Device design and fabrication
Vectorchips with integrated PDMS elastomer membranes were fabricated using a combination of laser patterning and soft lithography. We selected PDMS as the material of choice due to its low cost, biocompatibility, optical transparency, and chemical stability. PDMS-based devices can be fabricated using reproducible and scalable fabrication techniques26. Such devices have routinely been used for molecular analysis of small sample volumes using microfluidics27, and are compatible with molecular assays and cell cultures21,22. We used laser patterning to create open-ended arrayed compartments (3−4 mm deep) to capture saliva droplets from mosquito bites in PDMS micro-wells (Fig. 1a–d). In order to obtain ultra-thin PDMS films that the mosquito mouth parts can pierce, we utilized spin-coating of uncured PDMS over a flat silicon surface. We obtained membrane-integrated devices after plasma-induced PDMS-PDMS bonding of the laser-patterned array with the PDMS film (Fig. 1e). Detailed steps describing chip fabrication are included in the methods section below. The versatile fabrication process can provide user-defined variation in the size and density of the individual compartments. We were able to fabricate chips with hole diameters ranging from 150 μm to 1 cm (Supplementary Fig. 1) showing the scales of operation for the laser-ablation process while maintaining membrane stability. Additionally, the membrane can be easily integrated with fluidic networks for direct interfacing with mosquito bites, enabling assays involving on-chip fluid exchange (Fig. 1f). The microfluidic compartments on the chips can hold feeding media such as blood or sugar water (Fig. 1g), collect saliva during biting events, and act as isolated reaction chambers for molecular assays.
a Schematic showing the fabrication process of Vectorchip. Laser patterning was used to obtain through-holes in blocks of PDMS. A thin sacrificial layer of photoresist (Microposit™ S1813 G2, DOW Inc., USA) was spread onto a silicon wafer followed by spin-coating a thin film of PDMS on the wafer surface. The laser-patterned body of the chip was then bonded to the thin film of PDMS using plasma-activated PDMS-PDMS bonding. Membrane-bonded chips were obtained by placing the wafer in an acetone bath, where the sacrificial layer of resist was dissolved. b A PDMS block with laser-generated through-holes (diameter 250 μm). c A PDMS chip bonded to a 1.6 μm thin PDMS membrane on a 4-inch silcon wafer. d A Vectorchip with around 3000 wells (dia: 250 μm) is shown. e Chip with well diameter 1.75 mm and a visible PDMS membrane (thickness 1.6 μm). f Microfluidic channel-integrated chips with a PDMS membrane on top. g These wells can be loaded with feeding solution, reaction reagents, and can store mosquito saliva after biting assays. Wells loaded with feeding media (10% sucrose laced with blue food color) shown in the image. Scale bars represent 1 mm in (b), 5 mm in (e), and 2 mm in (f, g).
Mosquito-chip interactions
Next, we examined the ability of mosquitoes to pierce through the PDMS membranes. Mosquitoes were attracted to the chips using heat as a guiding cue although several other baiting methods could be used. We placed a camera above the chip and observed stylet insertion through 1.6 μm thick PDMS membranes (Fig. 2a, Supplementary Movies 1 and 2). Abdominal engorgement in mosquitoes was observed after biting, indicating that mosquitoes can successfully feed through these membranes. Four species of mosquitoes were tested (Aedes aegypti, Aedes albopictus, Culex tarsalis, Culex quinquefasciatus) and all demonstrated successful probing and feeding through 1.6 μm thick PDMS membranes.
a Stylet entry through a PDMS membrane for an Aedes aegypti female. b Mosquito feeding success as a function of membrane thickness for two mosquito species—Aedes aegypti and Culex tarsalis. We demonstrate that by changing the membrane thickness we can turn biting on or off. Biting off is defined as when no mosquito in the cage was able to feed from the chip in a duration of 30 min. The feeding assays were repeated at least three times (n = 3) at every given membrane thickness to confirm the ability of mosquitoes to feed through the membrane. c A Culex tarsalis female mosquito biting through a 20 μm thick PDMS membrane. d Bending of the proboscis observed for an Aedes aegypti female, indicating failure in biting through the 20 μm membrane. e Fluorescent salivary droplets expectorated by an Aedes aegypti mosquito while probing. These mosquitoes were fed on rhodamine-laced sugar water resulting in fluorescent saliva. Three salivary droplets (1, 2, and 3) are encircled. f Timelapse images show a magnified view of fluorescent salivary droplet deposition (droplets numbered 1, 2, and 3 as shown in panel (e) over a period of 4 s). Scale bars represent 500 μm in (a) and 2 mm in (c, d).
