attack. Larvae were deposited directly on the host’s integument and burrowed through intersegmental membrane to enter the host’s body. On average, E. erro larvae spent 88.0 h residing inside their host before leaving to pupariate, but residence time was strongly dependent on both ambient temperature and effective clutch size. Adult flies eclosed about 18 days after pupariation. Across all study sites, the mean parasitoid load of infected male T. dorsatus was 4.97 larvae/host, and the overall parasitism rate was 26.3%. Parasitism rates and parasitoid loads varied considerably amonghost population samples, and high parasitism rates were usually associated with high parasitoid loads.

Conclusions: Previously, detailed information about the infection behavior, life history, and host parasitism rates of sarcophagid acoustic parasitoids was only available for one species, E. auditrix. This study reveals that the infection behavior of E. erro is quite different from that of E. auditrix and, more broadly, unlike that known for any other species of acoustic parasitoid. The life histories of these two Emblemasoma are also divergent. These differences suggest that sarcophagid acoustic parasitoids are more behaviorally and ecologically diverse than previously recognized and in need of further study.

BACKGROUND

For female parasitoids, successful reproduction usually requiresfinding suitable hosts for their offspring. The problem, of course, is that potential hosts generally do their best not to be found. Sometimes, however, even well-hidden host insects must produce intraspecific communication signals, and these communication sig- nalscan be exploited by specialist parasitoids for use in efficient, long-range host location (Godfray 1994; Zuk and Kolluru 1998; Haynes and Yeargan 1999). Most often, such ‘eavesdropping’ parasitoids intercept chemical communications, but several species offlies (Diptera) from two families, Sarcophagidae and Tachinidae, use acoustic signals tofind their hosts (Cade 1975; Soper etal. 1976; Lakes-Harlan and Lehmann 2015). Acoustically orienting tachinid parasitoids (tribe Ormiini) parasitize crickets and katydids (Orthoptera) (Lehmann 2003), while sarcophagid acoustic parasitoids, which are currently placed in the genus Emblemasoma (sensu Pape (1990)), parasitizecicadas (Hemiptera: Cicadidae) (Soper et al. 1976;Schniederkötter and Lakes-Harlan 2004). Because acoustic signals are often more amenable to experimental manipulation than pheromones, acoustic parasitoids have becomevaluable model organisms for investigating sexual signal exploitation and its consequences (e.g.,Adamo etal. 1995; Allen 1998; Gray and Cade 1999; Müller and Robert 2002; Lehmann and Lehmann 2006; Beckers and Wagner2011).

However, current knowledge of acoustic parasitoidsis heavily biased toward the ormiine tachinids, whichhave receivedthe majority of research (reviewed in Lehmann 2003).In comparison, our understanding of sarcophagid acousticparasitoids is far more limited. Emblemasoma includes16 described species (Pape 1996), but nearly everything known about the basic biology, behaviors, andecology of these flies comes from study of a single species, Emblemasoma auditrix (Shewell), which is a specialist parasitoid of the cicada Okanagana rimosa(Say) (e.g., Lakes-Harlan et al. 2000; Köhler and Lakes-Harlan 2001; Schniederkötter and Lakes-Harlan 2004). No detailed information is available about the in- fectionbehaviors or life histories of any other Emblema- soma,and the only other record of phonotactic behavior comes from a study in which the species of Emblemasoma wasnot determined (Farris et al. 2008). Furthermore, no information about host parasitism rates or parasitoid loads isavailable for any species besides E.auditrix andits host O. rimosa.

Consideringthat Emblemasomaarefrequently refer- enced in discussions of insect hearing andparasitoid biology (e.g., Godfray 1994; Feener and Brown 1997; Yager 1999;Yack 2004; Robert 2005; Hedwig and Robert 2014; Strauß and Lakes-Harlan 2014) and that their ‘ears’havebeen the focus of multiple physiological investigations (Lakes-Harlan et al. 1999; Robert etal. 1999;Farris et al. 2008), it is perhaps surprising how lit- tlewe actually know about the basic biology and ecology of any of these flies besides E. auditrix. As a conse- quence,it is nearly impossible to make meaningful gen- eralizationsabout sarcophagid acoustic parasitoids, and drawing broader conclusions about acoustically orienting parasitoidsin general is similarly difficult.

In2008, I discovered that adults of the cicada Tibicen dorsatus(Say)(Figure 1), a large cicada that is common in the grasslands of central North America (Cole 2008), were sometimes infected with the larvae of a sarcopha- gid parasitoid. After a preliminary investigationrevealed that this parasitoid was Emblemasoma erro Aldrich (Figure2) and that these flies were most likely locating their hosts acoustically, I began a comprehensive study of the basic biology of this fly. The only information previously available about the biology of E.erro wasa recordof a single female fly that had been reared from a specimen of the cicada Quesadagigas (Olivier)in Brazil (Lopes1981) and a report of rearings from Tibicensp.in Texas, USA (Lakes-Harlan2009).

Fig. 1.

Figure 1 Male Tibicen dorsatus, Harvey Co., KS.

Fig. 2.

Figure 2 Female Emblemasoma erro, Ellsworth Co., KS.

Inthis paper, I 1) describe the host locating and larvi-position behaviors of E. erro; 2) describe thisparasitoid’s lifehistory; and 3) report the results of an investigation ofparasitism rates and parasitoid loads in natural popu- lationsof the host cicada, T.dorsatus.I then discuss the following: the infection behaviors of E.erro incompari- son with other acoustic parasitoids and othersarcopha- gid parasitoids, potential host defenses, and the causes of variation in host population parasitism rates and para- sitoid loads, including empirical evidence that superpar- asitism might contribute to high parasitoid loads in somehost populations. The results show not only that the behaviors and life histories of sarcophagid acoustic parasitoids are more diverse than previouslyrecognized, butalso that the infection behaviors of E.erro areunlike those known for any other acoustic parasitoid.

METHODS

Study sites

Surveysof host populations, collections of adult hosts and parasitoids, and field behavioral observations were conductedat six primary study sites located in Ellsworth, Harvey, McPherson, and Reno counties in central Kansas, Hamilton County in western Kansas, and Prowers County ineastern Colorado (Figure 3). The central Kansas sites are located within the Central Great Plains level III ecoregion, while the western Kansas and eastern Col- orado sites are located near the boundary between the High Plains and Southwestern Tablelands ecoregions (US Environmental Protection Agency 2013). All sites consisted ofnative midgrass or shortgrass prairie vege- tation intermixed with riparian, woody vegetation, or planted trees. Supplemental collections ofhost cicadas and adult flies for life history and behavioral study were made atfive additional locations in Kansas with habitat that was similar to that at the primary study sites (Figure 3).

Fig. 3.

Figure 3 Locations of study sites. Filled circles indicate the primary sites used for estimating host parasitism rates, and open circles indicate secondary sites used for additional collections of cicadas and flies. Primary sites are referenced in the text by the counties in which they were located: 1) Harvey Co., 2) McPherson Co., 3) Reno Co., 4) Ellsworth Co., 5) Hamilton Co., and 6) Prowers Co. Both T. dorsatus and E. erro were found at all 11 sites. The inset map indicates the location of the main map in the United States.

