Impact of Hemlock Woolly Adelgid and Elongate Hemlock Scale on Early Seasonal Development of Eastern Hemlock

Herbivores can have a significant impact on plant host development. While altered plant development most commonly results from defoliation, some causes are more cryptic. Effects on development can be difficult to detect in long-lived woody plant species. Hemlock woolly adelgid is a piercing-sucking herbivore that has been previously been observed to have substantial effects on eastern hemlock. Observations of bud break were carried out on a three-day-on, one-day-off rotation from April through May in 2016 and 2017. We found that hemlock woolly adelgid delayed bud break in eastern hemlock. Our findings suggest that piercing-sucking herbivores can also significantly affect the early development of long-lived woody plant species.


PREFACE
The following thesis will be submitted as a single chapter, in manuscript format following the formatting guidelines of the scientific journal Ecology.
v  done by feeding guilds such as leaf chewers and twig browsers is easily quantified, but some effects are far less obvious (Preisser and Bastow 2006). These 'cryptic' impacts can include resource loss or reallocation, defensive induction, and other alterations to plant physiology (Karban and Baldwin 1997). Because these changes often occur internally and take time to manifest, their effect on plant fitness can be difficult to quantify, and the impact of herbivores on their host plants may be underestimated.
While plant phenology is largely driven by abiotic factors like temperature and day length (Schwartzberg et al. 2014, Rossi andIsabel 2017), the influence of herbivores has received increasing attention. These impacts can be difficult to detect in long-lived plants like trees because herbivory and the resulting phenologic changes often occur in separate growing seasons, as when summer defoliation delays budburst the following spring. Several studies on outbreaking folivores have found that severe (50-100%) defoliation is required to alter bud break even slightly (Haukioja et al. 1988, Kaitaniemi et al. 1997, Quiring and McKinnon 1999. These phenologic effects may reflect altered plant resource levels and allocation (Tuomi et al. 1989), and may also affect the synchrony between bud break and folivore emergence.
The hatching of many early-season herbivores is timed to coincide with the presence of young leaves. Because newly-hatched folivores are highly vulnerable to starvation, even short delays in bud break may be harmful (Van Asch and Visser 2   2007). Several studies, however, have found that these delays do not affect, and may actually benefit, some herbivores (Kaitaniemi et al. 1997, Carroll andQuiring 2003).
Such mismatches should have even less effect on herbivores that do not consume leaves, and phenologic alteration by a non-folivore has only been reported once: Quiring and McKinnon (1999) found that moderate-to-high densities of the galling insect Adelges abietis delayed budburst by a single day.
Although herbivore-driven changes in tree phenology have been documented in multiple plant species, this research has been conducted on plants growing in moderate-to full-light conditions (Tuomi et al. 1989, Quiring and McKinnon 1999, Carroll and Quiring 2003. In such environments, the consequences of a delay in photosynthate production from new tissue is likely to be minimal across an entire growing season. In contrast, the seedlings of many canopy-dominant tree species begin their lives in shady forest understories (Canham et al. 1990 (Ellison et al. 2005). It is threatened throughout its native range by an invasive sap-feeding insect, the hemlock woolly adelgid Adelges tsugae ('HWA'). This insect is capable of killing even mature trees within four years (McClure 1991b), and few infested trees survive for more than a decade (Orwig et al. 2002). Even though HWA-mediated mortality of mature hemlocks in the Northeastern United States has been slowed by cold winter temperatures, hemlock seedling/saplings have virtually disappeared from HWA-invaded hemlock stands throughout this region ).
We report the results of three separate field experiments, conducted in consecutive years, measuring the impact of HWA infestation on bud break in understory eastern hemlock saplings. We also assessed the impact of another similarly-sized invasive phytophagous insect, the elongate hemlock scale Fiorinia externa ('EHS'), whose densities equal or exceed those of HWA but is less harmful to 4 hemlock health (Gómez et al. 2015). We hypothesized that HWA, but not EHS, would alter the timing of hemlock bud break, and that the magnitude of this effect would increase at higher insect densities.

