Electropenetrography with Alternating Current Reveals In Situ Changes of Aedes aegypti Probing Behaviors Associated with Dengue Virus Infection
Human infection with dengue virus (DENV) results in significant morbidity and mortality around the world. Current methods to investigate virus-associated changes in insect feeding behaviors are largely restricted to video analysis of feeding events outside of the host or intravital microscopy. Electropenetrography, a method originally developed for plant-feeding insects, offers a promising alternative by allowing high-resolution recording of voltage changes across the insect bite interface. We compared recordings from DENV-infected Aedes aegypti mosquitoes feeding on uninfected mice and uninfected A. aegypti feeding on DENV-infected mice to controls of uninfected A. aegypti feeding on uninfected mice. We found significant mosquito behavioral changes in both DENV-infected groups compared with controls including longer feeding times and longer preingestion probing events for A. aegypti feeding on DENV-infected mice and a higher number of sequential probing events in DENV-infected A. aegypti feeding on uninfected mice. By recording mosquito feeding and probing events beneath the surface of the skin, we have been able to both confirm and add new dimensions to previous findings regarding DENV-associated behavior changes in A. aegypti. This provides a foundation for increasingly in-depth studies focusing on the transmission of the DENV between vectors and hosts.
Introduction
The Aedes aegypti mosquito, recognized as the primary vector for dengue virus (DENV), plays a critical role in the transmission of dengue fever, a significant global health concern. Aedes-borne DENV represents a substantial public health challenge, affecting nearly half of the world’s population and leading to an estimated 96 million cases of clinically manifested DENV infection each year.5,7 The complexity of DENV transmission is further compounded by the intricate behaviors of its mosquito vectors, particularly in terms of their probing and ingestion activities, which are crucial for the virus to move between vertebrate hosts.
Previous work has demonstrated that A. aegypti infected with DENV behave in different ways that may promote virus transmission. Previous studies have demonstrated that A. aegypti infected with DENV have increased locomotion,29 longer feeding periods,34 and an increased motivation and avidity to feed.30 The tropism of DENV for nervous tissue in A. aegypti has been demonstrated and is a hypothesized mechanism for the observed behavioral changes in this arthropod-borne virus (arbovirus) system.34
Despite considerable progress, barriers remain to quantifying changes in mosquito ingestion behaviors. While published methods exist for intravital microscopy of mosquito bites on mice,10,19,41 these methods have a limited ability to scale or be easily adapted to varied vector-pathogen-animal systems. Methods of external observation32,34–36 and videography44 are fundamentally limited in their precision because they are based on externally visible indications of feeding and the binary factor of whether the stylets are visible within the skin. Efforts to develop vaccines6 and chemical therapeutics38 against mosquitoes that spread pathogens of human and veterinary importance rely on understanding the intricacies of the arthropod feeding site. This gap in knowledge represents a critical barrier to developing more effective public health interventions for mosquito-borne diseases and drug-based vector control strategies.
The recently validated use of electropenetrography (EPG) with alternating current as a methodological approach in mosquito bite studies with A. aegypti43 and, specifically, on mice42 offers the opportunity to explore in situ feeding behavior changes of arboviral systems in detail. The EPG system records changes in a small, applied voltage across the insect-host interface, and waveform families and types are named following established conventions43 (Figure 1). These voltage changes over time produce waveforms that reflect the overall electrical resistance of the insect-host system. Although histologic correlations between waveforms and stylet placement in this species have yet to be accomplished, basic inferences from these characteristic waveforms may be made including when ingestion of liquids, including blood, begins (waveform family M) and the period after the insertion of the stylets but before ingestion (waveform family L) when the mouthparts are presumably in the interstitial space between the skin surface and the targeted blood vessel.
Citation: Comparative Medicine 74, 4; 10.30802/AALAS-CM-24-030
By allowing real-time, detailed observation of mosquito probing activities, EPG has the potential to shed light on the subtle behavioral changes induced by DENV infection, thus providing valuable insights into the dynamics of virus transmission. Similar behavioral studies have been performed for decades in agricultural insect-pathogen systems with dependable results.2,22,23,39 We hypothesize that recordings of DENV-infected A. aegypti feeding events on mice made with a modern EPG system with alternating current will display significantly more probes per event and feed for longer than uninfected controls. To our knowledge, this study expands the use of modern alternating-current EPG systems to the investigation of arboviral systems for the first time.
Materials and Methods
Virus.
DENV type 2 (strain S-14635, reference NR-3796) was obtained through the Biodefense and Emerging Infections Research Resources Repository, National Institute of Allergy and Infectious Disease, National Institutes of Health. The virus was propagated by sequential passage first in baby hamster kidney fibroblasts stably expressing human DC-SIGN (BHK-DC) followed by the Aedes albopictus cell line C6/36. Titers from plaque assays and laboratory standards were correlated with quantitative RT-PCT (RT-qPCR) results to establish the plaque-forming unit equivalent (PFUe) of given cycle threshold values. Harvested virus stocks were concentrated by centrifugal ultrafiltration and assayed by RT-qPCR to determine virus concentration for infections. The RT-qPCR assay used was a previously described multiplex assay26 run as a single-plex assay with only the DENV-2 primers and probes. The iScript One-Step RT-PCR Kit for Probes (Bio-Rad, Hercules, CA) was used for all amplification reactions following the manufacturer’s recommendations and published assay specifications26 on a Bio-Rad CFX-96 thermal cycler.