Host locating and larviposition behaviors of E. erro

The host locating and larviposition behaviors of E.erro werestudied in three ways. First, natural cicada/fly inter- actionswere observed directly in the field whenever possibleduring the summers of 2010 to 2014. Second, artificialbroadcasts of acoustic stimuli were used to test forphonotactic behavior by female flies in the field; and third, cicada/fly interactions were observed in a con- trolled, outdoor laboratory environment. The methods for these latter two approaches are next described in more detail.

Field broadcasts of acoustic stimuli

Preliminary observations suggested that male cicadas’ acousticsignals played a role in host location by E.erro, but such observations cannot assess whether acoustic cues by themselves are sufficient to attract femalepara- sitoids. To separate acoustic stimuli from other possible sources of information about the location ofpotential hosts (e.g., visual or olfactory), a loudspeaker was used inthe field to broadcast audio that mimicked the calling song of a typical maleT. dorsatus.

Acousticsignals for attracting E.erro weregenerated by gathering audio recordings of calling male T. dorsatus, analyzingthese recordings to estimate the mean values of several acoustic parameters, then constructing model acoustic signals that matched, as closely as possible, the mean calling song of the species. To ensure that the model acoustic signals were broadly representative of T. dorsatusfromthe general study area, I obtained record- ings of 20 different individuals of T. dorsatus from six field sites in Kansas. All recordings were made as un- compressed, 16-bit PCM audio at a sampling rate of 44.1kHz using a highly directional shotgun microphone (Sennheiser ME66 or ME67 with a matched windscreen; Sennheiser Electronic GmbH & Co. KG,Hannover, Germany)and a digital audio recorder (Sony MZ-M200 or PCM-M10; Sony Corporation, Tokyo, Japan). To minimize background noise while also avoidingnear- field acoustic effects in the frequency range of the cica- das’ calling songs (Michelsen and Nocke 1974; Peterson 1980),recordings were typically made with the micro- phoneheld at a distance of between 0.5 and 2 m from thecalling cicada. Most recordings were made directly in the field, but in some cases, cicadas were captured and placed in mesh cages, then recorded once they resumed normal acousticactivity.

Eachrecording was analyzed to determine the values of three acoustic variables: peak frequency, pulse group (PG)length, and PG rate. A ‘pulsegroup’isdefined as a first-order assemblage of sound pulses (i.e., a train of sound pulses) that is separated by silence from the rest of the audio signal. It is the basic unit of temporal struc- ture in the call of T. dorsatus (see Cole (2008) for a spectrogramand oscillogram of the T.dorsatus calling song;note that Cole refers to PGs as ‘syllables’).Peak frequency was estimated by identifying the highestpeak ina power spectral density plot generated by Audacity® (Audacity Team 2012) using a 512-sample fast Fourier transform with the Hann window function. If there were two or more peak frequencies that differed by less than 0.5dB, their average was taken as the overall peak fre- quency. PG length and rate were determinedusing custom-writtensoftware to analyze 10 s of audio from the middle of each calling song recording. Following theseanalyses, a single model acoustic signal was con- structed using the T. dorsatus recording that was as closeas possible to the mean calling song observed for the species (mean peak frequency = 4,308 Hz [s=444]; meanPG length = 20.3 ms [s=1.45]; mean PG rate =37.04PG/s [s=2.25]). I also generated synthetic acoustic signals constructed from amplitude-modulated sine waves that exactly matched the observed meanacoustic variablevalues.

Acoustic signals were broadcast in the field with a custom-built,portable broadcasting system consisting of a12-V audio amplifier and a high-output, horn-loaded tweeterspeaker (PylePro PH44; Pyle Audio Inc., Brooklyn, NY, USA)mounted in the top of a wooden box. Acoustic signals were fed to the amplifier from either a portable CD player or a flash memory-based digital audio player. Broadcastsin the field were conducted either in the late morningor afternoon when cicadas were naturally active, usuallyfor a duration of 4 to 8 min at one time. Flies that wereattracted to the broadcast speaker were captured by hand.

Laboratory observations

Although natural cicada/fly interactions were occasionally observed in the field, opportunities for such observation were unpredictable and infrequent. Furthermore, close- rangeobservation was often impossible, and cicadas could rarelybe captured after an encounter with a fly to deter- mine whether larviposition occurred. Consequently, ob- servation ofcicada/fly interactions in a more controlled setting was also necessary.

Initialattempts to observe the infection behavior of E. erroin2010 used restrained cicadas and audio broadcasts inan approach similar to that used by Schniederkötter and Lakes-Harlan (2004) for their study of E.auditrix. This technique wasunsuccessful for E. erro, however, so further experiments with immobilized cicadas were abandoned.

Instead, unrestrained cicadas and flies were allowed to freely interact in outdoor cages during behavioral experiments in 2012 and 2013. For each infection be- havior trial, one female fly was released into a mesh cage containing one or two uninfected male cicadas. Three types of cage were used: a cylindrical cloth mesh cage approximately 27 cm in diameter and 39 cm high; a larger cylindrical cloth mesh cage approximately 46 cm in diameter and 66 cm high; and a much larger, rectangular screen ‘flight cage’ with asquare base and walls approximately 1.8 mon each side and just over 2.1 mhigh at the center of the top. The behaviors of the fly and cicada(s) were then observed carefully throughout the duration ofthe trial. If the fly appeared to directly contact the cicada with the tip of her abdo- men or otherwise attack the cicada, the cicada was im- mediately removed and inspected for the presence of fly larvae. In most trials, the cicada and fly were not physically disturbed inside of the cage, but in some cases the cicada was induced to flight by the experi- menter to observe the fly’sresponse. Trials ended when afly larviposited upon a cicada or the fly no lon- ger showed interest in the cicada(s) in the cage. To avoid overly stressing the animals, trials were also usu- ally terminated after afly made several attempts to at- tack a cicada even if larviposition was not observed. If the infection status of a cicada could not be deter- mined byvisual inspection immediately after a trial, the cicada was not used in further trials for at least 48h in order to verify whether it had become infected. No single female fly was used to infect more than two cicadas. Whenever a fly larviposited on a cicada, I attempted toimmediately count the number of larvae deposited. This was not always possible, though, and in these cases, the total number oflarvae was determined by rearing the parasitoids or dissecting the host.

Allcicadas used for this part of the study were mature adultmale T.dorsatus thatwere captured directly in the field. Captured cicadas were maintained outdoors in largecloth mesh cages placed over live branches of green ash (Fraxinuspennsylvanica),which provided the cicadaswith a suitable food source. After capture, and before exposing them to parasitoids, all cicadas were closely monitored for up to 9 days to determine whether they had already been parasitized in the field by E.erro. Onlyunparasitized cicadas were used for studying cicada/ flyinteractions. Adult female E.erro wereobtained by broadcasting the model call of T.dorsatus inthe field, asdescribed above, and collecting attracted female flies byhand. Flies were kept in small mesh cages in the labora- tory and provided with sucrose and water adlibitum.