We also examined whether the membrane thicknesses can be engineered to selectively prevent the biting of some species while letting other species feed through the barrier. We loaded the laser-patterned microfluidic compartments with blood and warmed the chips to attract mosquitoes (Supplementary Fig. 2 and Supplementary Movie 3). We considered biting to be off when none of the mosquitoes in the cage demonstrated abdominal engorgement within 30 min since the start of the experiment. Tests were performed with Aedes aegypti (~45 mosquitoes per cage) or Culex tarsalis (~25 mosquitoes per cage) while varying the thickness of the PDMS membrane from 900 nm to 100 μm. We observed significant differences in the biting ability of the two species, where complete inhibition of feeding by Aedes aegypti was observed with membrane thickness at or greater than 15 μm (Supplementary Movies 4 and 5). In contrast, complete inhibition of feeding by Culex tarsalis was only observed for membranes that were 100 μm thick (Fig. 2b). This remarkable difference in the biting capacity of these two species has not been reported earlier. While some studies have focused on understanding the biting mechanics of a single mosquito species (Aedes aegypti or Aedes albopictus)28,29, we are not aware of any study which quantifies the differences in biting strength between different species and the evolutionary or biomechanical differences involved in this variation. Evolutionary pressure on mosquitoes biting different hosts could generate great variability of biting strength in mosquito species, which may be needed to probe different skin thickness. This intriguing biomechanical phenomenon can be advantageous for deploying devices where membranes of desired thickness can act as species-selective filters, providing a means to better track highly relevant species and pathogens in the field.
In order to extract molecular information from mosquito bites, we exploited the release of saliva during probing and biting events. The volume of saliva expectorated by mosquitoes in oil has been reported previously to be around 0.03 nL/min for Culex pipiens30, and 6.8 nL in approximately an hour for Aedes aegypti31 using a forced salivation method. Aedes aegypti have been estimated to salivate around 4.7 nL during blood feeding events32. We demonstrate that feeding rhodamine-labeled sugar water to mosquitoes renders their saliva fluorescent such that the release of saliva droplets can be directly visualized using fluorescence imaging (Fig. 2e and Supplementary Movie 6). Time sequences (Fig. 2f) show magnified images of three locations on a membrane surface where short probing events result in the deposition of fluorescent salivary droplets. We estimate that the volume of salivary droplets expectorated during this process is approximately 0.66 nL (Supplementary Fig. 3). This calculated mean volume is considerably lower compared to the volume associated with blood-feeding events reported earlier32; however, this discrepancy likely reflects differences between probing events such as observed in Fig. 2e, f and full feeding events as reported by Devine et al.32. The deposited salivary droplets harbor several biomarkers in the form of mosquito salivary proteins33, active pathogens, mosquito cells, and/or nucleic acids that can be used for their identification14,20,34.
Distribution of bites on chip
An important consideration for molecular surveillance of mosquitoes is the granularity at which the population is sampled. Methods can range from pooling strategies with low coverage where samples are pooled together at the population scale, to the individual collection of saliva from mosquitoes while recording species identity. Vectorchip potentially provides an adjustable method between these two strategies by increasing or decreasing the density of bite wells presented to a population of mosquitoes.