Mosquitoes.
The Rockefeller strain of A. aegypti was used for all experiments. These mosquitoes served as the experimental unit for all studies. This strain is highly adapted to the laboratory environment and routinely used for arboviral experiments.1,27 Mosquitoes were maintained at 25 to 27 °C with a relative humidity of 75% to 80%. Colonies were maintained on a 16:8 (light:dark) photoperiod and housed in 30 × 30 × 30-cm plastic cages with 10% sucrose, ad libitum, until use. Blood-naïve, 10 to 12 d posteclosion females were used for EPG recordings.
Regardless of infection status, all mosquitoes were anesthetized on ice for a maximum of 10 m and allowed to fully recover for 3 to 4 h before experimental use to limit anesthesia-related behavioral differences. The wiring of mosquitoes for EPG followed previously published methods42 with minor modifications. Briefly, a vacuum aspirator was used to immobilize the female mosquito. A fine gold wire was glued to the pronotum of the mosquito with a mixture of silver flake and diluted white glue and allowed to dry. This gold wire was silver glued to an insect stub constructed of a brass-plated escutcheon pin to which a copper wire had been soldered. Specific protocol changes included wiring the DENV-infected mosquitoes in a glove box for enhanced arthropod containment as well as enhanced biosafety practices. Briefly, these included double-containment for transport and housing of insects, enhanced personal protective equipment, limited laboratory access, and the use of a self-contained, HEPA-filtered vacuum pump with a blunt metal needle to aid the wiring process. In addition, recording sessions using mosquitoes with confirmed dengue dissemination were limited to a smaller number of individuals (n = 5 to 6) for each session compared with sessions using uninfected mosquitoes (n = 10 to 12). All tethered mosquitoes were able to fly and walk while wired and were allowed 3 to 4 h to recover from the wiring process before recording.
Mammalian host.
Inbred congenic 4- to 20-wk-old B6.129S2-Ifnar1tm1Agt/Mmjax (IFN-aBR−) female (n = 15) and males (n = 12) mice (MMRRC-JAX; Strain Code 032045-JAX) were used in this study. Mice were assigned to one of 3 groups described below. Due to the immunodeficient nature of this strain, Jax Laboratories verifies mice are pathogen free of the following: Sendai virus, pneumonia virus of mice, mouse hepatitis virus, minute virus of mice, mouse parvovirus, murine norovirus, Theiler disease virus, reovirus, rotavirus, lymphocytic choriomeningitis virus, ectromelia, mouse adenovirus of mice, murine cytomegalovirus, hantavirus, K virus, LDH elevating virus, Murine chapparvovirus, mouse thymic virus, polyoma virus, Bordetella spp., Citrobacter rodentium, Corynebacterium kutscheri, Corynebacterium bovis, Filobacterium rodentium, Helicobacter spp., Klebsiella spp., Mycoplasma pulmonis, Pasteurella multocida, Rodentibacter spp., Pseudomonas aeruginosa, Pneumocystis murina, Proteus mirabilis, Trichomonads, Yersinia spp., Salmonella spp., Staphylococcus aureus, Streptobacillus moniliformis, Streptococcus pneumoniae, β-hemolytic Streptocuccus spp., Tyzzer’s disease, dermatophytes, and all ectoparasites and endoparasites.
Sentinel testing was not performed since this was a short-term study. Animals were maintained in an AAALAC-accredited facility in accordance with the Guide for the Care and Use of Laboratory Animals in a dedicated housing room.25 All biohazard agents and procedures for animal use were approved by the Tulane University IACUC (protocol no. 1484) and the Institutional Biosafety Committee (protocol no. 2201609).
Mice were group housed by sex and maintained in static polysulfone autoclavable microisolation mouse cages (186-mm width × 298-mm length × 128-mm height; Allentown, Allentown, NJ) containing hardwood maple bedding (no. 7090, Sanichips; Harlan Teklad, Madison, WI; 2 bedding changes weekly) and shredded paper nesting material (Bed-r’Nest; The Andersons, Maumee, OH). Cages were autoclaved and changed aseptically twice weekly. Mice were maintained on a 12:12 light-dark cycle.
Mice were housed in the Tulane University Department of Comparative Medicine vivarium for at least 1 wk before the start of the study with ad libitum access to acidified water and rodent chow (reference no. 5053 Irradiated Laboratory Rodent Diet; Purina, Richmond, IN). Mice were ear tagged, and weight was recorded before the start of the experiment.
Electropenetrography.
EPG recordings were undertaken using an AC-DC EPG monitor system (EPG Technologies, Gainesville, FL) and WINDAQ software (DATAQ Instruments, Akron, OH). A previously described experimental setup for safely conducting EPG recordings in a biocontainment setting was used.42 In brief, the EPG equipment including the head stage amplifier and host probe were housed inside a gasketed acrylic box surrounded by a Faraday cage inside a biosafety level 2 laboratory in addition approved as an arthropod containment level 2 facility. All recordings were made with an applied substrate voltage of 100 mV AC and an input resistance of 107 Ω. A 1.3 megapixel USB microscope with a long working distance (model AF4515ZTL; Dino-Lite, Torrence, CA) was positioned within the glovebox to allow for video capture of experiments.
Experimental outline.
