Hemlock woolly adelgid (HWA), Adelges tsugae (Annaand), is an invasive insect that has devastated millions of eastern hemlocks, Tsuga canadensis (L.) Carrière, in eastern North America since the 1950s. In Michigan, HWA was first detected in 2015 and has spread to several counties in west Michigan. Conservation practitioners are managing HWA through a statewide, landscape-level pesticide management plan. The traditional method of using randomized branch sampling to estimate infestation levels is time-consuming but allows managers to monitor how HWA responds to treatments. Our main objective was to determine if molecular techniques could replace traditional methods for estimating infestation levels. We selected six treated sites and three untreated sites in west Michigan. During the fall and winter of 2022, we estimated the densities of the sistens generation using a randomized branch sampling method, and during the summer of 2023, we deployed five traps at each site to capture adelgid crawlers and performed qPCR analysis to quantify HWA DNA. To determine if qPCR is an appropriate method to assess HWA infestation level, we used linear regressions to determine whether relationships existed between (1) crawler counts and qPCR values, (2) sistens counts and qPCR values, and (3) sistens counts and crawler counts. We found that qPCR analyses and adelgid crawler counts have comparable, strong correlations with traditional estimation methods, particularly during the peak crawler period. Managers can use either qPCR or crawler counts from the peak crawler period to understand how HWA infestations respond to treatment or to prioritize treatment in high-risk areas.

Site selection

We selected six treated sites with varying levels of HWA infestation and three infested, untreated sites. The infested, untreated sites are Meinert County Park (MNPK), Duck Creek Natural Area (FRPK), and Pioneer County Park (PIPK). Our treated sites included several private properties: BRHA, SPLT, HMDV, and BBAP. Two of the treated sites are public parcels owned by Golden Township (GDNR) and the Muskegon Conservation District (MKGR) (Table 1; Figure 1). Only abbreviations for private properties are given for anonymity. At the treated sites, crews from local organizations treated all hemlocks with imidacloprid products with a trunk injection or basal bark application, depending on the tree size and proximity to water.

Sistens Generation Counts

We used a modified version of the RBS methodology outlined by Evans and Gregoire (2007) to quantify the HWA infestation levels. During fall and winter of 2022 and into early 2023, we randomly selected 10 trees at each of the nine sites to sample, for a total of 90 hemlocks sampled. Because these sites were assessed for HWA by local and state groups, all hemlocks within stands were previously tagged with unique identifying numbers. At each tree, we randomly selected nine branches maximum within reaching distance. We collected new growth twigs from each of the selected branches and stored the twigs at 4℃ in Ziploc bags. We used a Nikon SMZ645 dissecting microscope to count the individuals from the sistens generation present near the base of 100 new growth needles on the front and back of each twig. Counting individuals from the sistens generation is an appropriate measure for HWA because it is this generation that settles on new growth, which allows us to compare infestation levels between sites and years (Evans and Gregoire 2007). To obtain the average number of sistentes for each site, we first averaged the sistens counts from all sampled branches per tree, and then we averaged the number of sistentes counted across the 10 trees per site.

During the fall and winter of 2023, we again sampled the same trees across six of the original sites (GDNR, BBAP, FRPK, PIPK, MKGR, and MNPK) and used the protocol described above to collect the average number of sistentes for each site. However, the number of total trees sampled in 2023 was lower than in 2022 because some trees no longer produced new growth due to a decline in hemlock health. In total, 50 trees were assessed across the six sites in 2023, ranging from 6-10 trees per site.  

Trap Deployment

Our lab developed 3-D printed traps that each hold four microscope slides (Figure 2) for HWA monitoring. The traps were initially designed based on a fungal spore trap developed by Quesada et al. (2018), and we modified them to increase the capture probability of HWA crawlers and associated HWA material (ovisacs) in infested areas (Sanders et al. 2023) while limiting the by-catch of larger arthropods. We dipped one side of each slide in sterilized petroleum jelly and placed the slides in the traps with the petroleum jelly face-up. We mounted the traps on a 1.5 m wooden pole and attached a tag with trap identification and project information.

