Longitudinal Measurements of Tarnished Plant Bug (Hemiptera: Miridae) Susceptibility to Insecticides in Arkansas, Louisiana, and Mississippi: Associations with Insecticide Use and Insect Control Recommendations

Concentration-response assays were conducted from 2008 through 2015 to measure the susceptibility of field populations of Lygus lineolaris (Palisot de Beauvois) from the Delta regions of Arkansas, Louisiana, and Mississippi to acephate, imidacloprid, thiamethoxam, permethrin, and sulfoxaflor. A total of 229 field populations were examined for susceptibility to acephate, 145 for susceptibility to imidacloprid, and 208 for susceptibility to thiamethoxam. Permethrin assays were conducted in 2014 and 2015 to measure levels of pyrethroid resistance in 44 field populations, and sulfoxaflor assays were conducted against 24 field populations in 2015. Resistance to acephate and permethrin is as high or higher than that previously reported, although some populations, especially those exposed to permethrin, appear to be susceptible. Variable assay responses were measured for populations exposed to imidacloprid and thiamethoxam. Average response metrics suggest that populations are generally susceptible to the neonicotinoids, but a few populations from cotton fields experiencing control problems exhibited elevated LC50s. Efforts to associate variability in LC50s with recorded use of insecticides and estimated cotton insect losses and control costs suggest that intensive use of insecticides over several decades may have elevated general detoxifying enzymes in L. lineolaris and some field populations may be exhibiting resistance to multiple classes of insecticide. These results suggest that efforts should be made to manage these pests more efficiently with a reduced use of insecticides and alternative controls.


Introduction
Increased use of insecticide sprays for the targeted control of tarnished plant bug (Lygus lineolaris (Palisot de Beauvois)) in the Midsouth has been widely discussed and highlighted by numerous authors over the last two decades [1,2]. The current number of sprays made for tarnished plant bug is increasing and is somewhat reminiscent of scheduled "calendar day" approaches to pest management from the 1950s and 1960s that led to adverse effects of over-reliance on chemical control, resulting environmental problems, outbreaks of secondary pests, and a pesticide treadmill that evolved around The methods for the bioassays conducted from 2008 through 2013 were those previously described by Snodgrass et al. [21] for imidacloprid and thiamethoxam assays, and Snodgrass et al. [21] for acephate and permethrin assays. The collection sites ( Figure 1) included much of the Delta regions of Arkansas, Louisiana, and Mississippi. Tarnished plant bugs were captured via sweep net primarily from weedy hosts in ditches and field borders as described by Snodgrass et al. [21]. As with previous studies, no attempts were made to associate assay response to host plants due to many collections being made from areas with more than one host present, but the geographic location of the collection site was noted and subsequently studied with specific locations on digital maps. Not all sites had specific geographic coordinates associated with them, and these locations were excluded from some analyses. Depending upon the number of individuals collected at a sample site, assays may have been conducted with multiple insecticides. All of the assays were conducted with adults held under laboratory conditions for 24 h prior to testing; these individuals were held in cardboard (0.95 L) cartons and fed cut-green beans (Phaseolus vulgaris L.) surface sterilized by washing in a 3% sodium hypochlorite solution as described by Snodgrass [16].

Glass Vial/Contact Bioassays
Glass vial assays with acephate and permethrin were the same as those described by Snodgrass et al. [21] using the procedures developed by Snodgrass [16]. Technical grade insecticide was purchased from Chem Service (West Chester, PA, USA), stored in a freezer at −20 °C, dissolved in pesticide-grade acetone (Fisher, Fair Lawn, NJ, USA) at known concentrations, and pipetted in 0.5