Life history of E. erro

Thetimings of key life history events for E.erro werees- timated by rearing parasitoids from hosts that were nat- urally infected in the field, infected during the behavior studiesdescribed above, or artificially infected in the lab. Toartificially infect cicadas in the laboratory, a female E. errowasfirst anesthetized by chilling the insect at ap- proximately 4°C for several minutes. The fly was then decapitated, placed on a piece of moistened filter paper on a watch glass, and live first-instar larvae were care- fully dissected from the fly’s abdomen. Individual larvae weretransferred to uninfected adult T.dorsatus cicadas usingthe moistened tip of a fine artist’sbrush. Most lar- vae were placed on the intersegmental membrane at the base of the cicada’s wings,but some were placed at the lateral junction of the metathorax and mesothorax or the junction of the metathorax and the first abdominal tergum. A fine insect pin was sometimes used to make a small puncture in the membrane at the wing base in order to facilitate the larvae’sentry into the host’sbody. Larvae from a single female fly were never used to infect more than two cicadas.

Allinfected cicadas were kept in outdoor, mesh cages as described above. Cicadas were checked several times daily,and any individuals that died or appeared mori- bund were moved indoors into small plastic emergence containers to capture emerging fly larvae. Once a host was moved to a larval emergence container, video re- cording was used to capture the precise time and loca- tionof larval egress from the host. If no larvae were observed in an emergence container approximately 48 h afterhost death, or if only undersized larvae emerged, the dead cicada was dissected to check for additional fly larvae.

Parasitoid larvae that emerged from their hostwere provided with fine sand in which to burrow and pupari- ate. Once pupariation was complete, bits of moist paper towels were placed in the emergence containers to help maintainsuitable humidity, and the puparia were kept at roomtemperature (generally 24°C to 28°C) and exposed to the approximate natural daily photoperiod. Theemer- gencecontainers were fitted with screen tops to capture any eclosing adult flies.

Tobetter understand how biotic and abiotic factors in- fluence larval development, I evaluated the effects of two key variables - effective clutch size (the number of larvae from a clutch that successfully develop inside a host) and the ambient temperature experienced by the host - on the total time larval parasitoids spent inside their host (the ‘larvalresidence time’).These variables were chosen because temperature affects the develop- ment and growth rates of insects in general (Harrison etal. 2012), and the number of larvae inside a host mightinfluence how rapidly the host is consumed. For thisanalysis, larval residence time was calculated as the totalelapsed time, in hours, from the moment a larva was deposited on a cicada until the larva emerged from itshost. Effective clutch size was taken as the total num- berof larvae that emerged from a host (because the host cicadas for this analysis were infected in the behavior studies or in the lab, all larvae inside a host were known to be from the same clutch). Temperature was calcu- lated as the mean ambient air temperature experienced byeach parasitized host. Temperature data were taken from the Daymet 1-km daily surface weather dataset (Thornton et al. 1997, 2014). The overall mean ambient air temperature experienced by a parasitized cicadawas estimated by averaging the daily minimum and max- imum temperatures for each day that the cicada was in- fected up to the time of larval egress from the host. The relationship among these three variables was analyzed usingmultiple linear regression with effective clutch size andtemperature as the explanatory variables. To avoid potential non-independence problems caused byrelated larvae sharing the same host cicada, the data were sum- marizedat the level of the host cicada; that is, for each hostwith multiple parasitoid larvae, the mean residence timefor all larvae from the host was used in the analysis instead of the residence time for each parasitoid larva. Diagnosticplots of the standardized residuals were used toverify the fit of the regression model. This, and all otherstatistical analyses, was conducted in R version 3.1.1(R Core Team 2014). Note that larvae from parasit-izedcicadas captured in the field could not be included inthis analysis because it was not known when these lar- vae were deposited on theirhosts.

Host parasitism rates and parasitoid loads

Toobtain population samples for estimating host para- sitism rates, adult T.dorsatus weresurveyed by walking through the habitat at a study site and attempting to capture all T.dorsatus thatwere observed perched in the vegetation or disturbed into flight. Teneral or recently emergedcicadas were excluded because male cicadas do not develop full calling capabilities or begin sexual acous- tic behaviors until several days after eclosion (Maier 1982; B. Stucky, unpublished data). Captured cicadaswere maintainedin captivity to rear the parasitoids from all in- fected cicadas, determine the total number of infected ci- cadas in each sample, and determine the parasitoid load of eachhost, following the methods described above.

Cicadapopulation surveys were conducted at the six primary study sites in July, August, or early September of 2011 to 2014, although not all sites were sampled all 3 years (Table 1). The survey dates were limited by when adult T. dorsatus were actually present in the field, whichvaried from year to year. In 2012, for example, adultT.dorsatus wereabundant in central KS by July 1, butthey did not reach similar abundance in 2013 until the latter half of July.

Table 1 Observed parasitismrates of male T.dorsatus in thefield.

Studysite	Dates	Infected	Uninfected	Total	%infected	95% CI
McPherson Co.	Sitesummary (2012 to 2014)	3	45	48	6.3	2.1to 16.8
	2012:July 2, 4	0	12	12	0.0
	2013:Aug. 2, 20	0	10	10	0.0
	2014:Aug. 3, 12	3	23	26	11.5
ProwersCo.	Sitesummary (2013 to 2014)	27	9	36	75.0	58.9to 86.2
	2013:Aug. 22, 28	11	4	15	73.3
	2014:Aug. 21, Sept. 4	16	5	21	76.2
HamiltonCo.	Sitesummary (2013 to 2014)	10	3	13	76.9	49.7to 91.8
	2013:Aug. 22, 28	6	3	9	66.7
	2014:Aug. 21	4	0	4	100.0
HarveyCo.	Sitesummary (2011 to 2014)	11	51	62	17.7	10.2to 29.0
	2011:Aug. 12	1	9	10	10.0
	2012:July 7, 12	5	13	18	27.8
	2013:Aug. 5, 12	3	14	17	17.6
	2014:Aug. 9, 12	2	15	17	11.8
RenoCo.	Sitesummary (2011 to 2014)	8	65	73	11.0	5.7to 20.2
	2011:Aug. 14, 15	1	3	4	25.0
	2012:July 16, Aug. 17	1	13	14	7.1
	2013:July 23	0	15	15	0.0
	2014:Aug. 4, 11	6	34	40	15.0
EllsworthCo.	Sitesummary (2011 to 2013)	11	23	34	32.4	19.1to 49.2
	2011:Aug. 11	5	3	8	62.5
	2012:Aug. 19	2	1	3	66.7
	2013:Aug. 10, 17	4	19	23	17.4
	Overall	70	196	266	26.3	21.4to 31.9

Observedparasitism rates are given for each study site for all sample years combined, with the yearly observations for each site given below the site summary rows. The overall totals for all sites and years combined are given at the bottom of the table. ‘Infected’isthe number of parasitized cicadas that were captured, ‘Uninfected’isthe number of unparasitized cicadas, ‘Total’isthe total number of cicadas captured, and ‘95%CI’isthe Wilson 95% confidence interval for the ntageof infected male cicadas. Refer to Figure 3for study site locations.