We first analyzed the distribution of bites on Vectorchip by tracking mosquito activity during biting assays. In order to perform sucrose biting assays for molecular diagnostics, we filled the wells with 10% sucrose, and covered the open side of the chip with a glass slide. A resistor that acts as a heat source was placed atop the glass slide and the chip-resistor assembly was placed on top of the mesh ceiling of a mosquito cage. A part of the chip (50 percent of the total chip area) was protected by paper tape such that mosquitoes could not bite into that section of the device. The area of the chip protected by tape served as negative control for the downstream molecular assays. A camera (Raspberry Pi 3 camera module V2.1) was placed on the bottom of the cage to track and record individual mosquito activity on chip (Supplementary Fig. 4). Image recognition and mosquito tracking algorithms have been described by us previously25, where we used computer vision to analyze image sequences and robustly track the presence, movement, and biting behavior of individual mosquitoes at high resolution. In this study, we have used an imaging setup where the camera is attached to the bottom of the cage (Supplementary Fig. 4a). The obtained image datasets reveal mosquito trajectories on the chip as well as areas with prolonged interactions (Fig. 3a–d, Supplementary Fig. 4, and Supplementary Movie 7). Unique trajectories are identified as movements performed by an individual mosquito while in the field of view (e.g., landing, exploration, biting, and take-off by a single mosquito would constitute a trajectory). The movement of mosquitoes on a chip is shown in Fig. 3e–g and Supplementary Fig. 4. Depending on the duration of mosquito-chip interactions, multiple trajectories do not overlap (Fig. 3e), partially overlap (Fig. 3f), or completely overlap (Fig. 3g). Trajectory plots can indicate the likelihood that a set of wells has been probed by more than one mosquito. These results indicate that by modulating the time of interaction between the mosquito population and the chip, the user can control the degree of probing experienced by the chip. As an example, a chip with no overlap between trajectories should exclusively contain wells probed by single mosquitoes. Tracking mosquito movement on chip was used to locate trajectories and thus regions with unique mosquito presence.
a Image recognition software was used to identify and track mosquitoes on the chip. Wells are highlighted in yellow. b Trajectories of mosquito movement can be plotted and indicate the overlap between different movement tracks. c Location plot of mosquitoes on the chip are indicated by red circles. Regions with overlapping red circles are darker in color. d Trajectory and location plot can be used to identify regions with individual mosquito locomotion activity as compared to regions with more crowding. e−g Varying experimental parameters (symbols indicate the duration of experiments and number of mosquitoes) can be used to obtain either individual mosquito plots or population data. Three cases are discussed with e no overlap between multiple probing trajectories, f partial overlap, and g complete overlap. h Representative image for a chip where the presence of fluorescent droplets on wells has been indicated by a red dot. i Zoomed-in fluorescent image of the same chip showing wells with single, or multiple droplets. j Number of bites per well were calculated by counting fluorescent droplets deposited on the wells after biting events. k We defined the number of mosquitoes = Nm, the number of wells = Nw, probing frequency = η, and time = t. The probability of obtaining ‘k’ bites on a well for the parameters (Nm, Nw, η, t) utilized in (h−j) has been shown. Scale bars represent 5 mm in (a−c, e−g), 2 mm in (d), and 1 mm in (i).
While tracking provided macro-scale information about mosquito position and activity, we also utilized the release of fluorescent salivary droplets deposited by biting mosquitoes to resolve mosquito probing events at higher resolution. We quantified the number of fluorescent spots seen in every well of the PDMS chip by using fluorescent scans of the chip before and after biting (Fig. 3h). No spots were observed in the negative control region where access was blocked to mosquitoes, while several fluorescent spots observed on the open areas of the chip indicate the wells which have been probed. The images show the presence of wells with multiple bite marks, as well as single bite marks.
We further examined the design parameters used to build Vectorchip that can dictate the resolution of this sampling strategy. The primary determinants of how often a single mosquito will bite a specific well are dependent on multiple factors, including the number of mosquitoes in the cage, the time mosquitoes have access to the chip and biting behavior (e.g., frequency), and the size and number of wells on the chip. Other additional factors like the feeding media may affect the number of biting events performed by a female mosquito. In this study, we focused on sucrose as the feeding media since it is a diet source that does not inhibit polymerase chain reaction (PCR) assays. We considered a simple mathematical model to arrive at a first approximation for the distribution of bites on a chip. In this model, we made a simple assumption that mosquitoes probe the chip at a fixed frequency (η). We also assumed that mosquito bites are random and independent events, since no spatial gradients have been created on the chip. Detailed analysis of this model can be found in Supplementary Section 1.