Glass Vial/Contact Bioassays
Glass vial assays with acephate and permethrin were the same as those described by Snodgrass et al. [21] using the procedures developed by Snodgrass [16]. Technical grade insecticide was purchased from Chem Service (West Chester, PA, USA), stored in a freezer at −20 • C, dissolved in pesticide-grade acetone (Fisher, Fair Lawn, NJ, USA) at known concentrations, and pipetted in 0.5 mL aliquots into 20 mL scintillation vials. Treated vials were rolled on a hotdog roller (Star MFG, Smithville, TN, USA) without heat under a fume-hood until the residues were dry. A 4 cm piece of green bean was added to the vial as a food source before two adult insects were exposed to the insecticide in each vial, with multiple vials making up a replicate. Each population was exposed to five to six concentrations of insecticides, and control vials were treated with acetone alone. Mortality was measured after 24 h of exposure using the criteria outlined in Snodgrass et al. [21]. Assays conducted from 2008 through 2013 were replicated three times, with ten bugs per concentration and replicate. During this period of time, researchers would sometimes return to a sample site to collect additional bugs within 24 h if needed to complete replicates and minimize variability within the collected population. Mortality from bioassays was corrected for control effects using Abbot's formula [32]. Data were analyzed by probit analysis (PROC PROBIT, SAS 9.4, SAS Institute Inc., Cary, NC, USA, 2013) to establish mortality regression lines, and comparisons were made to those obtained for a control colony from Crossett, AR, considered to be susceptible due to its distal location from insecticide treatments on cotton [20,21,31]. Insects from this location were collected yearly to add to the colony, and reared on broccoli. This colony has been the standard benchmark of susceptibility used by the Snodgrass laboratory for the past two decades.

Floral Foam/Feeding Bioassays
Oral bioassays with imidacloprid and thiamethoxam were developed following observations from Teague and Tugwell [33] that plant bugs feed on fluids in floral foam [20]. The assays conducted between 2008 and 2013 were conducted as described by Snodgrass et al. [20]. Briefly, 12 mm diameter pieces of floral foam were cut from large blocks of floral foam and placed in 20 mL glass scintillation vials. Test concentrations of imidacloprid (0.0, 1.0, 2.5, 5.0, 7.5, and 10 µg/vial) and thiamethoxam (0.0, 0.1, 0.5, 1.0, 2.5, and 5.0 µg/vial) [20] were pipetted onto the floral foam in a 10% honey solution in 0.5 mL aliquots. Observations of mortality for imidacloprid were made at 72 h post-exposure to the treated floral foam, which was based on preliminary observations that indicated low mortality at earlier observation times [20]. The mortality of individuals exposed to thiamethoxam was measured after 24 h of exposure. All of the other insect collection and assay procedures were the same as those previously outlined, including data analyses and replication.

2014 and 2015 Bioassays
The methodology used in 2014 and 2015 differed slightly, but overall, the procedures were similar to those used by Snodgrass from 2008 through 2013. The differences in the methodology used are noted below. Glass vial assays were essentially the same as those used previously, but the numbers of individuals assayed from each collection were more variable, with a greater emphasis on increasing the number of collections and individual assays. When sufficient numbers of tarnished plant bugs were obtained from a collection, bioassays with six test concentrations were repeated three to four times per field-collected population as described by Snodgrass et al. [21]. The test concentrations used for imidacloprid, thiamethoxam, and sulfoxaflor were 0, 0.15, 0.5, 1.5, 5, and 15 µg/vial, while assays with acephate and permethrin utilized test concentrations of 0, 1, 3, 10, 30, and 100 µg/vial. When the numbers of individuals from a collection were limited, assays were conducted with a minimum of 60 individuals (using six concentrations and 10 bugs per concentration). There were no attempts to return to a collection site in 2014 or 2015 to supplement collection numbers. Data analyses were exactly the same as those previously described, and the resulting data summaries were based only on statistically significant regression models (p < 0.05 for Chi-Square Tests of slope and p > 0.05 for Chi-Square Goodness of Fit). Individual regressions that were not significant with the above criteria were eliminated from both the summary of the Snodgrass laboratory assays for 2008 through 2013 and more recent assays in 2014 and 2015.
The floral foam assays with imidacloprid and thiamethoxam in 2014 and 2015 were exactly the same as those described for the Snodgrass laboratory during 2008 through 2013, except that the observations of mortality for imidacloprid were made at 24 h instead of 72 h based on a series of laboratory assays. Collections made from the Crossett, AR, location were also maintained as a "susceptible" control, collected, and assayed in order to calculate resistance ratios (RR 50 ) as a comparison with other locations.
Additionally, in 2014 and 2015, a laboratory colony was added to the routine assays as an experimental control. The laboratory-reared insects were from a colony established in 1998 and kept at the USDA ARS' Southern Insect Management Research Unit in Stoneville, MS. The colony is reared following details outlined in Portilla et al. [34], to provide large numbers of known-age insects. The insects are reared under controlled conditions in environmental chambers (constant 27 • C and a photoperiod of 16:8 (L:D) h). Assays with the USDA colony were conducted exactly as those with the field colonies, with insects of a uniform age (1-day-old adults) that had been fed on a meridic diet. Groups of mixed-age adults were not available from the USDA colony due to rearing procedures.
Assays with sulfoxaflor have not been previously published, so exploratory assays were conducted using both glass vial and floral foam assays with the USDA laboratory colony. The resulting LC 50 s (95% confidence limits) for the vial assays (9.420 (3.525-27.184) µg/vial) and the floral foam assays (53.236 (12.147-130.639) µg/vial) suggested higher contact activity using the glass vial bioassay. The glass vial procedures as described above were used to test field populations for response to sulfoxaflor in 2015. As with all of the other 2014 and 2015 assays, mortality was measured after 24 h of exposure to the insecticides.