Bythe time the first population samples were collected in2011, I had established that the male cicada’s calling songwas a critical cue used by female parasitoids to lo- cate their hosts. Consequently, surveys from 2011 to 2013 focused on male cicadas only (females do not pro- duce sound) in order to use the limited space available for housing these large insects as efficiently as possible. In2014, both female and male cicadas were sampled at the field sites in Harvey, McPherson, and Reno counties in central Kansas.

Logistic regression (generalized linear models with binomial-distributedresponse and logit link function) was used to evaluate whether host parasitism ratesvar- ied among the field sites and whether sample year or sampledate also influenced parasitism rates. The pro- portionof parasitized cicadas in population samples was modeledwith field site, year, and ordinal sample date as possiblepredictor variables. Both field site and year were treatedas categorical variables. To test the effects of in- dividual predictors and decide which variables to retain in the model, nested models were compared using the differenceof their deviance statistics (i.e., likelihood-ratio tests) (Dobson and Barnett 2008). Standardized residuals plots were examined to check for any problems with model specification. To further assess the final model fit,the likelihood-ratio (also known as McFadden) pseudo R²wascalculated (McFadden 1974; Menard 2000).

Early in this study, it became clear that parasitoid loadsvaried among the population samples. One pos- sible cause of such variation is superparasitism, which, forgregarious parasitoids such as E.erro,is expected to occur more often when unparasitized hosts are rare (Godfray1994). Unparasitized hosts are rare when para- sitismrates are high, so to test for this causal relation- ship,I used simple linear regression to evaluate whether high host parasitism rates corresponded with high para- sitoid loads. For this analysis, each data point was the es- timatedmean host parasitoid load and parasitism rate for a single study site in a given year. Yearly population samples for which fewer than three parasitized cicadas wereavailable to estimate the mean parasitoid load were excluded from the analysis. Parasitism rate was used as the explanatory (i.e., x-axis) variable, and because para- sitismrates were estimated from population samples, some of which were small, there was the possibility of substantial measurement error. Consequently, the re- gression analysis was likely to suffer from slope attenu- ation bias, in which the slope estimator is biased to be less than the true slope (Bulmer 1979; Smith 2009). To compensate for this, I used the sizes of each population sampleto estimate the mean measurement error variance across all population samples. I then used this estimate of theerror variance with the method of moments estimator (MME)of the bias correction factor (Carroll and Ruppert 1996;Smith 2009) to calculate an attenuation-corrected slope estimate. Diagnostic plots of the standardized resid- ualswere used to verify the fit of the simple linear regres- sionmodel.

Additional statistical methods

Allconfidence intervals (CIs) for population proportion estimateswere calculated using the Wilson method (also knownas the score confidence interval) because of its good performance across a broad range of sample sizes (Wilson 1927; Agresti and Coull 1998). CIs for the esti- mates of population means were constructed using the standard t-distribution method when possible (Whitlock andSchluter 2009), but in cases where the population distribution appeared to be non-normal (as determined by examining plots of sample distributions), CIs were calculatedusing the bootstrap-t resamplingmethod with 1,000,000replicates (Efron and Tibshirani 1993; Carpenter andBithell 2000). Bootstrap-tresamplingwith 1,000,000 replicateswas also used to compare the means of non- normallydistributed populations. Throughout this paper, ‘s’ isused to indicate the sample standard deviation.

RESULTS

Host locating and larviposition behaviors of E. erro Host locating behavior

Field and laboratory observations of cicada/fly interac- tionsand field broadcasts of the T. dorsatuscallall con- firmed that E.erro usesthe calling songs of male cicadas as the primary cue for locating potential hosts. In the field, I was able to observe, at relatively close range, the interactions between 14 T. dorsatus cicadas and E. erro flies. All of the cicadas involved in these interactions weremales, 13 of which were acoustically active while I observed them. Seven individual flies were observed in the process of locating a perched cicada (either by flight orwalking), and in every case, the perched cicada was calling while the fly was moving toward it. In the other observed cicada/fly interactions, the flies were already perched near the cicada when I first saw them and I did not observe how, or when, the flies actually arrived near thesecicadas. I never saw flies near perched female cica- das, which cannot producesound.

Duringthe infection behavior trials in outdoor cages in 2012 and 2013, female flies in the experiment cages typically showed an immediate, strong phonotactic re- sponse to calling cicadas. When a cicada in the cage began to call, a fly would either walk towards the calling cicada, fly to within several centimeters of the cicada and then walk towards it, or, in some cases, fly directly to, and land on, the calling cicada. Thus, both field and laboratory observations of cicada/fly interactions pro- vided strong circumstantial evidence that E.erro usethe acoustic calls of cicadas to locate their hosts, but such observations cannot definitively rule out the possibility thatsome other source of information was actually being used,such as visual or olfactory cues.

Field broadcastsof the model T. dorsatus calling song furnished unambiguous evidence that acoustic cues, by themselves,are sufficient to attract female flies. The model T.dorsatus callingsong was broadcast at least once at all six primary field sites. E.erro wereattracted to the broad- castspeaker at every location. Flies often arrived within a fewseconds of the start of a broadcast, and it was not un- common to see multiple flies perched on the top of the speaker box at the same time. At least two or more flies were collected at each primary field site, and more than 60 E.erro intotal were captured during this study. Often, many more flies arrived at the broadcast speaker than could be captured by hand. I did not attempt to quantify the number of flies that arrived and were not captured, however, because individual flies will sometimes arrive atand leave the speaker multiple times during a single broadcast (B. Stucky, personal observation), making any such counts unreliable. The broadcast apparatus never attracted any flies when the loudspeaker was not operating.

Larviposition behavior

A cousticcues and phonotaxis are clearly critical for E. erro to locate its hosts, but the calling song by itself never induced E. erro to larviposit in the absence of a host.Despite the large numbers of flies that were attracted to the calling song broadcasts, no fly larvae wereever foundon the speaker or surface of the speaker box follow- inga broadcast of the model T. dorsatuscall.

Further more, laboratory observations of infection be- haviorsrevealed that, even when a potential host was present,the cicada’scalling song was still not the stimu- lus that ultimately triggered larviposition. Although cicadasoften called during the infection behavior trials in2012 and 2013, in no case did this directly result in lar- viposition by a female fly. Instead, once a fly had moved to within a few centimeters of a calling cicada by orienting to the cicada’s callingsong, it would typically remain more or lessstationary next to the cicada with its head facing to- wardthe cicada’s body.At this point, repeated calls from the cicada usually resulted in relatively little additional movementfrom the fly. However, if the cicada moved ei- ther by walking or flight, the fly usually attempted to maintain its proximity to the potential host. Thus, if the cicadabegan walking, the fly would typically follow it from ashort distance (e.g., 2 to 3 cm away). If the cicada took flight,the fly almost always immediately took flight as well andattempted to follow the cicada in the air.