2015
Experiment #1 were used over each cage to protect against cross-treatment contamination. Trees were initially randomly assigned to one of four herbivore treatment of either: control (no HWA or EHS), HWA-only, EHS-only or both (Wilson et al. 2018).
Starting in 2011, trees were manually inoculated annually with either neither insect (control), HWA, EHS, nor both. The timing of these inoculations reflects the natural dispersal period for HWA and EHS; both insects overwinter as adults whose eggs hatch into mobile crawlers (McClure 1989). The dispersal of these crawlers is aided by the relatively high sub-canopy wind velocities that occur prior to canopy leafout (McClure 1990). Because HWA disperses approximately one month earlier than EHS, we inoculated them in April and May, respectively. We used foliage infested with each insect to inoculate the appropriate trees, a standard protocol (Butin et al. 2007); control trees were 'inoculated' with herbivore-free foliage to control for inoculation-related disturbance. These treatments would continue for four years 6 (HWA-4, EHS-4, and Both-4, respectively; 13, 9, and 6 trees respectively). In 2013, some HWA-and EHS-only trees were thereafter annually inoculated with both insects, creating two 'priority effect' treatments (i.e. HWA  Both, and EHS  Both; 9 and 12 trees respectively). In the same year (2013) some trees that were uninfested (controls) were transitioned into HWA, EHS, and both for two years (control  HWA-2, control  EHS, and control  Both-2; 10, 9, and 7 trees respectively; Wilson et al. 2018 each block contained eight rows and columns, with 1-1.5m between individual trees. Chicken-wire cages were placed over each tree to protect them from deer browse. To protect against cross-treatment contamination, mesh bags (Agribon-15, Johnny's Selected Seeds, Waterville, ME, USA; 90% light transmission) were placed over each wire cage; we used ground staples to attach each mesh-enclosed cage to the ground.
Within each block, 16 trees were randomly assigned to one of the following three treatments: control (no HWA or EHS), HWA-only, or EHS-only. There were 16 "both" treatments in blocks that were used in unrelated experiments but were not monitored for phenology.
In mid-spring 2014 and 2015 we inoculated trees in each treatment with the appropriate insect. Inoculations were carried out in the same manner as in 2015.
2016 data collection: In early spring 2016, we selected trees from four of the five 64-tree blocks for phenology monitoring; we excluded the fifth block because deer had broken into many cages. Within each block, we chose five trees from each of the three treatments for monitoring, a total of 60 trees (three treatments * four blocks * five trees). All selected trees had zero or very low densities of gypsy moth (Lymantria dispar) and spruce spider mite (Oligonychus ununguis), two herbivores common at our field site that occasionally colonized trees in our experiment.
In early April 2016, we surveyed all branches emerging from the main stem of each experimental tree to determine branch-level insect densities (measured in insects 8 cm -1 branch). Because of our interest in assessing the impact of herbivory on bud break, we selected the highest-insect-density branch on each tree (the 'marked branch') for the bud phenology survey detailed below. We recorded the length of each marked branch, defined as the distance from the main trunk to the furthest terminal bud.
On April 15th, prior to the start of bud break, we began daily counts of the opened terminal and axillary buds on the single marked branch. Daily counts were taken for three days, followed by a one-day break; this 3-1 rotation continued until May 20th, one week after the last bud-break was observed.
After all bud count data were collected we determined the date of first, mean, and last bud break for each tree. First bud break was defined as the first day that a tree had one or more open buds. Mean bud break was defined as the day at which at least half of the buds that would break had broken bud. Last bud break was defined as the first day on which all of the buds that would break had broken. In early April 2017, we chose one branch from each remaining tree, recorded its length, and measured insect density on it (insects cm -1 branch). As with the year before all selected trees had zero or very low densities of gypsy moth (L. dispar) and spruce spider mite (O. ununguis), which were common in the field experimental site for 2016-2017 data collection.
On April 12th (prior to the start of bud break), we began conducting daily counts of the opened terminal and axillary buds on each marked branch. Daily counts were taken for three days in a row, followed by a one-day break; this continued until May 26th, one week after bud-break ceased.

2015:
We fit linear mixed effects models and used a backward-model-selection approach to examine the individual and interactive effects of HWA and EHS on bud break. HWA and EHS were treated as fixed factors, each with three levels corresponding to the length of infestation (0, 2, or 4 years) and an interactive term (HWA × EHS). Random effects included in models are described below. Full and reduced models were ranked and compared based on Bayesian Information Criterion (BIC) values, a standard criterion for model selection. The lme4 package was used to generate and compare models (Pinheiro et al. 2014).
In the full model, initial trunk diameter at planting was included as a covariate and row position (1-20) of each tree was included as a random effect in the linear mixed effects model. We used this approach to examine how HWA and EHS affected the date of bud break. All analyses were performed using R v. 3.2.2 (RCoreTeam

2016:
We first analyzed among-treatment differences in the date of first, mean, and last bud break, as well as the number of buds broken. We excluded data from a single EHS tree on which only a single bud broke (i.e., same date for all three bud break variables). We used mixed linear models (normal distribution, identity link) fit using restricted maximum likelihood methods, with 'branch length' and 'block' as covariates. Chi-square and p-values were obtained for response variables via L-R χ 2 tests.
We also analyzed the within-treatment relationship between insect density and the date of first, mean, and last bud break, as well as the number of buds broken. We used mixed linear models (normal, identity) fit using restricted maximum likelihood methods. Both 'branch length' and 'block' were included as covariates. P-values were again derived using L-R χ 2 tests.