We defined the number of mosquitoes = Nm, the number of wells = Nw, and probing frequency = η, and estimated the probability (P) that a well receives k bites in given time t as (Supplementary Note 1):
$$P[k]=({N}_{m}eta t)!/(k!times ({N}_{m}eta t-k)!)times 1/{N}_{w}^{k}times {(1-1/{N}_{w})}^{{N}_{m}eta t-k}$$
(1)
This was approximated by the form (Supplementary Note 1):
$$P[k]approx {{{{{{mathrm{e}}}}}}}^{-{N}_{m}eta t/{N}_{w}}times {({N}_{m}eta t/{N}_{w})}^{k}/k!$$
(2)
and represents a Poisson distribution of the form,
$$P[k]={{{{{{mathrm{e}}}}}}}^{-lambda }times {(lambda )}^{k}/k!$$
(3)
with the expected rate of events (i.e., bites per well), λ = Nmηt/Nw. Probability distributions for the expected number of bites on a well were plotted while varying the number of mosquitoes, wells, and probing frequencies (Supplementary Fig. 5). The probing frequency of Aedes aegypti mosquitoes on a parafilm membrane has been estimated previously to be in the range of 0–1 bites per min35. Supplementary Fig. 5 shows a diverse phase space where a single bite per well is the most dominant outcome, i.e., λ was less than or equal to 1, when up to 50 mosquitoes interact with a 150-well Vectorchip for 15 min—in line with our experiments. In the context of a larger number of mosquitoes interacting with the chip, increasing the number of wells can ensure that the majority of wells will not get bitten more than once. We quantified bites on wells as indicated from fluorescent droplets (Supplementary Table 1), and compared it to the analytical model (Fig. 3j, k). The fitting parameter was η = 0.1 bites per minute, which is close to the range of probing frequencies reported recently35. With water as the feeding media and parafilm as the probing membrane, Jove et al. report that the majority of tested mosquitoes showed a probing frequency below 0.5 bites per minute35. While we observe and model slightly lower probing frequencies, this difference is possible due to the change in the membrane material from parafilm to PDMS, and other differences in the setup. The methods discussed above provide multiple ways to understand mosquito−chip interactions as well as help us define parameters to obtain data at single-bite resolution.
PCR diagnostics on-chip
Having established that mosquitoes can bite through the membrane and deposit salivary droplets in microwells while biting on a Vectorchip, we analyzed these bites for their DNA and RNA content. Running PCR-based nucleic acid amplification directly on individual wells establishes a low-cost assay to detect bite content in a high-throughput manner. We utilized devices with a PDMS membrane thickness of 1.6 μm and well diameter of 1.75 mm to perform molecular analysis of salivary samples; providing 145 independent reactions per chip (within a chip area of 5 cm × 2.5 cm).
Detection of mosquito mitochondrial DNA (mtDNA) and RNA from viruses in saliva was performed using on-chip PCR with end-point fluorescence readout. A detailed protocol for performing PCR in Vectorchip is provided in the Methods section and Supporting Information file (Supplementary Fig. 6). We tested the efficiency of PCR amplification in Vectorchip by manually loading a known concentration of DNA and RNA into the reaction wells (Fig. 4a–c). We performed the PCR for 42 cycles and detect DNA amplification from approximately 5 DNA copies in 4 μL of reaction mix (~1 copy/μL) (Fig. 4b). Simultaneously detection of Zika virus RNA using reverse transcription-PCR (RT-PCR), was successfully demonstrated from as low as 15 RNA copies in 4 μL of reaction mix (Fig. 4c). PCR reactions were also performed in 96-well plates to test if amplification accuracy and sensitivity were similar as compared to Vectorchips. Both qPCR amplification curves and end point fluorescence information was obtained from the well plates. The 96-well plate reactions were performed in DI water and also in the presence of dried sucrose (10%) to ensure that the amplification response provided a good comparison to the Vectorchip-based reactions using manually spiked concentrations, and to test if the presence of sucrose would inhibit the reactions (see Supplementary Fig. 7a, b).
The symbols indicate manually spiked assays, uninfected mosquito assays, and Zika infected mosquito assays from top to bottom. a–c A spiked assay with successful amplification of Aedes aegypti DNA and Zika RNA on chip. a Indicates wells filled with PCR mix, with ROX dye providing the background fluorescence. Fluorescent wells indicate amplification of b mosquito DNA, and c Zika RNA on chip. Assays were performed with uninfected mosquitoes where d tracking patterns were collected and demonstrate extensive translocation activity on the chip. e PCR shows detection of mosquito DNA on the chip directly from biting. f Percentage of wells available for biting on Vectorchip that display a positive PCR outcome. The PCR assay was repeated 6 times on different chips (n = 6). The false positive rate (amplification in wells that were covered by tape) was close to zero (0.23%). Assays performed with Zika infected mosquitoes indicate the presence of both g Aedes aegypti DNA and h Zika virus RNA after bites on chip. i Percentage of wells available for biting that show a positive signature for Zika RNA, mosquito DNA or both. The PCR assay was repeated 3 times on different chips (n = 3). Scale bars represent 5 mm in (a−e) and 3 mm in (g, h). Source data are provided as a Source Data file.