Data Analyses
Once response regressions were completed and the entire assay data set was assembled and available for collective study, we examined the temporal (year) and spatial (state) patterns of response to each insecticide using standard least square ANOVA models in JMP, Version 11.1. The spatial patterns (latitude and longitude) of insect response for each field collection to each insecticide (LC 50 s) were examined by linear models using JMP, Version 11.1. To explore the possible linkages across the landscape for tarnished plant bug susceptibility (average LC 50 s), insecticide use measured as kg of insecticide per hectare of harvested cropland (Figure 2), and annual estimates of cotton insect control costs and losses [35], were averaged across years for each of the three states and studied by Pearson's pairwise correlation using the Multivariate Procedure in JMP (Version 11.1). Additional exploratory research was conducted by examining possible relationships to estimated insecticide use by collecting county-level information from the USGS National Water Quality Assessment Program's Pesticide National Synthesis Project [36][37][38] Linear relationships between observed LC 50 s and county-level estimates of kg of insecticide used per hectare of harvested cropland were developed and studied for each of the insecticide classes and the total set of LC 50 estimates for each insecticide. The insecticide use groups were organophosphates, pyrethroids, neonicotinoids, and sulfoxaflor insecticides. General changes in tarnished plant bug susceptibility over the last two decades were also examined via pairwise correlation analyses relative to recommended control practices by the Mississippi Cooperative Extension Service [39][40][41][42] and annual insect loss estimates for Arkansas, Louisiana, and Mississippi from 2008 through 2015 [35].

Results
Benchmark comparisons are important criteria in tracking temporal changes in insecticide susceptibility. For tarnished plant bugs in the Midsouth, most of the historical benchmarks are previous studies by the Snodgrass group that utilized field collections from Crossett, AR, as a source of benchmark susceptibility. Table 1 reports recent measurements of insecticide susceptibilities of tarnished plant bugs from the Crossett location.
The annual variability among field populations of tarnished plant bug in measured LC50s for acephate, imidacloprid, thiamethoxam, permethrin, and sulfoxaflor is illustrated in Figure 3. Table 2 summarizes the overall assay data for 2008-2015 with LC50s averaged across all significant regressions, including all significant assays with field populations and all significant assays

Results
Benchmark comparisons are important criteria in tracking temporal changes in insecticide susceptibility. For tarnished plant bugs in the Midsouth, most of the historical benchmarks are previous studies by the Snodgrass group that utilized field collections from Crossett, AR, as a source of benchmark susceptibility. Table 1 reports recent measurements of insecticide susceptibilities of tarnished plant bugs from the Crossett location.
The annual variability among field populations of tarnished plant bug in measured LC 50 s for acephate, imidacloprid, thiamethoxam, permethrin, and sulfoxaflor is illustrated in Figure 3. Table 2 summarizes the overall assay data for 2008-2015 with LC 50 s averaged across all significant regressions, including all significant assays with field populations and all significant assays conducted using the USDA lab colony. The USDA lab colony has only once previously been reported as an experimental control [14]. Tables 3-6 compare annual measurements of susceptibility to previous Snodgrass benchmarks [20] and assays with the more recent assays with insects from the Crossett control location (Table 1).