Sometimes,cicada locomotion resulted in a larviposi- tion attack by the fly, and all evidence suggested that at leastsome movement by the cicada was essential for lar- viposition. Seventeen incidents of successful larviposi- tion were obtained during the infection behavior trials in 2012 and 2013, and in every case, larviposition was only observedwhen cicadas were in motion. I was able to de- termine the moment of larviposition for 15 of these at- tacks. Of these, five (33.3%) occurred while the cicada wasin flight; the remaining ten attacks (66.7%) occurred whilethe cicada was either walking, flapping its wings, or both. In a few cases, the cicadas that were attacked nevercalled at all during the time the fly was in the cage.These cicadas were ‘discovered’bythe flies purely due to their physical movement in the cage, further indi- cating that movement by a potential host, not sound, providesthe visual cues that ultimately trigger larviposi- tion.As a further example, in 2012, I experimented with releasinga female T.dorsatus (whichcannot produce sound at all) in the air in front of a female fly. The fly eventuallyfollowed the cicada and larviposited on it in flight.

Successful larviposition attempts resulted in the depos-itionof one or more tiny first-instar larvae directly upon theexterior of the cicada’s body(Figure 4). A sticky secre- tion also usually accompanied the larvae, presumably to helpthem adhere to the host. Flies appeared to briefly contactthe cicada with their abdomens during larviposi- tion,but high-speed video recording would be needed to revealthe exact mechanics of this process. After larviposi- tion, the larvae immediately began searching for an area ofintersegmental membrane through which to burrow andenter the host’s body.The larvae typically entered the hostquite rapidly, disappearing after anywhere from a matter of seconds to a few minutes.

Fig. 4.

Figure 4 Larviposition by E. erro. A first-instar larva of E. erro on the right fore wing of a T. dorsatus moments after larviposition (larva indicated by blue arrow). The cicada’s head and foreleg are at top center.

Although the location of larvae deposition couldnot bedetermined in all cases because the larvae sometimes disappearedinto the host’sbody before they could be observed,the evidence suggests that flies prefer to attack thebase of a host’swings. Of 15 attacks for which the exact location of larviposition was determined (out of 17 total successful attacks), 1 (6.7%) was on the base of a fore leg, 2 (13.3%) were on the abdomen, and 12 (80%) were either directly on the wings (usually near the base) oron the pterothorax or the first two abdominal seg- mentsnext to the base of the wings. Left/right orienta- tion was recorded for 14 of the 15 attacks, and of these, 9 (64.3%) were on the left side of the cicada’s body,3 (21.4%)were on the right side, and 2 (14.3%) were ap- proximately medial. In their studies of the infection be- haviors of E.auditrix,Schniederkötter and Lakes-Harlan (2004) discovered that E.auditrix preferentiallyattacked the left side of potential hosts. Most laterally oriented at- tacks by E.erro alsooccurred on the left side of the cica- da’s body(9 of 12, or 75%), but this asymmetry was not statistically significant for this sample size (exact bino- mial test, p =0.146).

The number oflarvae deposited by a single female fly on a host cicada during the infection behavior trials (i.e., the clutch size) varied from a minimum of one to a maximum ofsix, but more than 80% of the time, flies (14 of 17) de- posited three orfewer larvae. The mean clutch size was 2.53 (95% bootstrap-t CI: 1.85 to3.45 larvae/host, s=1.50, n = 17 hosts) and the median was 3.

Observations of fly behavior in the field appearedto corroboratethe infection behaviors in the lab. I observed eightmale T.dorsatus thateach produced one or more complete calling songs (as many as five in one case) while an E.erro wasperched next to the cicada. In no casedid the calling song appear to trigger an attempt at larviposition.Just as in the laboratory, flies often waited, nearly motionless, next to calling cicadas, and if a cicada in the field crawled up or down the vegetation it was perched on, the fly usually followed it. If the cicada took flight, the fly usually also took flight and followed the ci- cada in the air.

Asin the cage infection behavior trials, flies in the fieldonly seemed to attack a cicada if the cicada was in motion or had just moved, and most apparent attacks occurred when the cicada was in flight. On two such oc- casions, cicadas that were evidently struck in the air by E.erro hadtheir flight disrupted to such an extent that the cicadas crashed to the ground. Unfortunately, I couldnot determine with absolute certainty whether any ofthe flies I observed in the field actually larviposited on the attacked cicadas. I was able to capture eight cicadas shortly after their interactions with E.erro,but I was un- able to locate first-instar fly larvae on any of them. Con- sidering how rapidly larvae can burrow into their host’s body,it is likely that they had already disappeared from view by the time I was able to look for them. Parasitoids were reared from all eight cicadas, though, so it is very likely that at least some became infected during the ob- served cicada/fly interactions.

Life history of E. erro

Fromthe moment of larviposition until they completely exited the host’sbody, E.erro larvaespent, on average, 88.0h residing inside their host (95% CI: 81.19 to 94.76h, s=17.1, n=27 larvae from 13 host cicadas and 10female flies, range = 61.3 to 116.0). Multiple regression analysis ofthese data revealed that both temperature and effective clutch size had significant effects on lar- val residence time. Together, these two variables ex- plained more than 93% ofthe observed variation in residence times (R2=0.934; pvalues for the coeffi- cients ofboth explanatory variables were <0.002). In- creases in either ambient temperature or the number of larvae in a host were associated with a decrease in residence time (the estimated relationship was residence_ time=211.2 −4.14temperature−5.50effective_clutch_ size)(Figure 5). By the time all larvae left an infected host, it was common to find all soft tissues inside the cicada’s bodyentirely consumed so that nothing but the exoskel- etonremained.

Fig. 5.

Figure 5 Relationship of effective clutch size and temperature to larval residence time. Each data point represents the mean residence time of the parasitoid larvae inside a single host cicada along with the effective clutch size (number of larvae emerging from the host) and the mean air temperature experienced by the host during parasitoid development. The planar surface represents the multiple linear regression model of the effects of temperature and effective clutch size on larval residence time. Lines connected to the data points indicate the vertical distance of each data point from the regression surface (i.e., the residuals).

Toexit their host, larvae used their oral hooks to bur- rowthrough intersegmental membrane, and they usually emergedby squeezing between one of the cicada’s oper- culaand its abdomen (Figure 6). The exact location of egress was observed for 83 larvae from 28 T. dorsatus hosts,and of these, 64 (77.1%) exited from behind one of the cicada’s opercula. Of the remaining larvae, 16 (19.3%)exited next to the pygofer or terminal abdominal segments at the apex of the abdomen, and 3 (3.6%) bur- rowed through the membrane between the head and prothorax.

Fig. 6.

Figure 6 Emergence of E. erro from its host. A mature larva of E. erro emerges from between the left operculum and the abdomen of a deceased male T. dorsatus from Prowers Co., CO.

Afterleaving their host, larvae immediately burrowedinto the soil (or sand, in the case of theemergence containers)to pupariate. Although more than 300 E.erro larvae were obtained from infectedT. dorsatus speci- mensduring the course of this study, relatively few of these were successfully reared to the adult stage. Fifty- three flies survived to adulthood, and the times of both larval egress from the host and adult eclosion were ob- tained for 31 of these flies. Adult flies eclosed 18.4 days, on average, after leaving their host (95% CI: 18.02 to 18.69days, s=0.91, n=31 flies from 15 host cicadas, range = 16 to 20 days). The lifespan of adult flies in the field is unknown. Adult flies maintained in the labora- tory survived as long as 92 days.