For each of the nine sites, we used ArcGIS to map 1-acre grids on the properties and randomly selected a point in five different grids to deploy traps within each site (ESRI 2023). The traps were placed within the hemlock stand closest to the randomly selected point. This resulted in a total of 45 traps across the nine sites (five traps per site). On April 24th, 2023, we deployed the traps and exchanged slides every two weeks until August 1st, 2023, totaling seven collection periods. This allows us to sample across most of the adelgid crawler period. For each two-week collection period, we collected one field blank per site to ensure we did not contaminate the samples during sample collection. We stored samples at 4℃ in 50 mL tubes until slide processing.

In spring 2024, we deployed the traps in six of the nine original sites (GDNR, BBAP, FRPK, PIPK, MKGR, and MNPK). We deployed five traps within each site, and each trap was placed in the same location as those deployed in 2023. Traps were deployed on April 23, 2024, and slides were exchanged every two weeks until July 3, 2024, totaling five collection periods. These traps were only assessed for crawler counts and not for qPCR.

Crawler Counts

We inspected each slide under a Nikon SMZ645 dissecting microscope to count the number of HWA crawlers on the slides and noted any evidence of other adelgid material (i.e., ovisac material). Since each trap had four slides, we summed these counts to give the total number of crawlers per trap. To limit DNA contamination, we counted the crawlers in a separate lab than the DNA extraction lab. The microscope and lab bench were cleaned between each trap with either 10% bleach or DNA Away.

DNA Extraction

Immediately after counting the crawlers from the 2023 slides, we removed the petroleum jelly from each of the four slides on a trap using a stainless-steel micro spatula and placed it in a single Zymo bead bashing tube containing 0.1 mm and 2.0 mm beads (Zymo Research, CA, USA), resulting in 315 samples, which represented one tube per trap per collection period. In between processing traps, we cleaned the micro spatula using flame sterilization and DNA Away.  For the remainder of the benchwork, we moved to a different lab to reduce the risk of DNA contamination.

We used Quick-DNA Tissue/Insect Microprep Kits (Zymo Research, CA, USA) and modified the manufacturer’s protocols to separate the DNA from the petroleum jelly. We placed the tubes in a heat block at 65℃ for 15 minutes to liquefy the petroleum jelly. Then, we homogenized the samples in a Qiagen TissueLyser LT (Hilden, Germany) at 50 rps for 2 minutes. We repeated the heating/bashing cycle until the tubes were homogenized for at least 10 minutes. After the final heat cycle, we separated the petroleum jelly from the homogenized samples by centrifuging the tubes at 11,000 rpm for 1 minute. This allowed for any aqueous material to collect at the bottom of the tube, and the petroleum jelly to collect at the top of the tube. We used a sterilized toothpick to poke a hole through the upper layer of the petroleum jelly and pipetted out the remaining aqueous solution into a Zymo-Spin III F filter column inside a collection tube. We followed the manufacturer’s protocols for the remainder of the procedure, except that we eluted the DNA in a final volume of 20 mL. For each set of samples, we processed a petroleum jelly blank and an extraction blank using the same method to ensure the petroleum jelly and reagents were not contaminated. We stored the samples at -20℃ until qPCR analysis.

qPCR Protocol

We created a seven-point standard curve using a gBlock standard for HWA designed by Kirtane et al. (2022) (Integrated DNA Technologies, IA, USA). The standard curve consisted of a 1:10 serial dilution with concentrations ranging from 2x10⁶ gene copies (gc)/µL – 2.0 gc/µL.

We ran the standard curve, all samples, and no template controls in triplicate. The qPCR reactions consisted of 1X TaqMan Environmental Master Mix, 0.6 µM final concentration of the forward and reverse primer, 0.3 µM final concentration for the probe (Fam labeled with a Zen/Iowa Black^TM FQ double quencher; Integrated DNA Technologies, IA, USA), 2.0 mg/mL final concentration of bovine serum albumin (BSA), and 2.0 µL of DNA for a total volume of 20 µL. Kirtane et al. (2022) developed the primers and probes, which target a region of the cytochrome oxidase 1 (COX1) gene. These were previously tested against multiple adelgid species and determined to be HWA-specific (Sanders 2021, Kirtane et al. 2022). We ran the samples on a Step One Plus Real-Time PCR System (Applied Biosystems, MA, USA). The thermal cycle profile included a pre-holding stage at 60ºC for 30 seconds, followed by a holding stage of 50ºC for 2 minutes and then 95ºC for 10 minutes. The cycling stage consisted of 95ºC for 15 seconds, followed by 60ºC for 1 minute for 40 cycles. All standard curves needed an efficiency between 90% – 110% and an R² > 0.98 to pass quality control and be included in the analyses, which are the standard thresholds used in other studies (Rogers-Broadway and Karteris 2015). We processed 315 samples (45 traps across 7 sampling periods) but removed four samples before analysis due to technical errors during laboratory processing.