Experiments were designed to detect changes in A. aegypti probing and feeding behaviors associated with DENV-2 infection in the mosquito by comparing 3 experimental groups: group 1 (control): uninfected A. aegypti biting uninfected mice; group 2: uninfected A. aegypti biting DENV-infected mice; and group 3: DENV-infected A. aegypti biting uninfected mice (Figure 2). Each group contained 9 mice (5 female, 4 males). The sample size was determined in consultation with the IACUC based on previous arbovirus studies with blood-feeding arthropods12,33,40 and considering the committee’s mandate to minimize the use of animals. The experimental unit for this study is based on the number of analyzable waveforms produced by each mosquito-mouse pairing. Female mosquitoes for all experiments were selected randomly from a colony-rearing cage. Each experimental condition drew mosquitoes from a minimum of 3 separate generations of the mosquito colony reared under standardized protocols. Recordings were consistently carried out at the same time of day to reduce any potential temporal bias in feeding behaviors. Treatment allocation could not be concealed from technical staff.
Citation: Comparative Medicine 74, 4; 10.30802/AALAS-CM-24-030
Group 1 (Control).
Mice were anesthetized intraperitoneally with a cocktail of ketamine (90 mg/kg), xylazine (10 mg/kg), and acepromazine (2 mg/kg) based on standard formularies,15,28 recommendations with the IACUC, and previous experience with rodent sedation requirements for EPG recordings.42 Mice were shaved on the ventrum and dorsum, and eye lubricant was applied before being placed in sternal recumbency on the host probe of the EPG system. The mouse was provided heat during recordings by way of a 75-mL sodium acetate heating pad beneath the host probe. A small section of bubble wrap was used to prevent direct contact between the heating pad and the host probe. Electrode gel (Spectra 360 Electrode Gel) was applied to the ventrum of the mouse at the site of host probe contact to ensure a satisfactory electrical connection.
EPG recordings were conducted following previously published methods.42 Briefly, wired mosquitoes were attached to the head stage amplifier of the prepared EPG system and allowed 10 min to attempt probing on the mouse host. Once a mosquito was fed to repletion and withdrew its mouthparts, the mosquito was returned to the small holding box in the glove box. For the control group only, if the mouse was still on a surgical plane of anesthesia, multiple mosquitoes were allowed to attempt probing. Mice were then injected intraperitoneally with a 1 mg/kg atipamezole diluted in 0.9% saline as an anesthetic reversal and allowed to recover. Terminal bleeds were collected at 72 h post-EPG recording.
Group 2.
All experiments were performed as in group 1 with the following exceptions: mice were inoculated subcutaneously in a rear footpad with 20 µL of media containing 2 × 105 PFUe of DENV-2 48 h before EPG recording. DENV infection was confirmed by RT-qPCR of 72 h post-EPG terminal blood collections by RT-qPCR with a cutoff Ct of 38. Only one A. aegypti was allowed to probe and feed on each DENV-infected mouse.
Group 3.
All experiments were performed as in group 1 with the following exceptions: Mosquitoes (2 d posteclosion) were inoculated with DENV-2 by intrathoracic injection using a Nanoject II system (Drummond Scientific, Broomall, PA) and standard methods.11 Briefly, capillary tubes were heated, pulled into points, and beveled to create fine needles. Using a positive displacement action, the Nanoject II system allowed for the injection of 69 nL of concentrated DENV-2 (1.0 × 108 PFUe/mL) into the thorax of cold anesthetized mosquitoes. Injected mosquitoes were allowed to recover for 7 d. A single hind leg was removed from each injected mosquito and tested by RT-qPCR for DENV-2 dissemination. Mosquitoes with confirmed viral dissemination were used for EPG recordings on days 9 to 10 postinjection. Only one DENV-infected A. aegypti was allowed to probe and feed on an uninfected mouse.
Statistical analysis.
Recorded waveforms were manually annotated following published conventions for A. aegypti.42,43 The unreliable visibility of waveform J and infrequent and treatment group-biased occurrences of waveform N resulted in these 2 waveform families being omitted from the analysis. Data files were preprocessed and transition probabilities were calculated with the FileManipW2 082117 and Error Checker 100716 SAS programs. The data were analyzed with the Ebert Generic14 and Backus 2.03 SAS programs for sequential and nonsequential variables, respectively (all programs are available from https://crec.ifas.ufl.edu/extension/epg/).
For sequential variables, a dummy variable was used immediately preceding waveform K to allow for analysis of the major waveform families: K, L, and M. The default Gaussian distribution was used for all variables except for NumPrbsAftrFrstXY, which used the γ distribution. As the Ebert Generic program is designed to work with both long- and short-duration feeding insects, only a subset of the variables calculated by the program provide meaningful results in quickly feeding insects like mosquitoes. For this reason, 7 of the 14 variables were excluded a priori, as described previously.42 Nonsequential variables were calculated with the Backus 2.0 SAS program (Table 1). Both programs calculate differences between groups by least squares means and use the GLIMMIX function for analysis of variance to test the overall significance of experimental treatment as a fixed effect. For both programs, the appropriateness of the underlying distribution was assessed by residual plots and χ2 goodness-of-fit statistics calculated within each program. All analysis steps were performed using SAS software, version 9.4 of the SAS System for Windows.