Thelifetime reproductive potential of female E.erro wasnot determined, but I did dissect 14 gravid female fliesthat were collected at audio broadcasts of the T. dorsatuscallingsong in 2013 and 2014 and counted all larvae contained within their abdomens. These flies car- riedas few as 3 and as many as 174 larvae in their incu- batorypouches, with a mean of 60.7 larvae per fly (s= 57.5). The observed distribution of larvae counts was strikinglybimodal: Three flies had more than 150 larvae, whileall of the rest had fewer than 80. The larvae of the four flies with the largest larvae counts were noticeably smaller than those from the remaining flies and gener- ally had less well-developed bristles. Remnants of egg- shell were still visible in the incubatory pouches of three of these flies, suggesting that the larvae had recently hatched.

Host parasitism rates and parasitoid loads Parasitism rates

Theresults of the T.dorsatus populationsurveys for the prevalence of E.erro infectionare presented in Table 1. Parasitized T. dorsatus were collected at all six of the primary study sites, although infected cicadas were not detected in all population surveys. All parasitoids that were reared to the adult stage were identified as E.erro, and the morphologies of all other larvae and puparia thatwere obtained were also consistent with E.erro.No hyperparasitoids of E. erro wereobserved.

Acrossall four sampling years (2011 to 2014) and all six primary study sites, the overall observed parasitism ratefor T.dorsatus maleswas 26.3% (95% CI: 21.4 to 31.9%, n=266 cicadas). The surveys in 2014 also in- cluded a sample of 28 female T.dorsatus fromthe cen- tral KS field sites (in Harvey, McPherson, and Reno counties),and of these, one female cicada was infected with E. erro larvae (3.7%; 95% CI: 0.7% to 18.3%).

Therewas substantial variation in observed parasitism rates among the population samples (summarized in Table 1). The results of the logistic regression analysis suggested that much of this variation was due to differencesamong field sites, with sampling year possibly also havinga small effect (likelihood-ratio tests of field site andyearaspredictors: p<0.00001 and p=0.0663, re- spectively). The model including these two predictor variables seemed to explain the data reasonably well, withpseudo R2=0.549. Nevertheless, this result must be interpreted with caution. The westernmost field sitesoften had the highest sample parasitism rates, but be- cause of logistical constraints, these sites were always surveyed later in the summer than the other field sites (Table 1). Thus, the variable field site was at least par- tiallycollinear with sampledate.Consequently, the high parasitismrates observed at these sites could have been due,at least in part, to seasonal effects rather than in- herent site differences.

Parasitoid loads

The mean parasitoid load of all field-collected infected cicadaswas 4.97 larvae/host (95% bootstrap-tCI:4.23 to 5.92larvae/host, s=3.95, n=91 hosts, range = 1–19lar- vae/host) and the median was 4, reflecting the strong rightskew of the distribution (Figure 7).

Fig. 7.

Figure 7 The distribution of parasitoid loads (larvae per host) of infected cicadas in the field.

The parasitoid loads of field-collected infected cicadas wereoften much higher than the clutch sizes of larviposit- ing females in the laboratory infection trials. A bootstrap-tcomparisonof means confirmed that the mean clutch size of female parasitoids (2.53 larvae/host) was significantly lessthan the mean parasitoid load of hosts in the field (4.97larvae/host) (95% bootstrap-tCI:1.32 to 3.52 fewer larvae/host,p<0.0001).

Overall,there was a strong, positive relationship be- tween host cicada parasitism rates and meanparasitoid loads per host (Figure 8), with parasitism rate explaining about 65% of the variation in mean parasitoid load (simplelinear regression: b=8.29, R²=0.650, p=0.0048). The estimated slope of the relationship was 8.29, but the imprecision of the parasitism rate estimates meant that this slope estimate likely suffered from attenuationbias. The estimated bias correction factor was approximately 1.117, giving an attenuation-corrected slope estimate of9.25.

Fig. 1.

Figure 8 Relationship between host parasitism rate and mean parasitoid load per host. Each data point represents 1 year of host population sampling data for a single study site. The solid line (blue in the color figure) represents the linear regression model for the data.

DISCUSSION

Theresults of this study provide the first detailed infor- mation about the infection behaviors and life history of any species of Emblemasoma besides E. auditrix. Both laboratoryand field observations reveal that E.erro find their hosts by eavesdropping on the sexualcommunica- tionsignals of male cicadas. When a female E.erro lo- catesa calling cicada, she waits to attack until the host is in motion, and larviposition on flying cicadas is not uncommon. The results also show that male T.dorsatus arecommonly parasitized by E.erro andthat there can be substantial variation in population parasitism rates andparasitoid loads. I next discuss the behavior and life history of E. erro, especially in comparison to other acoustic parasitoids and other sarcophagid parasitoids; assess possible host defenses; and discuss possible causes ofvariation in host parasitoid loads and parasitism rates.

Host locating and infection behaviors of E. erro

E.erro’s useof phonotaxis to locate potential hosts is similarto that reported for other acoustically hunting parasitoids (Soper et al. 1976; Lehmann 2003; Lakes- Harlan and Lehmann 2015), but E.erro’s preferencefor attacking moving targets is apparently unique among known acoustic parasitoids. For example, E.auditrix,the only other Emblemasoma for which larviposition behav- iors are known, will aggressively attack stationary or restrainedcicadas. Upon finding a male cicada, female E. auditrixexhibita stereotyped behavioral sequence in which the female fly immediately attempts tosqueeze underneath the perched cicada’s wings to gain accessto the cicada’s timbalregion. She then uses specialized ter- minalabdominal sternites to cut through the cicada’s timbal membrane and injects larvae directly into the host’s body (Schniederkötter and Lakes-Harlan 2004). Female E. erro lack any comparable abdominal modifica- tions, but larvipositing through the host’s timbal would likelybe impossible for E.erro anyway,because male Tibicendorsatus havetimbals that are fully protected by well-developed timbal covers. In contrast, E. auditrix’s hostcicada, O.rimosa,lacks timbal covers entirely.

Tachinid acousticparasitoids of the tribe Ormiini will alsoattack stationary hosts, and they will even larviposit without visual or tactile confirmation of a host’slocation. For example, Homotrixa alleni Barraclough,Ormia depleta (Wiedemann), and O. ochracea (Bigot) will all depositlarvae at a sound source regardless of whether or not a potential host insect is actually present (Cade 1979; Fowler 1987; Allen et al. 1999). For E. erro, the host’scalling song was never sufficient by itself to trigger larviposition, even when a potential host was present. In contrastto E.erro,ormiine tachinids are all nocturnal parasitoids of Orthoptera, and their willingness to larvi- positin the absence of a host probably reflects an almost totalreliance on acoustic cues at night. For acoustic par- asitoids such as E. erro that are active during theday, requiring visual confirmation of a suitable host prior to larviposition allows for more precise placement oflarvae andundoubtedly decreases the number of larvae that are wasted by the female fly.