Acephate Assays
A total of 252 regressions described individual field and laboratory assays with acephate. Of these, 229 (91%) were measurements of different field collections with an average LC50 of 12.249 μg/vial ( Table 2). The average LC50 for the USDA laboratory colony (25.509 μg/vial) was significantly higher than the average LC50 for field collections based on a non-overlap of 95% fiducial limits (FL). The laboratory colony also exhibited wide variability in response to acephate (12-fold differences in LC50s among the 23 individual tests). Field populations varied more than 60-fold with a maximum LC50 of 68.669 μg/vial. The highest average LC50 was observed in 2006 (16.1 μg/vial). No assays were conducted with acephate in 2011 (Table 3 and Figure 2). When the collective acephate data were studied by analysis of variance (ANOVA), no significant difference was found in average LC50s among collections from the three different states (df = 2, F = 1.7919, p = 0.1691), though differences

Acephate Assays
A total of 252 regressions described individual field and laboratory assays with acephate. Of these, 229 (91%) were measurements of different field collections with an average LC 50 of 12.249 µg/vial ( Table 2). The average LC 50 for the USDA laboratory colony (25.509 µg/vial) was significantly higher than the average LC 50 for field collections based on a non-overlap of 95% fiducial limits (FL). The laboratory colony also exhibited wide variability in response to acephate (12-fold differences in LC 50 s among the 23 individual tests). Field populations varied more than 60-fold with a maximum LC 50 of 68.669 µg/vial. The highest average LC 50 was observed in 2006 (16.1 µg/vial). No assays were conducted with acephate in 2011 (Table 3 and Figure 2). When the collective acephate data were studied by analysis of variance (ANOVA), no significant difference was found in average LC 50 s among collections from the three different states (df = 2, F = 1.7919, p = 0.1691), though differences among years (df = 6, F = 4.0498, p = 0.0007) were detected. The highest annual average LC 50 was observed in 2014 (19.079 µg/vial) ( Table 3). LC 50 was negatively associated with latitude (n = 165, r 2 = 0.0259, Intercept ± standard error (SE) = 69.3435 ± 28.0621, Slope ± SE = −1.7512 ± 0.8407) but not longitude (n = 165, r 2 = 0.0075, F = 1.2348, p = 0.2681). Most of the populations had LC 50 s with a lower fiducial limit greater than 3.6 (the upper level fiducial limit (reported by Snodgrass as confidence limit) reported for the Crossett location [21]). From 4% (2015) to 52% (2008) of the average annual LC 50 s observed were significantly more than those at the Crossett location in 2014 (Table 3). a All LC 50 s expressed as µg of insecticide/vial. b 16.9 µg/vial is the highest upper level fiducial limit (reported by Snodgrass as confidence limit) reported for a field population [21]. c 3.6 µg/vial is the upper level fiducial limit (reported by Snodgrass as confidence limit) reported for the Crossett control population [21]. d 9.2 µg/vial is the upper level fiducial limit (reported by Snodgrass as confidence limit) measured for the Crossett control population in 2014 (Table 1)