| Program | Response variable name | Description |
|---|---|---|
| Ebert Generic | TtlDurXY | The total duration of XY per insect |
| MeanXY | The mean duration of XY per insect | |
| MeanNXYPrb | The mean number of XY events per probe per insect | |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first XY | |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first XY event to the first XY event | |
| NumPrbsAftrFrstXY | The number of probes after the first XY event | |
| MaxXY | The longest duration of an XY event per insect | |
| Backus 2.0 | WDE | Waveform duration per event |
| PDE | Probing duration per event | |
| NWEI | Number of waveform events per insect | |
| WDI | Waveform duration per insect | |
| WDEI | Waveform duration per event (per insect) | |
| WDP | Waveform duration per probe | |
| NWEP | Number of waveform events per probe | |
| NPI | Number of probes per insect | |
| PDI | Probing duration per insect |
Results
Following DENV confirmation in all groups and assessing the quality of EPG recordings, group 2 (4 males, 3 females) and group 3 (3 males, 4 females) each contained 7 mosquito-mouse pairs with recordings suitable for analysis. Each of these groups had 2 mosquito-mouse pairs with unreadable recordings due to mosquito wiring complications or connectivity issues with the EPG. Group 1 controls (2 males, 5 females) had many waveforms available for analysis as a result of allowing multiple mosquito bites per mouse. To reduce bias in the analysis, the waveform from the first mosquito bite on each mouse in the control group was used. Of these, 7 of the 9 mice had analyzable waveforms for the first mosquito to probe and feed. Thus, the number of mice with interpretable waveforms for initial mosquito bites in each group was n = 7. The summary of calculated sequential and nonsequential response variables across waveforms and groups is presented in Table 2. To further examine the relationships among the treatment groups within significant response variables, pairwise differences in means were used.
| Programmatic variable name | Descriptive variable name | n | Group 1 | Group 2 | Group 3 | Num df | Den df | Pr > F | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Est. mean | SE | Est. mean | SE | Est. mean | SE | ||||||
| Waveform K | |||||||||||
| TtlDurXY | The total duration of K per insect | 39 | 4.90 | 2.12 | 4.90 | 1.80 | 9.55 | 1.34 | 2 | 36 | 0.0654 |
| MeanXY | The mean duration of K per insect | 39 | 4.45 | 0.50 | 3.70 | 0.42 | 4.28 | 0.31 | 2 | 36 | 0.4395 |
| MeanNXYPrb | The mean number of K events per probe per insect | 39 | 1.00 | 0.08 | 1.00 | 0.07 | 1.10 | 0.05 | 2 | 36 | 0.3872 |
| NumPrbsAftrFrstXY | The number of probes after the first K event | 25 | −3.59 × 10−14a | 0.3024 | 0.36a | 0.16 | 1.29b | 0.17 | 2 | 22 | <0.0001 |
| MaxXY | The longest duration of an K event per insect | 39 | 4.58 | 0.87 | 5.27 | 0.74 | 6.78 | 0.55 | 2 | 36 | 0.0755 |
| NWEI | Number of waveform events per insect | 21 | 1.14 | 0.47 | 1.57 | 0.47 | 2.56 | 0.47 | 2 | 18 | 0.0465 |
| Waveform L1 | |||||||||||
| TtlDurXY | The total duration of L1 per insect | 42 | 22.91 | 10.55 | 15.74 | 8.42 | 39.08 | 5.70 | 2 | 39 | 0.0653 |
| MeanXY | The mean duration of L1 per insect | 42 | 17.01a | 2.56 | 6.42b | 2.04 | 11.37 | 1.39 | 2 | 39 | 0.0091 |
| MeanNXYPrb | The mean number of L1 events per probe per insect | 42 | 1.79 | 0.40 | 1.55 | 0.32 | 1.97 | 0.22 | 2 | 39 | 0.5456 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first L1 | 42 | 17.54a | 35.06 | 265.00b | 27.97 | 70.59a | 18.94 | 2 | 39 | <0.0001 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first L1 event to the first L1 event | 42 | 17.54a | 33.49 | 189.51b | 26.72 | 44.07a | 18.09 | 2 | 39 | <0.0001 |
| NumPrbsAftrFrstXY | The number of probes after the first L1 event | 27 | −3.54 × 10−14 | 0.47 | 0.69 | 0.19 | 0.90 | 0.11 | 2 | 24 | 0.1602 |
| MaxXY | The longest duration of an L1 event per insect | 42 | 20.34 | 3.82 | 10.88 | 3.05 | 19.30 | 2.07 | 2 | 39 | 0.0617 |
| WDE | Waveform duration per event | 42 | 17.01 | 3.20 | 6.42 | 2.55 | 11.37 | 1.73 | 2 | 39 | 0.0432 |
| Waveform L2 | |||||||||||
| TtlDurXY | The total duration of L2 per insect | 60 | 46.39a | 28.66 | 236.03b | 23.05 | 91.05a | 16.80 | 2 | 57 | <0.0001 |
| MeanXY | The mean duration of L2 per insect | 60 | 30.82a | 6.40 | 60.04b | 5.15 | 22.40a | 3.75 | 2 | 57 | <0.001 |
| MeanNXYPrb | The mean number of L2 events per probe per insect | 60 | 1.68 | 0.36 | 1.75 | 0.29 | 2.36 | 0.21 | 2 | 57 | 0.1261 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first L2 | 60 | 6.11a | 0.63 | 3.96b | 0.51 | 3.60b | 0.37 | 2 | 57 | 0.0041 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first L2 event to the first L2 event | 60 | 6.11a | 0.63 | 3.96b | 0.51 | 3.60b | 0.37 | 2 | 57 | 0.0041 |