2017:
Our analysis of the 2017 data was identical to that used for the 2016 data, with the following two differences. First, we excluded data from a single HWA tree on which only a single bud broke. Second, since the 'block' term was never significant as a covariate, it was removed and the analyses were rerun without it.
The 2016 and 2017 analyses were conducted using JMP 10 (SAS Institute, Cary North Carolina USA).

2015:
Adelgid feeding delayed bud break by 3-4 days (F 2 ,79 = 8.88, p < 0.001; Fig. 1a). Adelgid-free trees broke bud on average by May 7th, whereas trees infested with adelgid for two and four years had a mean bud break date of May 10th and 11th, respectively. In contrast, scale did not alter bud break (Fig. 1).
There were, however, no treatment-level differences in the date of first, mean, or last bud break (all p > 0.2). Insect densities in the HWA treatment ranged from zero to 0.56 cm -1 (mean: 0.11 ± 0.035); EHS densities ranged from zero to 0.84 cm-1 (mean: 0.26 ± 0.067).
EHS density did not affect most of the measured variables (p > 0.2 for all). The one exception was the date of first bud break (χ 2 = 5.52, p = 0.019), where increasing 12 EHS density from zero to 0.56 insects cm -1 accelerated bud break by 4.3 days (r 2 = 0.14).

2017:
HWA density on marked branches ranged from zero to four insects cm -1 (mean: 1.32 ± 0.335 [SE]). There was a positive relationship between HWA density and the date of first bud break ( Fig. 2d; χ 2 = 4.47, p = 0.0345) and mean bud break EHS density ranged from 0.2 to 33 insects cm -1 (mean: 5.90 ±1.970). There was no relationship between scale density and any of the examined variables (all p > 0.10).

DISCUSSION
In this study, we found that a piercing-sucking herbivore can have residual effects on eastern hemlock phenology, as hemlocks that were infested with HWA were delayed in bud break the following spring. There was, however, substantial annual variation in the effect of HWA on bud break. For example, depending on year, trees with moderate to high (2016: 0-0.56 insects cm -1 ; and 2017:0-4 insect cm -1 ) HWA densities, densities that were representative of the field experiment site, broke bud anywhere from 1.4 to 14.0 days later than control or EHS-infested trees. Despite this difference, there were no treatment differences or density effects on overall bud production. The HWA-caused delay in bud break could have major implications for this conifer, as its understory saplings rely heavily on early spring carbon-acquisition for growth, which almost entirely occurs prior to overstory canopy leaf-out in the spring (Hadley 2000b).

HWA vs. EHS
EHS was included in this study as a means of observing another prevalent invasive within the eastern hemlock system. EHS feeding occurs through similar mechanisms to that of HWA, although it feeds on needles rather than at the twigneedle junction (McClure 1979). The separation in feeding locations allows these two insects to coexist on a shared host (McClure 1991a, but see Gomez et al 2014).
Whereas HWA has been shown to manipulate tree chemistry (Gómez et al. 2012), EHS is known mostly to cause discoloration in needles and, if in high enough densities, kill a hemlock in approximately 10 years (McClure 2002). HWA results in a hypersensitive-like response, where hydrogen peroxide (Radville et al. 2011), amino acids (Gómez et al. 2012), and salicylic acid ) accumulate, and the emission of methyl salicylate increases (Pezet et al. 2013, Pezet andElkinton 2014). HWA also appears to cause water stress and false ring formation (Gonda-King et al. 2012, Domec et al. 2013), but EHS does not stress trees to the same extent.
These differences in effects between HWA and EHS in the eastern hemlock system could contribute to the reasons why we did not observe a delay in bud break in eastern hemlock in the presence of scale.

Relevance of new flush
Hemlock is similar to other conifer species (ex. Abies sp.) in its long period of needle retention (Baiser et al. 2014). While this may allow for continued Bielinis (2017) found that net CO2 assimilation rate in Abies alba decreased over time with needle age, and that CO2 assimilation rates of one-year-old foliage were six times higher than that of seven-year-old needles (Robakowski and Bielinis 2017).
Abies sp., like eastern hemlock, are relatively shade-tolerant conifers whose dense canopies can easily shade out their own foliage. Similar relationships between CO2 assimilation and foliar age have been found in a variety of Abies, Picea, and Pinus sp. (Freeland 1952, Weikert et al. 1989, Wang et al. 1995, Brooks et al. 1996, Kitajima et al. 1997, Olekysn et al. 1997, Jach and Ceulemans 2000, Crous and Ellsworth 2004, and Warren 2006.