Permethrin Assays
Wide variability was observed in the responses of field collections (n = 44) and the lab colony (n = 26) to permethrin assays in 2014 and 2015 (Table 2). We observed a similar range of responses to those previously reported by Snodgrass et al. [31] across the 44 field colonies tested in 2014 and 2015 (0.284-53.414 µg/vial), though a few populations were more susceptible ( Figure 2).
Using comparisons to former Snodgrass data (Table 4), none of the populations tested had 95% fiducial limits greater than the highest upper fiducial limit (reported by Snodgrass as confidence limit) reported for a field population (77.4 µg/vial) [31]. Three (7%) had LC 50 s with lower fiducial limits greater than the upper fiducial limit reported for the Crossett location, while 13 of the 45 (33%) had fiducial limits that did not overlap with the limits estimated from the Crossett location in 2014 (Table 1). One population of interest was a field population from cotton in Humphreys County, MS, that experienced control problems with thiamethoxam. The measured LC 50 for this population was 5.493 (2.080-16.226) µg/vial, high enough to be included in the group of colonies with fiducial limits greater than the Crossett control. A corresponding regression model with thiamethoxam for this collection location was eliminated from the overall summary because the model was not statistically significant.   267)). These suggested a 3.33-fold difference in LC 50 s measured at 24 h and 72 h for the USDA laboratory colony exposed to imidacloprid in the floral foam assays. The mortality was lower at 24 h as indicated by Snodgrass et al. [20], and the efficiency of making all observations on the same day facilitated the ability to conduct additional assays. While the change in methodology between research programs was not ideal, it may actually underestimate susceptibility. If assay data observed in 2014 and 2015 are divided by 3.33 to correct for differences in observation time as described in the methods section, the populations assayed in 2014 and 2015 were at least as susceptible as those assayed from 2008 through 2013.
Imidacloprid LC 50 s were significantly influenced by year of assay (df = 5, F = 6.1881, p < 0.0001). The average LC 50 for the 24 field populations assayed in 2015 (5.029 µg/vial) was greater than the average of the populations assayed in other years (Table 5). There were no observed differences among states (df = 2, F = 0.1981, p = 0.8205), and LC 50 values were not influenced by the latitude (n = 146, F = 0.7297, p = 0.3944) or longitude (n = 146, F = 2.5675, p = 0.1113) of the collection location.
Comparisons of LC 50 s to those previously reported by Snodgrass [20] and the 2008 measurement of imidacloprid susceptibility in the Crossett control generally found no evidence of a changed susceptibility to imidacloprid (Table 5). One collection made in 2008 from Concordia Parish, LA, had an elevated response (LC 50 (95% FL) of 6.470 (4.351-12.761)). In the 2014 and 2015 assays, a colony collected from a cotton field experiencing control problems in Tallahatchie County, MS, had an LC 50 of 13.997 µg/vial when measured at 24 h, and the estimated LC 50 at 72 h would be 4.203 µg/vial. Many of the field populations tested had LC 50 s with lower fiducial limits greater than 1.09 µg/vial (the upper fiducial limit (reported by Snodgrass as confidence limit) reported for the Crossett control [20]). Half of the collections tested in 2009 and 2010 had LC 50 s greater than the Crossett control population.

Thiamethoxam Assays
A total of 208 field populations were studied for susceptibility to thiamethoxam over an eight-year period ( Table 2). The average LC 50 was 2.066 µg/vial, with two populations in 2010 and two populations in 2014 exhibiting LC 50 s in excess of 10 µg/vial ( Figure 2). Two of the collections were made in August of 2014 (with a maximum LC 50 measured of 27.213 µg/vial) from the same cotton field that was experiencing control problems, taken approximately a week apart in Tallahatchie County, MS. The insects from this collection exhibited high LC 50 s to thiamethoxam, acephate, and permethrin. Thiamethoxam assays with the USDA lab colony (n = 16) produced LC 50 s slightly higher than those reported by Snodgrass et al. [20] for the field colonies, but not significantly different than the Crossett location (Table 1).
Similar to the imidacloprid assays, no differences were observed in LC 50 s for thiamethoxam among states (df = 2, F = 0.1581, p = 0.8539), but there was a significant effect of year (df = 7, F = 2.1952, F = 0.0363). However, Tukey's HSD test failed to separate differences among years at p = 0.05. The highest average LC 50 (Table 1).