| NumPrbsAftrFrstXY | The number of probes after the first L2 event | 41 | −1.12 × 10−9 | 0.31 | 0.46 | 0.15 | 1.10 | 0.11 | 2 | 38 | 0.0004 |
| MaxXY | The longest duration of an L2 event per insect | 60 | 45.54a | 24.73 | 223.06b | 19.89 | 57.73a | 14.50 | 2 | 57 | <0.0001 |
| Waveform M1 | |||||||||||
| TtlDurXY | The total duration of M1 per insect | 37 | 40.56 | 26.66 | 72.75 | 27.75 | 43.47 | 27.75 | 2 | 34 | 0.6610 |
| MeanXY | The mean duration of M1 per insect | 37 | 27.69 | 18.99 | 64.59 | 19.77 | 35.17 | 19.77 | 2 | 34 | 0.3799 |
| MeanNXYPrb | The mean number of M1 events per probe per insect | 37 | 2.50a | 0.27 | 1.63b | 0.29 | 1.01b | 0.29 | 2 | 34 | 0.0026 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M1 | 37 | 65.19a | 45.80 | 250.36b | 47.67 | 205.65b | 47.67 | 2 | 34 | 0.0213 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first M1 event to the first M1 event | 37 | 55.23 | 14.24 | 84.85 | 14.82 | 81.99 | 14.82 | 2 | 34 | 0.2911 |
| NumPrbsAftrFrstXY | The number of probes after the first M1 event | ID | ID | ID | ID | ID | ID | ID | ID | ID | ID |
| MaxXY | The longest duration of an M1 event per insect | 37 | 50.24 | 30.88 | 111.33 | 32.14 | 40.43 | 32.14 | 2 | 34 | 0.2510 |
| Waveform M2 | |||||||||||
| TtlDurXY | The total duration of M2 per insect | 41 | 85.65 | 33.14 | 173.98 | 32.01 | 92.02 | 35.79 | 2 | 38 | 0.1177 |
| MeanXY | The mean duration of M2 per insect | 41 | 44.66a | 21.72 | 118.79b | 20.99 | 39.00a | 23.46 | 2 | 38 | 0.0215 |
| MeanNXYPrb | The mean number of M2 events per probe per insect | 41 | 2.54a | 0.23 | 1.69b | 0.22 | 1.00c | 0.25 | 2 | 38 | 0.0003 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M2 | 41 | 77.76a | 36.94 | 224.66b | 35.68 | 213.52b | 39.90 | 2 | 38 | 0.0129 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first M2 event to the first M2 event | 41 | 68.52 | 18.27 | 54.71 | 17.65 | 75.62 | 19.73 | 2 | 38 | 0.7196 |
| NumPrbsAftrFrstXY | The number of probes after the first M2 event | ID | ID | ID | ID | ID | ID | ID | ID | ID | ID |
| MaxXY | The longest duration of an M2 event per insect | 41 | 60.80a | 30.81 | 190.23b | 29.77 | 79.38a | 33.28 | 2 | 38 | 0.0093 |
| Composite parameters | |||||||||||
| PDI | Probing duration per insect* | 38 | 188.31 | 75.47 | 336.35 | 64.36 | 105.02 | 48.97 | 2 | 35 | 0.0253 |
Estimated means (Est. means) and SEs are least squares means calculated by the GLIMMIX procedure in Ebert Generic. a,b,cStatistically relevant groupings by line. Different letters denote a statistical difference between groups of P ≤ 0.05 where the overall type III tests of fixed effects (Pr > F) are also significant at P ≤ 0.05 with the denoted numerator and denominator df (Num df and Den df, respectively). n, number of observations used for the calculation; ID, indeterminant: insufficient number of occurrences were available to calculate these response variables.
Group 2: Uninfected A. aegypti biting DENV-infected mouse.
Sequential variables.
The difference in means between controls and group 2 was calculated for significant sequential variables. These are listed in Table 3 with each response variable’s descriptive name and programmatic variable name. This group with DENV-infected mice/uninfected mosquitoes had waveforms that were statistically different from controls in 4 of the 5 waveforms analyzed. The only waveform for which this group did not differ significantly from controls was during the K waveform, which is generated during initial penetration of the skin. Group 2 demonstrated longer durations from overall initial skin penetration to reach the L1 waveform (average = 247.45 s longer, TmFrmFrstPrbFrstXY, P < 0.0001) and longer durations to reach L1 during any probing event that contained L1 (average = 171.97 s longer, TmBegPrbFrstXY, P = 0.0003). This group also spent less than half as long in L1 as controls, on average (MeanXY, P = 0.0025). Conversely, for L2 this group entered L2 more quickly (average = 2.16 s faster, TmFrmFrstPrbFrstXY, P = 0.0098) and had longer duration L2 waveforms than controls by several metrics (average = 189.64 s longer, TtlDurXY, P < 0.0001; MeanXY, P = 0.0008, MaxXY, P < 0.0001). With respect to ingestion waveforms, this group averaged fewer M1 waveforms per probe (average = 0.88 fewer, MeanNXYPrb, P = 0.0339) and M1 was reached much later than controls (average = 185.17 s later, TmFrmFrstPrbFrstXY, P = 0.0083). Similarly, group 2 averaged fewer M2 waveforms per probe (average = 0.85 fewer, MeanNXYPrb, P = 0.0125) and M2 was reached much later than controls (average = 146.90 s later, TmFrmFrstPrbFrstXY, P = 0.0068). In addition, this group displayed much longer M2 waveforms than controls by 2 metrics (mean duration averaged 74.13 s longer, MeanXY, P = 0.0188; MaxXY, P = 0.0045).