Vectorchip provides an opportunity for tracking both pathogens in salivary bites, as well as identifying host species by detecting DNA from the host biting the well. Mosquito DNA has previously been reported in salivary droplets deposited by feeding Culex mosquitoes20. Utilizing uninfected Aedes aegypti mosquitoes, we performed PCR assays on-chip to detect the presence of mosquito DNA in deposited saliva droplets. Figure 4d, e shows the tracking data and PCR detection of mtDNA for a chip, which was placed on a cage with 75 uninfected Aedes aegypti females for 45 min. Areas highlighted in green were protected by paper tape such that mosquitoes could not probe them and serve as negative controls. The tracking data obtained from the chip indicate significant activity of the mosquitoes on-chip as represented by several overlapping trajectories. PCR data confirms that mosquito DNA indeed can be detected in probed wells. Multiple experiments indicate that a subset (3−12%, n = 6) of the wells where mosquitoes are active test positive for the presence of mosquito DNA (Fig. 4f). The rate of false positives obtained from wells not accessible to mosquitoes was very close to zero (0.23%, n = 6). In order to further validate the rate of detection of mosquito DNA on Vectorchips, we performed a biting assay using a 96 well plate loaded with agarose and covered with a parafilm membrane. Agarose gel was used as feeding media during this test, to minimize the risk of cross-contamination between wells during the peeling off of the parafilm membrane (details in “Methods” section). We recorded the tracking and DNA amplification response from uninfected mosquito bites on the well plate (Supplementary Fig. 7c, d). Approximately 15% of the wells where mosquitoes were active tested positive for mosquito DNA, which is close to the fractions observed on Vectorchips. Since mosquito DNA can only be detected in a subset of wells showing mosquito activity, this suggests that the presence of detectable levels of mosquito DNA in saliva is a noisy physiological process, likely dependent on multiple variables such as duration of feeding and time of last bite. It is furthermore important to note that our tracking algorithm only detects the presence of mosquitoes, but does not provide information regarding if a well was bitten or not.
Next, we utilized the device to perform assays on infected mosquitoes, for the detection of both mosquito species and deposited pathogens. Figure 4h shows results obtained when Aedes aegypti infected with Zika virus (ZIKV)36 interacted with the chip for 20 min. We subsequently performed RT-PCR on the chip and observed that Aedes aegypti mtDNA and ZIKV RNA could be detected in 24/118 and 37/118 wells, respectively. We summarize the detection rates of mosquito DNA and ZIKV RNA for three assays performed with ZIKV-infected mosquitoes in Fig. 4i. Interestingly, not all wells positive for ZIKV RNA showed positive for mosquito DNA and vice versa, highlighting the stochastic genetic composition of bite-derived nanoliter saliva droplets. A higher fraction of wells (mean ~ 22%, n = 3) showed positive for ZIKV RNA as compared to mosquito mtDNA (mean ~ 17%, n = 3), indicating a higher prevalence of viral genomic material in mosquito saliva as compared to mosquito mtDNA. Detection of mosquito DNA and virus RNA in these assays demonstrates the capacity of this tool to identify mosquito and pathogen species directly from mosquito bites on-chip.