Impact of herbivore-induced phenologic delays on plant performance
Photosynthesis in hemlock does occur through the autumn and winter months, but at substantially reduced levels due to cold temperatures and low photon flux densities (PFD) (Hadley 2000a). Photosynthesis through the winter months (roughly December through March) accounts for only about 10% of the annual total of photosynthesis for hemlock (Hadley 2000b). Indeed, throughout the months of April, May, and into July, hemlocks in the understory of a deciduous canopy experience maximum carbon gain due to high enough levels of PFD (Hadley 2000b, Hadley andSchedlbauer 2002). In early spring after deciduous canopy leaf-out, understory hemlocks subsist on a drastically-reduced quantity and quality of light, and 'sunflecks' (i.e., short periods of direct sunlight lasting on the scale of minutes) become the only source of light capable of activating meaningful (i.e., carbon-fixing) levels of photosynthesis (Hadley 2000b).
After leaf abscission in the fall, the low solar angle (and therefore low PFD) reduces light levels enough to preclude substantial carbon fixation. This means that early spring, prior to canopy leaf-out, is the critical growth window for understory hemlocks.
The narrow growth window for understory hemlocks, limited in the early spring by temperature and solar angle and in the early summer by canopy leaf out, means that delays in spring bud break and the full leaf-out of new foliage are likely to be harmful with respect to putting on new growth. Hadley (2000) found that during the summer (after canopy leaf out) understory hemlocks get an average PFD of 20µmol m-2s-1; only in instances of strong sun flecks did this value rise to 100µmol m -2 s -1 . In contrast, during early spring hemlocks regularly experienced PFDs above 200µmol m -2 s -1 (Hadley 2000). Between the early spring and summer, based on Hadley's (2000) PFD values, hemlocks experience at least a 50-90% decrease in available photon flux density. The same research also showed that photon flux density shared a linear relationship with net photosynthesis (Hadley 2000). HWA-infested understory hemlocks would thus gain less photosynthate than uninfested trees.
In addition to photosynthate losses caused by delayed-bud break-related decreases in PFD availability, HWA infestation can also reduce hemlock needle biomass (Wilson et al. 2018). Hemlocks infested with HWA have 15% lower above-/belowground biomass ratios and 16% lower aboveground needle/woody biomass ratios, than non-infested controls (Wilson et al. 2018). These reductions are likely to interact with the negative effects of delayed bud break to strongly inhibit the growth of HWAinfested understory hemlocks.

Herbivore induced delays and herbivore performance
Herbivore-induced shifts in the timing of bud burst during the following season can have major implications for the population dynamics of subsequent herbivores (Quiring and McKinnon 1999). Shifts in bud break phenology due to previous-season herbivory have been documented in woody plants primarily for defoliating herbivores (Haukioja et al. 1988, Kaitaniemi et al. 1997, Quiring and McKinnon 1999. However, Quiring and McKinnon (1999) reported a slight (one day) but significant delay in white spruce (Picea glauca) bud break following infestation of the pineapple galling adelgid, Adelges abietis. The authors attributed this delay to the ability of the gall to act as a sink, pulling resources away from shoots. In our system, the magnitude of HWA-mediated delay in bud break (e.g. a 14 day-delay in 2016; Figure 2), coupled with the scale of the HWA invasion in North America, means that this stylet-feeder could have enormous effects on other herbivores that rely on a hemlock as a food source. Many hemlock folivores such as the hemlock looper (Lambdina fiscellaria) are dependent on the synchronization between larval emergence and bud break (Butt et al. 2010).
Chewing herbivores such as the hemlock looper may not be the only species affected by a delay in bud break. The summer-emerging sistens generation of HWA depends on new foliage that develops from burst buds (McClure 1989). The ability of piercing-sucking insects like HWA to insert a stylet into foliage that is more lignified could be advantageous. Delays in bud break and inevitable canopy closure from a deciduous leaf out in the canopy could give second generation crawlers better potential synchrony for settling on hemlock foliage that takes longer to fully elongate and may be more lignified.

Future directions
External environmental factors can also significantly impact phenological timing (Chmielewski andRötzer 2001 andCleland et al. 2007) and a warming climate may result in earlier bud break. Many deciduous trees can accelerate the timing of bud break to coincide with earlier onset of spring (Kramer et al. 2000, Sparks et al. 1997, 2000, Chmielewski and Rötzer 2001. The effect of climate change on conifer phenology is less clear, with most studies pointing to genus-or species-specific responses to altered climates and local environmental changes (Kramer et al. 2000, 18 Royce and Barbour 2001, Richardson et al. 2009, Ribbons 2014. Future research might address the interaction between climate change and delayed bud break in this system.
While bud break has largely been explored in relation to external environmental factors and folivore impacts, piercing-sucking insects such as HWA can also alter the phenology of long-lived woody plant species. Our findings highlight a cryptic but