Sulfoxaflor Assays
Benchmark data for sulfoxaflor were not previously obtained because of the unavailability of technical grade insecticides prior to the commercial use of the formulated product. Data collected in 2015 were taken several years after the initial use of the insecticide, and field populations may have already experienced some selection. For the 21 field populations tested in 2015 for susceptibility to sulfoxaflor using a glass vial bioassay and 24 h observations, an average LC 50 of 9.042 µg/vial was measured with 95% confidence limits of 4.924 to 13.159 µg/vial. The range in response across the field populations was high (179-fold) as compared to the range in response for the USDA Laboratory colony (11-fold). Average LC 50 s were similar for the field and laboratory assays ( Table 2).  Table 7 provides regression models that illustrate the connections among insecticide use across the different insecticide classes. Subsets of these data associated with individual counties and parishes for our sample sites ( Figure 1) were created and paired with our measured LC 50 s to study possible relationships between insecticide use and resulting LC 50 s. LC 50 s for the different insecticides were also examined for paired linear relationships. The resulting regression models are reported in Table 8. Acephate LC 50 (Table 8). There were no other significant relationships between thiamethoxam LC 50 s or sulfoxaflor LC 50 s with insecticide use and the insecticide assay data.

Relationships to Insecticide Use and Control Recommendations
The comparisons were further refined by calculating average LC 50 s for each county/parish to compare to the annual average insecticide use data for counties/parishes (Table 9). Significant (p < 0.05) regression models included positive relationships between kg of pyrethroids and kg of sulfoxaflor applied per hectare of cropland and resulting LC 50 s for permethrin (Table 9). Thiamethoxam LC 50 s were a positive predictor of imidacloprid LC 50 s, while the amount of pyrethroid insecticide applied per acre of harvested cropland was a negative predictor of imidacloprid LC 50 (Table 9).
Recommendations for the control of tarnished plant bug in cotton by the Mississippi Cooperative Extension Service are summarized in Table 10 for the years 1983, 1993, 2003, and 2013. These snapshots of time are presented to illustrate the evolution of plant bug control procedures and changing availability of and preferences for different insecticides. The recommended insecticides in 1983 and 1993 were almost entirely organophosphates, with the exception of the carbamate carbaryl in 1983 and oxamyl in 1993. The four organophosphates still recommended in 2013 were all recommended at use rates generally two-to three-fold higher than those recommended in 1983, with the exception of malathion, which was still recommended at a similar rates (Table 10). Three classes of insecticide chemistry were available for use in 2003, and six different types of insecticide chemistries were available for plant bug control in 2013.  Annual insect loss estimates included the number of insecticide applications for plant bugs, number of insecticide applications for bollworms, acres of cotton harvested, yield, percent crop loss to insects, average number of total insecticides sprays per year, and cost of all foliar insecticides. Significant correlations were observed between a variety of variables examined (Table 11). No significant correlations (p = 0.05) were observed for estimated percent crop loss to insects, average LC 50 s for sulfoxaflor, average LC 50 s for imidacloprid, or kg of sulfoxaflor applied per acre of harvested cropland.