Detection of infectious viral particles directly from bites
While PCR-based end-point assays are an important strategy for multiplexed detection of vectors and pathogens in various settings, they cannot determine the presence or concentration of infectious virus particles in bites. Detection of infectious pathogen load in mosquito bites is important to understand the vector competence (including potential and efficiency of pathogen transmission through bites) of mosquito-pathogen systems and its dependency on factors such as local climate37, mosquito species38, physiology39, and presence of endosymbionts (e.g., Wolbachia)40. Typical methods to measure vector competence rely on manual “forced salivation” of individual mosquitoes12, where the viral loads may differ from biting events and mosquitoes are sacrificed preventing time-course measurements. Sugar feeding on filter papers has also been utilized recently, and provides promising results towards longitudinal sampling of viral transmission by individual mosquitoes20. As compared to filter paper-based approaches, the Vectorchip uses an integrated PDMS membrane which can enable purely bite-based collection and avoid possible contamination from e.g., excretion events. Furthermore, our method improves the throughput of the assay as mosquitoes do not need to be individually housed and sampled. Finally, filter paper assays are reliant on sampling nucleic acids, and cannot perform live cell culture assays to detect infectious pathogens.
We tested the potential of Vectorchip to perform focus forming assays (FFA) and detect infectious viral particles directly as a result of mosquito bites on chip (Fig. 5). Vectorchip wells were filled with cell culture media (Dulbecco’s modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS). The efficiency of FFA on-chip was tested using manual spiking of viral particles into Vectorchip (Supplementary Fig. 8). In order to perform biting assays, Aedes aegypti mosquitoes were infected with Mayaro virus, an emerging zoonotic pathogen that causes a dengue-like disease41. The mosquitoes were able to bite through the membrane into the wells and transfer live Mayaro viral particles directly into the cell culture medium (Fig. 5a). These viral particles then infect a monolayer of vero cells cultured on the PDMS membrane. These infections can be identified via fluorescent antibodies specific to viral envelope proteins (Fig. 5c). Each fluorescent patch (focus) is attributed to an infection resulting from a single viral particle. Foci counted on two sample chips can be seen in Fig. 5d. No foci were visible from negative control samples. The distribution of viral foci on the chips indicates the heterogeneity of infectious particle dose in mosquito bites. Figure 5 shows that the Vectorchip supports the growth of a cell monolayer on the chip and can enable the direct collection of mosquito saliva through bites in cell culture media. Antibody assays can be used to directly quantify infectious viral particles in the wells. These results demonstrate the suitability of using the Vectorchip to probe the transmission of infectious viral particles.
Focus forming assays on Vectorchip can directly quantify the number of active viral particles in mosquito bites, without the need for isolation of individual mosquitoes and manual salivary extraction. a Image shows Aedes aegypti mosquitoes infected with Mayaro virus biting on Vectorchips filled with DMEM cell culture media. b A monolayer of vero cells was cultured on Vectorchip membrane. The formation of vero cell monolayers on membranes was verified on 5 independent chips (n = 5). c FFA performed on a Vectorchip. Symbols indicate wells in the top row which were bitten by mosquitoes with no infection (negative control), and rows of wells bitten by mosquitoes infected with Mayaro virus. The green channel shows fluorescence from an antibody against a viral envelope protein. Every green island represents infected cells likely to have resulted from an active viral particle. No active viral particles were seen in the control wells. FFA on Vectorchips was verified on 3 chips (n = 3) for both uninfected and Mayaro infected mosquitoes. d FFA formed on two chips and viral foci were counted using the antibody fluorescence. Scale bars represent 2 mm in (a), 30 μm in (b), and 100 μm in (c).
Finally, we examined if the feeding media provided in Vectorchip can enable blood meal-mimic biting behavior through the addition of phagostimulants such as ATP, while maintaining accurate molecular analysis (Supplementary Note 2). We observed that the addition of 100 μM ATP, (which is a phagostimulant and used by mosquito sensory neurons to identify blood35), to the feeding media did not impact nucleic acid amplification reactions (Supplementary Fig. 9a) (n = 3). Interestingly, mosquitoes feeding on DMEM supplemented with 10% FBS and ATP (1 mM) led to a similar or higher number of abdominal engorgements in 45 min as compared to blood meals (n = 3) (Supplementary Fig. 9b). This is reasonable as DMEM (supplemented with FBS and ATP) contains a variety of the components used by mosquito sensory neurons to identify blood at relevant concentrations (Supplementary Note 2)35). In summary, our data indicate that Vectorchip can provide physiologically relevant mosquito feeding behavior, enabling accurate assessment of vector and pathogen identity through nucleic acid amplification reactions for field surveillance, as well as on-chip assessment of transmission of infectious viral particles in bites through focus forming assays.
Source: Ecology - nature.com
