Multi-trait selection to build resilience in conifer forests: a case study on spruce-shoot weevil interactions

Tree planting programs now need to consider climate change increasingly, therefore, the resistance to pests plays an essential role in enabling tree adaptation to new ranges through tree population movement. The weevil Pissodes strobi (Peck) is a major pest of spruces and substantially reduces lumber quality. We revisited a large Interior spruce provenance/progeny trial (2,964 genotypes, 42 families) of varying susceptibility, established in British Columbia. We employed multivariate mixed linear models to estimate covariances between, and genetic control of, juvenile height growth and resistance traits. We performed linear regressions and ordinal logistic regressions to test for impact of parental origin on growth and susceptibility to the pest, respectively. A significant environmental component affected the correlations between resistance and height, with outcomes dependent on families. Parents sourced from above 950 m a.s.l. elevation negatively influenced host resistance to attacks, probably due to higher P. engelmannii proportion. For the genetic contribution of parents sourced from above 1,200 m a.s.l., however, we found less attack severity, probably due to a marked mismatch in phenologies. This clearly highlights that hybrid status might be a good predictor for weevil attacks and delineates the boundaries of successful spruce population movement. Families resulting from crossing susceptible with resistant parents generally showed fast-growing trees were the least affected by weevil attacks. Such results indicate that these “hybrids” might be genetically better equipped with an optimized resource allocation between defence and growth and might provide the solution for concurrent improvement in resistance against weevil attacks, whilst maintaining tree productivity.

leading to a diminished sink strength [5]. 23 To keep pace with the current speed of climate change in terms of sustainable forest 24 productivity, effective management of forest genetic resources must be developed. The 25 assisted gene flow approach implementing transfer functions [6] shifts genetic material 26 from different provenances towards their future optimal growth conditions to maintain 27 the current level of productivity. However, with change in average annual temperature 28 and precipitation, there is also increased frequency of biotic and other abiotic stresses 29 that will require establishing higher tolerance not only to drought, but also to an 30 increased risk of frost damage due to rapid dehardening under a warming climate, and 31 attacks from various pathogens. For Canada, boreal conifer forests are especially at risk 32 for a reduction in timber volume, and by 2100, the unaffected wood volumes are 33 projected to be lower than those currently harvested [7]. Therefore, the implementation 34 of resiliency measures to cope with these disturbances is paramount for a sustainable 35 forestry. Aside from crucial intervention measures for forest insect pests and pathogens, 36 tree breeding for resistance against these forest nuisances is urgent to limit their burden 37 on forest ecosystems. 38 To achieve tolerance to various stressors, Janes and Hamilton [8] emphasized the 39 importance of hybrid zones to select individuals with a favourable combination of 40 features from both parental ancestries. Hybridization enhances broader genetic 41 variation and increases the potential for future adaptation which might be important 42 especially under climate change [9]. Additionally, hybrids are usually more adapted 43 under a broader scale of environmental conditions, while the success of hybridization 44 depends on the phylogenetic distance of the parental species [10,11]. In some cases, 45 however, there are obvious reproductive barriers due to large phylogenetic distance, 46 causing structural or physiological incompatibility of pollen tubes to establish viable 47 embryos [12]. Therefore, species hybridization is not a universal solution for the 48 generation of resilient genotypes for future environments, and its appropriateness needs 49 to be evaluated on a case-by-case basis. 50 Spruces are the most important reforestation species in Canada, with hundreds of 51 millions of seedlings planted each year. Interior spruce, a natural hybrid, has a long 52 history of introgression between white spruce (Picea glauca (Moench) Voss) and 53 Engelmann spruce (P. engelmannii Parry ex Engelm.) as shown by the high levels of 54 recombination events uncovered in hybrid populations [13]. Analyses using genetic 55 markers also revealed asymmetric introgression from white spruce into the local 56 Engelmann spruce species as the initial founder events of the Interior spruce hybrid 57 zone, with eastern British Columbia as the centre of this hybrid zone [13,14]. Still, the 58 January 15, 2022 2/19 ancestral species maintain their integrity due to environmental selection and limited 59 recent interspecific gene flow (ibidem). In fact, both spruce species occupy distinct 60 ecological niches. While the boreal white spruce has an extremely broad longitudinal 61 range of 111°, the habitat of the mountainous Engelmann spruce is fragmented in 62 western North America covering a much narrower longitudinal range of 23° [13,15]. 63 Pure species occur below 600 m a.s.l. (P. glauca) and above 1,800 m a.s.l. (P. 64 engelmanii ) according to recent hybrid index assessment using genetic markers [16]. 65 Interior spruce is also one of the economically most important species in western 66 Canada and, due to its hybrid nature, can provide genetic resources needed to cope 67 with climate change [17]. The white pine or spruce shoot weevil (Pissodes strobi (Peck); 68 Coleoptera: Curculionidae) is one of the most important biotic threats for Interior 69 spruce [18] and the coastal Sitka spruce (Picea sitchensis (Bong.) Carr) species [19] 70 reforestation in western Canada. The weevil is most damaging to young trees [20]. North America, is also susceptible to the weevil [21]. The white pine weevil's life cycle 74 starts with adults overwintering in the litter on the ground. From April to May, adults 75 reemerge and start feeding on the terminal shoot. They favour feeding on the bark 76 (phloem tissue) that is close to the dormant terminal buds. Female weevils lay their 77 eggs into the bark of the terminal shoot, and eggs hatch shortly thereafter (7-10 days). 78 The larvae continue feeding on the leader until July when they reach maturity, and 79 adults emerge after 10-15 days and continue feeding on old and new growth [22]. Thus, 80 terminal growth can be severely affected by weevil feeding and significant losses in 81 productivity due to substantial stem deformations are the anticipated consequence, 82 rendering a spruce plantation meant to provide sawlogs at rotation age worthless.