LOD and LOQ Determination for qPCR

We compiled data from 35 of our HWA qPCR standard curves to determine the LOD and LOQ of our HWA assay. In addition, we included several samples with HWA concentrations of 1.0 gc/µL, 5.0 gc/µL, 10.0 gc/µL, and 15.0 gc/µL to help determine our LOD. We defined the LOD as the lowest number of gc/reaction that produced a positive reaction in at least 95% of the samples (Klymus et al. 2019). We defined the LOQ as the lowest gc/reaction that could be quantified with a coefficient of variation (CV) of less than 35%. We estimated the LOD and LOQ in R (version 4.4.0; R Core Team 2024) using code developed by Merkes et al. (2019, Klymus et al. 2019). To quantify gc/reaction, we obtained the number of gc/µL produced from the standard curve and multiplied this by the 2.0 µL of DNA that we added to the reaction. For example, our lowest standard curve template was 2.0 gc/µL, so the number of gc/reaction would be 4.0 gc.

Relationship between 2023 Adelgid Crawler Counts and qPCR Values Across Sites

To identify whether there was a correlation between the number of adelgid crawlers captured on the trap slides and the qPCR values associated with those traps, we first ran a linear regression between the number of adelgid crawlers captured per trap and the qPCR value of that trap for each of the seven collection periods across all nine infested sites. We then ran individual linear regressions for each site to examine how this relationship varied across sites. Before the regression analysis, all data were transformed using log(value + 1) due to the large variance between individual counts and gc/reaction values across samples. We performed all analyses using R (version 4.4.0; R Core Team 2024).

Relationship between 2022 Sistens Generation Counts and Trap Values During the 2023 Crawler Peak

Our initial goal was to evaluate the general relationship between sistens generation counts and the trap data. To do this, we focused on our 2022 sistens counts to see how they compared to our 2023 trap data. We used the average sistens counts for each site and only used the average adelgid crawler counts and the average qPCR values per site generated during the peak crawler period. This is the period when we would expect to see the strongest relationship between HWA qPCR values and sistens generation counts since this is when the majority of the adelgid crawlers (i.e., the generation produced by the sistens generation) are mobile and will be moving to new locations to settle. The peak crawler period for 2023 was the June 7^th, 2023 collection period, which was the deployment period from May 24^th, 2023 – June 7^th, 2023.

We analyzed the relationship between average sistens generation counts, average adelgid crawler counts from the traps, and their corresponding qPCR values using a correlation analysis. Firstly, we performed a Spearman correlation analysis between average sistens generation counts and the average adelgid crawler counts per site. Secondly, we performed a Pearson correlation analysis between average sistens generation counts and the average qPCR values per site. Before the correlation analysis, some data were transformed using log(value + 1) due to the large variance between individual counts and gc/reaction values across samples and assessed for normality. We performed all analyses using R (version 4.4.0; R Core Team 2024).

Assessment of 2024 Peak Crawler Data to Predict 2023 Sistens Generation Counts

Because crawler counts had the strongest correlation to sistens generation counts (see results below), we evaluated whether we could use average crawler counts to predict average sistens counts within a site. To do this, we assessed how well data collected the following year (2023 sistens counts and 2024 crawler counts) fit within a generalized linear model developed from the 2022 sistens and 2023 crawler count data. Average crawler counts during the peak crawler period in 2024 were assessed as described above, except the peak crawler period was earlier and ranged from May 8, 2024 – May 29, 2024, due to a warmer winter. Before our analysis, we removed data from one of the traps (FRPK Trap 1). Crawler data from this trap appeared to be influenced by one highly infested tree, with crawler counts averaging over 23,000 crawlers for the peak crawler period. This was significantly higher than the other traps in the site and it was considered an outlier.

We used a Spearman correlation to assess the relationship between the 2023 sistens generation counts and 2024 crawler counts to compare to the previous year’s data. We then ran a generalized linear model (glm) based on the previous year’s data and used the glm.predict function to see how well the 2024 crawler counts predicted the 2023 sistens generation counts. We performed all analyses using R (version 4.4.0; R Core Team 2024).