Discussion
This paper continues a history of the reporting of assay responses of tarnished plant bug to the major classes of insecticide used in the Delta. There is a wealth of previous information on tarnished plant bug response to insecticides in this area, including early work [43,44], numerous papers from the Snodgrass laboratory [7,[14][15][16][17][19][20][21][22]45], studies in Arkansas [29,33], studies in Louisiana [46,47], a summary of small-plot field experiments in Mississippi [2], and a number of efforts to understand tarnished plant bug resistance mechanisms [1,23,[25][26][27][28]. Understanding the changes in susceptibility over time are important, but understanding how these changes evolve and how management practices can be refined with this information is particularly relevant to managing tarnished plant bugs in the future. Snodgrass [7] concluded a summary of 30 years of research with the tarnished plant bug in the Mississippi Delta by writing "as long as insecticides are the main method for controlling TPB (tarnished plant bug) in cotton, the TPB will remain a serious pest".
While the data presented here include previously unpublished resistance information from the Snodgrass laboratory, it also attempts to transition the research to new approaches. Research conducted in 2014 and 2015 introduced the use of the USDA lab colony as an experimental control. The historical use of collections from Crossett, AR, as an index of susceptibility may not be sustainable (Table 1), and the historical approach potentially confounds insect nutrition and age with measurements of insecticide susceptibility. Indirect comparisons back to previous published benchmarks are useful and important, but paired comparisons to a susceptible lab colony would strengthen experimental measurements and allow researchers to control experimental error. Both Zhu and Luttrell [23] and Zhu et al. [24] made comparisons between field-collected strains and a meridic diet-fed laboratory colony of tarnished plant bugs, and Fleming et al. [1] discussed their omission of a laboratory-susceptible strain and the possibility that all of their strains may have had some level of insecticide resistance. The high tolerance and variability of the USDA lab colony to acephate, permethrin, and thiamethoxam needs additional research if it is to be used as an experimental control. Regardless, there needs to be more experimental consistency and more understanding of the relationships between laboratory susceptibility and control in the field. Because of the high tolerance and variability observed with the lab colony, comparisons of insecticide susceptibility were based on previously published benchmarks from the Snodgrass laboratory using insects from Crossett and our more recent direct measurements of response from field collections of insects from the Crossett location (Table 1). Based on these traditional benchmark comparisons to the Crossett susceptible location, the USDA lab colony would be judged to be resistant to most of the insecticides tested. The variability of response in the USDA lab colony was greatest for acephate and permethrin assays, while the variability with the neonicotinoids and sulfoxaflor was less, perhaps suggesting the presence of resistance genes for acephate and pyrethroid resistance within the USDA lab colony. A more plausible explanation is likely linked to the differences in food between the lab colony's meridic diet and the field-collected insects that have fed on a wide range of nutritionally variable native plants and crops. Nutrition and host plant development have been shown to affect measurements of insecticide susceptibility in several other insect species [48][49][50][51][52].
Based on the acephate assay results of this study (Tables 2 and 3, Figure 2) and comparisons to declining levels of susceptibility reported by Snodgrass et al. [21], tarnished plant bug populations in the Delta region of Arkansas, Louisiana, and Mississippi have high frequencies of resistance to this organophosphate insecticide. However, there is still wide variability among populations and some populations appear to be relatively susceptible (Figure 2). Research by others [1,23,24] has found similar results and elevated activities of metabolic enzymes in field populations collected from the Delta region. Both Zhu and Luttrell [23] and Fleming et al. [1] have reported elevated levels of esterase activity in Delta populations of tarnished plant bugs. Zhu and Luttrell [23] associated these elevated levels with reduced susceptibility to acephate, while Fleming et al. [1] observed differences in esterase activities in bugs from two different regions of Mississippi, but did not associate the differences with individual assay results. Both Zhu and Luttrell [23] and Fleming et al. [1] measured variable levels of glutathione S-transferase in the Delta populations of tarnished plant bug. Zhu and Luttrell [23] indicated that inhibitors of glutathione S-transferase exhibited less suppression in 2010 as compared to 2006, and suggested that this may indicate a potential shift in the genetics of the pest populations. Recommended application rates of acephate were increased from 0.23-0.33 lb ai/acre to 0.5-1.0 lb ai/acre during this time period (Table 10), indicating a growing concern for the level of field control being achieved. Reed et al. [2] summarized results of replicated field experiments conducted to measure tarnished plant bug control in cotton with organophosphate insecticides, and reported that the average control measured from 1982 to 1997 was 57%. If this level of field control compares to the previous assay data of Snodgrass et al. [21], and field control is even loosely related to changes in assay response, acephate applications alone do not adequately control tarnished plant bugs in the field. Bioassays were conducted with permethrin only in 2014 and 2015, as most of the previous systematic monitoring for pyrethroid resistance was based on the use of a diagnostic-dose assay [18,20]. Based on our 2014 and 2015 assays ( Figure 3 and Table 4), tarnished plant bugs still express resistance to permethrin. Pyrethroid-resistant populations are common, but there is wide variability in response among populations and some are relatively susceptible. Perhaps, a return to careful monitoring of populations prior to spraying a pyrethroid would enhance the efficiency of insecticide selection decisions and allow growers to more carefully determine when to use pyrethroid insecticides [18].