| Difference by least squares means: Group 1 compared with group 2: Significant sequential variables | ||||||
|---|---|---|---|---|---|---|
| Programmatic variable name | Difference in means estimate | SE | df | t value | Pr > |t| | |
| Waveform L1 | ||||||
| MeanXY | The mean duration of L1 per insect | 10.5942 | 3.2737 | 39 | 3.24 | 0.0025 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first L1 | −247.45 | 44.8510 | 39 | −5.52 | <0.0001 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first L1 event to the first L1 event | −171.97 | 42.8454 | 39 | −4.01 | 0.0003 |
| Waveform L2 | ||||||
| TtlDurXY | The total duration of L2 per insect | −189.64 | 36.7785 | 57 | −5.16 | <0.0001 |
| MeanXY | The mean duration of L2 per insect | −27.216 | 8.2191 | 57 | −3.55 | 0.0008 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first L2 | 2.1574 | 0.8071 | 57 | 2.67 | 0.0098 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first L2 event to the first L2 event | 2.1574 | 0.8071 | 57 | 2.67 | 0.0098 |
| MaxXY | The longest duration of an L2 event for each insect | −177.51 | 61.7363 | 57 | −5.59 | <0.0001 |
| Waveform M1 | ||||||
| MeanNXYPrb | The mean number of M1 events per probe per insect | 0.8750 | 0.3959 | 34 | 2.21 | 0.0339 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M1 | −185.17 | 66.1086 | 34 | −2.80 | 0.0083 |
| Waveform M2 | ||||||
| MeanXY | The mean duration of M2 per insect | −74.132 | 30.2028 | 38 | −2.45 | 0.0188 |
| MeanNXYPrb | The mean number of M2 events per probe per insect | 0.8468 | 0.3228 | 38 | 2.62 | 0.0125 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M2 | −146.90 | 51.3588 | 38 | −2.86 | 0.0068 |
| MaxXY | The longest duration of an M2 event per insect | −129.43 | 42.8447 | 38 | −3.02 | 0.0045 |
| Difference in least squares means: Group 1 compared with group 2: Significant nonsequential variables | ||||||
| Waveform L1 | ||||||
| WDE | Waveform duration per event | 10.5942 | 4.0880 | 39 | 2.59 | 0.0134 |
The estimated difference in means and SE were calculated using a t test with the denoted df to examine pairwise relationships between these groups for the significant response variables. The variable ‘time from beginning of first probe to the first [waveform]’ differs from ‘time from the beginning of the probe with the first [waveform] event to the first [waveform] event’ in that not all probes have all waveforms. The former measures from the very beginning of the first probe while the latter measures only from the first probe where the waveform of interest actually occurred.
Nonsequential variables.
The Backus 2.0 program determined that there was one nonsequential variable that was statistically different between the DENV-infected mouse/uninfected mosquito group and the controls. This was the response variable waveform duration per event for waveform L1 (Table 3). In this program, the term ‘waveform event’ is used to refer to a single, uninterrupted occurrence of a waveform during a single probing attempt. The mean duration of L1 was found to be significantly shorter in duration in group 2 compared with controls (10.59 s shorter on average, P = 0.0134).
Group 3: DENV-infected A. aegypti biting uninfected mouse.
Sequential variables.
The difference in means between controls and group 3 was calculated for significant sequential response variables. These are summarized in Table 4 with each response variable’s descriptive name and programmatic variable name. The DENV-infected mosquito/uninfected mouse group had waveforms that differed significantly from controls in 4 of the 5 waveforms analyzed. This group was not statistically different from controls with respect to L1 waveforms in the variables analyzed. The DENV-infected A. aegypti group demonstrated an average of 1.29 probes (initiated by Waveform K) after the first probe, which is significantly more than controls (NumPrbsAftrFrstXY, P = 0.0006). In addition, group 3 was found to have longer maximum waveform K’s durations than controls (average = 2.20 s longer, MaxXY, P = 0.0399). Like group 2, group 3 reached the L2 phase of the pathway more quickly than controls (average = 2.52 s faster, TmFrmFrstPrbFrstXY, P = 0.0011). DENV-infected A. aegypti averaged fewer ingestion waveforms, M1 and M2, per probe than controls (MeanNXYPrb, M1: average = 1.49 s fewer, P = 0.0007; M2: 1.54 s fewer, P < 0.0001), and they had an extended time from the beginning of the first probe to the first respective ingestion waveform event (TmFrmFrstPrbFrstXY, M1: average = 140 s longer, P = 0.0410; M2: average = 135.75 s longer, P = 0.0170).