Incomparison to the larviposition behaviors of other acousticparasitoids, E.erro’stendency to attack flying ci- cadasis especially striking. One third of the successful attacks observed in the experiment cages took place while the cicada was in flight, but this is almost certainly an underestimate of the true frequency of flight-based attacks in nature. Due to the size of the cages used in the trials, most attempts by flies to follow cicadas in the air resulted in failure because the cicada crashed into a side of the cage before the fly could approach and orient itself to the flying cicada. It was hoped that the large’- flightcage’wouldalleviate this problem, but even it ap- peared to be too small for most aerial attacks to succeed. Nevertheless,flies seemed much more reluctant to attack potentialhosts that were not in flight.

Thisconclusion is further supported by observations in the field, where nearly all apparent larviposition at- tacks occurred while cicadas were in flight. Flies some- times even followed a single cicada from perch to perch, waiting patiently next to the cicada each time it landed, but never attempting to attack while the cicada was not flying. As an example, in 2013, I observed a male T.dor- satus callingfrom a grass flowering culm with a female E.erro perchedon the opposite side of the stalk near the cicada’s abdomen.The fly was nearly motionless until thecicada backed a short distance down the stalk, caus- ingthe fly to move with him nearly in unison, but the flymade no move to attack the cicada. When the cicada flew a short distance (approximately 1 to 2 m) to a new perch, the fly closely followed him in the air, landed next to the cicada, and again remained nearly motionless while the cicada began calling. The cicada flew twice more, with the fly following both times, and after the final flight of at least 30 m, I captured the cicada and later reared two E.erro larvaefrom it.

WhileE.erro’s behaviorof larvipositing on hosts while theyare in flight or otherwise in motion might be different from E. auditrix and tachinid acoustic parasitoids, it is remarkablysimilar to the larviposition behaviors reported for some sarcophagid parasitoids of the genus Blaesoxipha that parasitize acridid grasshoppers.B. aculeata (Aldrich), B. caridei (Brethes), B. kellyi (Aldrich), B. redempta (Pandellé),and B.reversa (Aldrich),among others, have all been reported to attack grasshoppers while in flight (Coquillett 1892; Kelly 1914; Aldrich 1916; Lloyd 1951; Rees1973; Povolný and Verves 1997). Kelly (1914) pro- videda detailed description of the larviposition behaviors ofB.kellyi,reporting that grasshoppers were attacked ei- theron the wing or on the ground, and that grasshoppers wereonly attacked when they were in motion (but see Smith 1915). Furthermore, both B. kellyi and B. reversa typicallyplace larvae near the base of a host’s wings,much likeE. erro(Kelly1914; Rees 1973).

It is worth noting that early last century, Beamer (1928)and Kelly (1914), both working in Kansas, reported seeingcicadas pursued by flies while in flight. Beamer noted that ‘theflies follow but a few inches away, and sometimes seem almost to alight on the body of the ci- cada.’ Although their observations were largely adventi- tiousand incidental, and neither author identified the flies involved, it seems plausible in retrospect that their papers might have been the first published records of E. erro’s host infectionbehavior.

Infection of female hosts

GivenE.erro’sprimary host-finding mechanism, male ci- cadasare clearly the primary targets of infection by this parasitoid.However, the observation of a fly larvipositing ona female cicada in the laboratory, along with the 2014 survey of female T. dorsatus in the field, confirms that femalecicadas are also sometimes attacked.

Sincefemale cicadas are silent, how are they discov- ered by E. errointhe field? One possibility is that, sim- ply by chance, they happen to fly within the visual range of a perched female E.erro.Perhaps more likely, though, female T. dorsatus and female E. erro might sometimes encounter one another while seeking male cicadas. Like E.erro,female cicadas perform phonotaxis in response tomales’calls,so female cicadas could become parasit- ized if they were attracted to the same calling male as a female E.erro. Inany case, despite many hours spent observingcicadas in the field, I never witnessed any inter- actionsbetween female E.erro andfemale T.dorsatus,so suchencounters must be rare in comparison to encoun- tersbetween male cicadas and female E.erro.However, E. erro’s occasionaluse of female hosts is not unique. Several other species of acoustic parasitoids that primarily attack male hosts are also known to sometimes parasitize females (Soperet al. 1976; Lehmann 2003).

Phenology and fecundity of E.erro

Littleis known of the seasonal phenology of E.erro.In this study,adult flies were observed in the field as early as June 13(in 2012) and as late as September 4 (in 2014), and these were also the earliest and latest dates that I attemptedto find them. The rearing data strongly suggest thatE.erro ismultivoltine in the geographic area covered bythis study. With a total development time from larvipo- sition to adult eclosion of about 22 days, it seems possible thatthere could be at least three generations per year. E. auditrix,in contrast, is apparently univoltine (Soper et al. 1976; de Vries and Lakes-Harlan 2005).

FemaleE.erro wereobserved with as many as 174 first-instarlarvae, nearly 3.5 times the maximum of 50 observed for E. auditrix (De Vries and Lakes-Harlan 2005).The apparently large difference in fecundity be- tween these two species might be at least partially ex- plainedby their larviposition behaviors and life histories. E. auditrix deposits larvae directly inside a host’s body, onelarva per host, and all available evidence suggests that E.auditrix isa solitary parasitoid (Soper et al. 1976). By injecting larvae into its hosts, E.auditrix likely‘wastes’ relativelyfew larvae during larviposition, and as a solitary parasitoid, it is plausible that multiple larvae inside a sin- glehost would physically attack one another (Godfray 1994). Under these conditions, females might benefit by producingfewer, larger larvae to increase their chances of survival. In contrast, because E.erro depositsits larvae on theexterior of a host, it is likely that some percentage of these larvae never manage to make it inside the host’s body.Moreover, E.erro isa gregarious parasitoid, and as such,larvae probably face little direct physical aggression from conspecifics (Godfray 1994). For E.erro,then, invest- ingfewer resources in more larvae might increase a female’s lifetime reproductive success. Some tachinid acoustic parasitoid species, which deposit their larvae even morehaphazardly, also have large larval complements (Wineriter and Walker 1990; Allen et al. 1999; Kolluru andZuk 2001), and although behavioral data for other sar- cophagid parasitoids is extremely limited, at least some parasitoidspecies in the genus Blaesoxiphaalsoappear to followthis pattern (Middlekauff 1959).

Host defenses and mortality

Once discovered by a female E. erro, male T. dorsatus appearedto have relatively few viable options to defend themselves.When approached by a parasitoid fly, calling maleT.dorsatus cicadasresponded either by flying, im- mediately terminating their call and remaining motion- lesson their perch (hereafter referred to as ‘hiding’),or simply continuing their calling behavior. The latter seemed to be the most common. Cicadas often called re- peatedly and walked freely about the walls of the experi- ment cages despite being followed by a fly only a few centimeters away. Cicadassometimes even called with a flyperched right on top of them. However, stationary ci- cadas that were directly contacted by a fly would often vigorously flick their wings to try to repel the parasitoid. Unfortunately,given the relatively small space inside the cages, evaluating the effectiveness of any of these behav- iors was nearly impossible because a cicada could never truly escape from thefly.