Snodgrass et al. [21] reported variability in the response of tarnished plant bugs to both imidacloprid and thiamethoxam, and suggested that some imidacloprid resistance was present. At the time, it was also generally concluded that the neonicotinoids were the only widely used insecticides to which tarnished plant bug populations are still susceptible. In our studies (Tables 2 and 5, Figure 2), we found no evidence of increasing resistance to imidacloprid, although the populations commonly had higher LC 50 s than the Crossett collections. With thiamethoxam, a few colonies collected from cotton fields experiencing control problems, in particular the colony collected from Tallahatchie County, MS, in 2014, were reason for concern, as was a population with an elevated response in 2010 ( Figure 2). Because of the importance of these insecticides to the effective control of tarnished plant bugs, and the possible selection for elevated esterase, P450, and glutathione S-transferase genes that may confer resistance to multiple classes of insecticides [24], new approaches to tarnished plant bug control that lessen the current intensive and repeated use of insecticides of all classes are needed. As with the other insecticide classes, additional research is needed to associate variability in laboratory assays to insect survival and crop damage in the field.
Those bioassays conducted with sulfoxaflor may serve as future benchmarks, but these data were collected several years after the insecticide chemistry was commercially deployed and variability was observed in our 24 h glass vial assays ( Figure 3, Table 2). Additional research is needed to refine the assay methods and perhaps extend the observation time, but as with the other insecticides, these studies need to be related to field observations of plant bug survival. The lack of benchmark information pre-commercial release of sulfoxaflor may hinder future resistance assessments.
Estimated insecticide use in Arkansas, Louisiana, and Mississippi varied considerably across years and states depending on the chemical class. Regressions between the estimated use of the four insecticide classes were all highly significant when all three states were pooled. Thiamethoxam LC 50 s were a positive predictor of imidacloprid LC 50 s, while the amount of pyrethroid insecticide applied per acre of harvested cropland was a negative predictor of imidacloprid LC 50 . The subset data utilizing only counties and parishes with collection records revealed linkages between LC50 values and estimated use for several classes of insecticides ( Table 8). The observations of linkages between pyrethroids and neonicotinoids in measured LC 50 s and the positive influence of estimated pyrethroid use on neonicotinoid susceptibility is concerning, especially in light of the research of Zhu and Luttrell [23], Zhu et al. [24], and Fleming et al. [1] that report elevated levels of broad-based metabolic enzymes capable of detoxifying multiple classes of insecticides.
Thirty years of tarnished plant bug thresholds and insecticide recommendations from Mississippi State University's Cooperative Extension Service provide snapshots of the evolution of thresholds, changes in control strategies, and availabilities of different classes of insecticides. Thresholds have been refined [12,13] with adjustments for relative critical densities among different sampling procedures, and treatment levels have generally decreased (i.e., sprays recommended at lower plant bug densities) as the recency of recommendations increased (Table 10). For the first time since pyrethroids were introduced into Mississippi cotton production in 1979 [53], specific pyrethroid insecticides are now listed in combination with other insecticides as recommended options for the control of plant bugs and fleahoppers [54]. Additionally, pyrethroids are also no longer recommended for the control of bollworms and budworms for the first time since 1979. Correlations utilizing data from the annual loss estimates [35] are not unexpected, with significant positive relationships observed between the total numbers of foliar applications, the number of applications made for tarnished plant bugs, and the cost of all foliar applications.

Conclusions
The exploratory examination of associations among measured plant bug susceptibility, reported insecticide use for individual counties/parishes, and estimated cotton insect losses and control costs is likely influenced by a number of spurious relationships, including temporal and spatially dynamic patterns of cotton acreage across the landscape. Hopefully, the detailed information on tarnished plant bug susceptibility to insecticides in the Delta over time will allow future pest managers to develop more efficient and sustainable insect management programs with lessened use of insecticide. Developing alternative control options should be a priority for future research [55]. Given the wealth of historical assay data on tarnished plant bug susceptibility to insecticides and the ability to link this historical information to future research, a priority should be placed on field research, especially additional empirically based research comparing varying levels of tarnished plant bug control across the range of insecticide tools available and including assays to link historical data. Empirical, field-based research will enhance our understanding of how the variable expressions of survival in assay experiments relate to the survival of tarnished plant bugs and crop damage in insecticide-treated fields.