| Difference by least squares means: Group 3 compared with group 1: Significant sequential variables | ||||||
|---|---|---|---|---|---|---|
| Programmatic variable name | Difference in means estimate | SE | df | t value | Pr > |t| | |
| Waveform K | ||||||
| NumPrbsAftrFrstXY | The number of probes after the first K event | 1.2879 | 0.3207 | 22 | 4.02 | 0.0006 |
| MaxXY | The longest duration of an K event per insect | 2.2013 | 1.0327 | 36 | 2.13 | 0.0399 |
| Waveform L2 | ||||||
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first L2 | −2.5168 | 0.7290 | 57 | −3.45 | 0.0011 |
| TmBegPrbFrstXY | The time from the beginning of the probe with the first L2 event to the first L2 event | −2.5168 | 0.7290 | 57 | −3.45 | 0.0011 |
| Waveform M1 | ||||||
| MeanNXYPrb | The mean number of M1 events per probe per insect | −1.4861 | 0.3959 | 34 | −3.75 | 0.0007 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M1 | 140.46 | 66.1086 | 34 | 2.12 | 0.0410 |
| Waveform M2 | ||||||
| MeanNXYPrb | The mean number of M2 events per probe per insect | −1.5357 | 0.3417 | 38 | −4.49 | <0.0001 |
| TmFrmFrstPrbFrstXY | The time from the beginning of the first probe to the first M2 | 135.75 | 54.3698 | 38 | 2.50 | 0.0170 |
| Difference in least squares means: Group 1 compared with group 3: Significant nonsequential variables | ||||||
| Composite parameter | ||||||
| PDI | Probing duration per insecta | −0.8984 | 0.3483 | 35 | −2.58 | 0.0143 |
| Waveform K | ||||||
| NWEI | Number of waveform events per insect | 1.7143 | 0.6598 | 18 | 2.60 | 0.0182 |
The estimated difference in means and SE were calculated using a t test with the denoted DF to examine pairwise relationships between these groups for the significant response variables.
Nonsequential variables.
The Backus 2.0 program determined that there were 2 nonsequential variables that were statistically different between group 3 and group 1 controls (Table 4). The composite variable probing duration per insect was found to be significantly lower in group 3 than in controls (P = 0.0143). In addition, the number of waveform events per insect was found to be significant for waveform K with the number of waveform events per insect being higher in group 3 than in controls (average = 1.71 more events, P = 0.0182).
Transition probabilities.
Transition probabilities calculated by the Error Checker program were used to manually construct kinetograms in Microsoft Word (Figure 3). These plots outline the percent probability of transition from one step of the probing process to another. In uninfected A. aegypti (group 1), this process typically follows a relatively consistent pattern: skin penetration (K), to pathway (L), to ingestion (M), to withdrawal (W) with some percentage moving into a postingestion waveform N before W. Only a small percentage of mosquitoes in the pathway abort feeding and withdraw. When A. aegypti were allowed to feed on DENV-infected mice (group 2), the feeding kinetogram noticeably shifts to include a higher percentage of mosquitoes aborting feeding during pathway (36%), and no observed instances of waveform N are observed. The kinetogram of group 3 is noticeably more convoluted than that of either comparator group. These DENV-infected A. aegypti vacillated between steps within the feeding process and had the highest percentage of transitions to W from L of any experimental group (58%).
Citation: Comparative Medicine 74, 4; 10.30802/AALAS-CM-24-030
Discussion
Previous work with the use of EPG technology with mosquitoes on a mouse host42 provided critical information needed to perform similar studies with risk group 2 pathogens; however, the current study required further refinement of the previous methods especially with respect to anesthesia. EPG is very sensitive to movement and vibrations, including movements from the mouse host. To minimize these movements, we refined our earlier methods42 and developed an anesthetic protocol that provides a deep and long-duration anesthetic plane. The deep anesthesia reduces the risk of the mouse moving while recording is in progress, and the long activity allows for multiple mosquito recordings and ample time to safely negotiate biocontainment barriers. The addition of acepromazine and a lower dose of xylazine in the anesthetic cocktail provided a stable and consistent plane of anesthesia needed for this study.
Behavioral changes in probing observed in DENV-infected A. aegypti were in agreement with previously published studies that found infected A. aegypti had longer feeding periods34 and probed more often to reach engorgement.44 The significantly lower number of ingestion waveforms per probe and longer duration to arrive at the first ingestion waveform as detailed in Results are a result of repeated short probes that abort before ingestion. This behavior is also reflected in the kinetograms as a 58% transition from L to W. The increased probing measured is of full withdrawal of the mouthparts and the beginning of a new probe. Partial withdrawal and reinsertion, visualized as an L to K path on the kinetograms occurred in only group 3 and only 4% of the time. This repeated puncturing of the skin is thought to be paired with saliva deposition that recruits proinflammatory cell populations21 and increases the exposure of tissue-dwelling dendritic cells, which serve as a primary infection target for DENV.37 It should be noted, that although M1 and M2 are associated with blood ingestion based on correlation experiments in A. aegypti and another mosquito species13,43 and a temporal association between these waveform types and visualizing blood through the pleural membrane of the abdomen was present in the current work, rigorous correlation experiments to prove the assumed ingestion association of the M-waveform family were not completed here.
Interestingly, DENV-infected A. aegypti entered the L2 waveform of the preingestion pathway phase more than 2.5 s faster than controls, on average. Because the number of L1 and L2 events were not statistically different between these groups, this may suggest that the poorly defined behavioral processes resulting in the irregular L2 waveform are associated with DENV transmission. In particular, there may be an association between the preference for L2 and the documented change in expectorated saliva in DENV-infected mosquitoes.9 Future histologic validation of stylet placement and saliva deposition in L1 and L2 will be needed to clarify this association.