Nevertheless, observations in the field suggested that boththe flight and hiding strategies do sometimes work. Inat least one case, a fly lost interest in a hiding cicada and left before the cicada resumed calling, and in an- other, a cicada that was contacted by an approaching fly managed to escape by flying away. Most of the time, though, flies simply waited until a hiding cicadabecame active again, and they usually had little difficulty in fol- lowing aflying cicada from one perch to another. As a defensive strategy, flying seems especially risky given E. erro’s aptitude for aerial larviposition.

After being larviposited upon, cicadas had yetanother optionfor defending themselves. I repeatedly observed cicadas perform ‘wing flipping’ behavior immediately afterbeing attacked, characterized by rapidly flapping their wings several times while perched. In this way, one cicada managed to completely dislodge the single larva that had been deposited on the cicada’s rightfore wing, thus avoiding infection completely. This was the only case for which I confirmed that a cicada was able to re- move all larvae from its body, but it is possible that some larviposition events were not detected during the behavioral experiments. Wing flipping by T.dorsatus ap- pearsto be functionally similar to the grooming behav- iors used by the cricket Gryllus texensis Cade and Otte to prevent infection by the larvae of Ormia ochracea (Vincent and Bertram2010).

Although the hosts of some other sarcophagidparasit- oids have been reported to occasionally survive parasit- ism (Spencer and Buckell 1957; Danyk et al. 2000), infection by E. erro appears to be invariably fatal forT. dorsatus.In most cases, hosts died several hours before the parasitoid larvae emerged. Host death was usually preceded first by loss of wing function, then loss of leg function beginning with the hind legs and ending with the fore legs. Prior to death, a cicada’s antennaewere typicallythe last appendages to display a visible response toexternal touch. After a cicada died, small, rhythmic movements of the legs or head capsule were often visible as the parasitoid larvae used their oral hooks to scrape muscle and other soft tissue from the integument.

Sometimes,though, when a cicada was infected with only a single larva, the larva emerged before the cicada died, leaving the host in a severely weakened, moribund state.Cicadas in this condition usually succumbed after afew hours. In one exceptional case, a large male T.dor- satus from the Prowers Co., CO site that was infected with a single E. erro larva survived for more than 24h following parasitoid emergence. Although sluggish, it was still able to cling to and crawl on a perch, weakly flutter its wings (but not fly), and was even observed attempting to feed before its movements became unco- ordinated and it, too, died. Overall, E. erro must be a major cause of mortality for adult male T. dorsatus,es- pecially considering the very high parasitism rates ob- served in some cicada populations.

Variation in host parasitism rates among study sites

Hostpopulations at the two westernmost field sites ap- pearedto have consistently higher parasitism rates than sitesfurther east (Table 1, Figure 3). The biogeography of potential host cicadas might offer one explanation for this pattern. The western sites were located on the semi- arid High Plains, where there are fewer species of large cicadapresent than on the more mesic midgrass prairies ofthe study sites further east. E.erro parasitizesother ci- cada species besides T.dorsatus (B.Stucky, in prep.), so higher parasitism rates of T.dorsatus onthe High Plains could be a consequence of local differences in the com- munities of potentialhost species.

However,as noted in the ‘Results’ section, because these western sites were also sampled later inthe sea- son than the eastern sites, higher parasitism rates could have also been caused by seasonal effects rather than intrinsic differences among the sites. One might expect parasitism rates toincrease throughout the sea- son as E. erro populations reach their peak and host populations decline, as has been observed for several other species ofdipteran parasitoids, including some acoustic parasitoids (e.g., Tamaki et al. 1983; Allen 1995; Lehmann 2008). It seems likely that this ac- counts for at least some of the among-site differences in parasitism rates found in this study. Furthermore, both host andparasitoid population sizes undoubtedly also play arole in determining parasitism rates. As evi- denced by some of the small population sample sizes, host cicadas were uncommon and difficult to collect for some years at some field sites, which suggests that there was variation in host population sizes from year toyear. Future studies that estimate host and parasitoid populationsizes and sample both High Plains and cen- tral Plains sites multiple times throughout the season will be needed to fully disentangle the effects of these variables on host parasitism rates.

Superparasitism by E. erro

Thestrong, positive relationship between parasitism rate andparasitoid load (Figure 8), as well as the significant difference between the mean parasitoid load of field- collected hosts and the mean clutch size of larvipositing females (4.97 and 2.53 larvae/host, respectively), can both be explained as a consequence of superparasitism inthe field. If at least some host cicadas are superparasi- tizedin the field, then we should expect the mean para- sitoid load of host cicadas to be larger than the mean clutch size of individual female flies. Furthermore, for gregariousparasitoids such as E.erro,superparasitism is expectedto be more common when unparasitized hosts arerare, simply because female parasitoids have a harder timefinding hosts that have not already been infected (Godfray1994). Unparasitized hosts are rare when para- sitismrates are high, so higher parasitism rates should correspond with increasing rates of superparasitism.In- creased superparasitism would, in turn, likely result in larger parasitoid loads per host, which means that higher population parasitism rates should correspond with higher parasitoid loads. This prediction matches the pattern of thedata quite well (Figure 8).

Additionally, anecdotal evidence of superparasitism wasfound in the relative sizes of larvae emerging from someof the most heavily parasitized hosts. In some cases,two distinct larval size classes were evident, pre- sumablydue to the smaller larvae having been deposited onthe host later than the larger larvae. In other cases, though, all larvae emerging from heavily parasitized hosts were approximately the same size, suggestingthat eithera single female deposited all of the larvae at once, or more likely, that two (or more) female flies discovered an uninfected host at nearly the same time.

CONCLUSIONS

E.erro isa widespread, common parasitoid of the cicada T. dorsatus on the grasslands of the Great Plains inthe central United States. Female flies locate potentialhosts byeavesdropping on the acoustic mating calls of male cicadas, then use visual cues to larviposit on the host while it is in motion. Larviposition often occurs while the cicada is in flight. Parasitization by E.erro isalways fatal for T. dorsatus, which seems to have few consist- ently effective defenses against attack. Parasitism rates formale T.dorsatus canexceed 70% in some host popu- lations.Parasitoid loads of infected cicadas average about fivelarvae per host but can be as high as 19 larvae per host.At least some variation in parasitoid loads is likely due to superparasitism in host populations with high parasitism rates.

Eventhough E. errois,like E. auditrix,an acoustically orienting parasitoid of cicadas, the close-range infection behaviorsof these two species are highly divergent, and the infection behaviors of both species are very different from tachinid acoustic parasitoids of the tribe Ormiini. Indeed,the infection behavior of E.erro isunlike that knownfor any other acoustic parasitoid. There are im- portant life history differences between E.erro andE. auditrixaswell: E.auditrix isapparently a solitary, univol- tineparasitoid with relatively low larval production per fe- male; E. erro is a gregarious, multivoltine parasitoid with highlarval production per female. Given the marked dif- ferencesbetween E.erro andE.auditrix,the results of this study suggest that more work is needed to characterize the diversity of Emblemasoma parasitoids. An improved understandingof sarcophagid acoustic parasitoids would make it possible to more meaningfully compare sarcopha- gid and tachinid acoustic parasitoid lineages, and it would also allow for more robust inferences about eavesdropping parasitoids ingeneral.