The use of EPG with applied current as a tool to understand bite site interactions between animals and mosquitoes may be beneficial in the development of novel mosquitocidal vaccines6,17 and drugs.16,24,38 The rationale for the utility of EPG in this area is a direct translation from agricultural disciplines where EPG has been used for several decades. Major applications in this field of research include determining the effectiveness of systemic insecticides8,20,31 and assessing immune or other factors that impact the feeding process and alter pathogen transmission.4,18 In much the same way, EPG may play an important role in the understanding the mechanism of and screening systemic drugs that negatively impact mosquito feeding like ivermectin.16 With increased development in this area, EPG will be able to provide highly precise measurements of changes in mosquito feeding behaviors and thereby meaningfully move this field forward.
Strong mosquito behavioral changes were also associated with DENV presence in the mouse host. DENV-2 has been reported to increase the attractiveness of infected mouse hosts by altering the skin microbiota.45 While skin flora was not tested in the present study, marked differences in feeding behaviors were observed in uninfected A. aegypti feeding on DENV-infected mice. Most notable among this group is the extended periods spent in the L-waveform family and the simplified path to completing a feeding event once it has begun, as observable in the kinetograms. The increased time spent in L may have implications for pathogen acquisition. It is assumed that during this phase the stylets are migrating through and saliva is being deposited within tissues while searching for a capillary. Increased dwelling time in this phase may allow time for recruitment and ingestion of infected dendritic cells or similarly infected cell populations to the bite site. The biologic reasons for why the kinetograms appear to be so different among the treatment groups remain unknown. New statistical tools are in development to allow for rigorous analysis of these data. This will allow for the comparison of kinetograms between biologic replicates as well as between independent research groups and allow for more narrow research questions to be asked about specific transition probabilities that are consistently shifted in association with DENV.
Increased time spent in an early waveform by the DENV-infected mouse group influences the statistics of later waveforms as sequential statistics are calculated using the time from the beginning of the file. This is a known limitation of sequential statistics3 and should be interpreted with caution. For this reason, we also performed an analysis of nonsequential response variables.
The present study has some limitations that should be noted. As a preliminary study to validate the utility of EPG to discern behavioral changes in arboviral infections, the sample size was kept small and the age of mice was not controlled for. While the assumptions for the statistical methods were met and the methods are appropriate, a study seeking to dissect behaviors associated with arboviral infections would benefit from an increased sample size and an age-matched randomization strategy. The use of the Rockefeller strain of A. aegypti is routine for such studies given its high degree of lab adaptation and low genetic diversity. However, this limits the generalizability of results to wild-type A. aegypti, which may be better represented by a recently field-derived colony. In addition, this study did not include sham injections of saline or an alternate pathogen. To this end, it is not possible to distinguish between behavioral changes that are the result of DENV exposure and those that are purely related to immune activation or the infection procedure itself. While this limits the interpretation and generalizability of the results, it does not impact the study’s overall hypothesis. That is, EPG with alternating current is capable of detecting mosquito behavioral changes associated with DENV infection. This is a critical distinction as it was previously unclear whether this system, which records the sum of the electrical resistance across the entire mouse-mosquito system, would be able to discern mosquito behavioral changes given the larger size and endogenous electrical complexities of the mouse host.
Finally, it is not known if the removal of a hind leg to test for virus dissemination alters the feeding behaviors of A. aegypti. This study did not remove hind legs from all tested mosquitoes as an internal control. Notably, mosquitoes routinely lose legs during manipulation and the tracking and reporting of missing legs during virus transmission studies is not a standard practice. In addition, A. aegypti does not stand on its hind legs during feeding or at rest, and thus, it is unlikely to have caused behavioral changes based on balance or surefootedness. However, because a single hind leg was not removed from all mosquitoes and routine leg loss was not monitored, this remains a possible confounding variable. Future studies should include an increased sample size, sham microinjections, and experimental controls for leg removal procedures.
A) Representative waveforms from a mosquito probe on a mouse host with the waveform families K, L, M, and W denoted; 2.2 s/div, 8x WinDaq gain; box (a) denotes a selected region of M family waves that are highlighted panel B; (B) A zoomed in sample of ingestion waveforms demonstrating the transition from the regular M1 waveform to the irregular M2 waveform; 0.2 s/div, 64x WinDaq gain; (C) A still image of a wired Ae. aegypti feeding on an IFN-αβR- mouse during an EPG recording; (D) The glovebox containing the head stage amplifier (center, ring stand mounted) and host probe (center, large exposed copper wire) used for the EPG recordings in this study.
Experimental design. Group 1: Uninfected mosquitoes paired with uninfected mice. Group 2: Dengue virus (DENV)-infected mice paired with uninfected mosquitoes. Group 3: Uninfected mice paired with DENV-infected mosquitoes. Mosquitoes in all groups were 10 to 12 d old at the time of EPG recording. Created with Biorender.com.
Kinetograms of transition probabilities between different waveform families. NP, nonprobing; J, preprobing; K, insertion; L, preingestion/pathway; M, ingestion; N, postingestion/undefined; W, withdrawal of mouthparts.
Contributor Notes