83
Adding to this, usually it is the tallest and best growing trees that are preferentially 84 infested by the weevil [22]. Adult weevils can persist over several years and the most 85 important natural control of weevil population numbers is through winter mortalities.

86
However, the ongoing shift to milder winters due to climate change might now create 87 the optimal conditions for maintaining higher weevil population numbers and thus 88 increase the frequency of infestation outbreaks. Therefore, resistance against white pine 89 weevil attacks might have to gain higher importance in breeding objectives for spruces. 90 While conifers have more or less effective defence mechanisms in place [23,24], the 91 expression of these defences depends on several factors that is the genetic makeup of the 92 host tree and the local environment where host and pest coexist [25]. For example, it 93 was shown for Sitka spruce that synchrony of weevil activity and budburst phenology of 94 the host is determinant for successful oviposition. Faster budburst phenology of the host 95 may contribute to resistance to the white pine weevil [26]. Environmental cues that 96 indirectly influence host preference include overstory shading (shade-grown trees are less 97 preferred [27]) and fertilizer application (fertilized trees due to their increased bark 98 thickness and leader size enhance their attractiveness for weevil oviposition [28,29]).

99
One of the best studied direct host defences are the bark resin canals related chemical 100 defences [30] that feature oleoresin blends of a great variety of terpenes, generated 101 thanks to an important functional diversity of terpene synthases found in spruces [31]. 102 However, geography and climate may play an important role in the effectiveness of such 103 resin canal defenses, as shown in trees from the northwestern British Columbian zone of 104 introgressions between Sitka and white spruces [32]. Based on the pure species' 105 biogeography, O'Neill and colleagues concluded that higher P. glauca proportions in 106 hybrids meant increased resistance to the weevil. In general, host tree defences are most 107 effective against local pests, while they weaken under relaxed or absent pest pressure.
did not coevolve with the pest, since they were never exposed to this threat, and 111 therefore, are less likely to express the appropriate defense response.

112
The selection of initial plus trees entering the breeding program, and the subsequent 113 selection of parental individuals for advanced generation breeding, has favoured trees 114 with superior growth characteristics. Therefore, screening for resistance of seed sources 115 usually occurs posterior to the detection of a pest or pathogen problem within the 116 breeding program. Therefore, a way to perform mass screening for suitable genotypes is 117 to establish new progeny trials from known seed sources and use artificial infestations to 118 assess genetic control of resistance [33]. These trials can further be used to evaluate 119 how resistance (or tolerance) covaries with growth rate. To do so, the growth rates of 120 tree hosts need to be known before a pest or pathogen occurs to avoid any assessment 121 bias. Long-term breeding programs, such as for the Interior spruce complex Picea glauca 122 (Moench) Voss × P. engelmannii Parry ex Engelm., exist. It was started in 1973 with a 123 seed collection from 173 'plus tree' Interior spruce individuals growing in natural stands 124 across north central British Columbia (Prince George Selection Unit), where 125 subsequently four progeny test sites involving these open-pollinated (OP) families were 126 established, with subsequent evaluation of the genetic control of juvenile height growth 127 [18,34]. Furthermore, due to the presence of endemic weevil populations at three test 128 sites, resistance rankings of parents based on retrospective weevil damage assessments 129 were performed over the accumulated sustained damages in each of the 173 OP families 130 [18,35]. These resistance rankings of parents formed the basis for the selection of 131 parents for the progeny trial at Kalamalka Research station (Vernon, British Columbia) 132 that involved controlled crosses that were generated by mating resistant with resistant, 133 resistant with susceptible, and susceptible with susceptible parents (cross types; [35]). 134 At 5 years of age (in the fall of 1999), this entire Interior spruce trial was exposed to 135 herbivory through an artificially augmented weevil population occurrence. This 136 established the first large white pine weevil trial for spruces in Canada, and screening 137 for weevil attacks took place from 2000 to 2003 [35]. The outcome of this progeny 138 assessment confirmed for the most part the original parent groupings into resistant or 139 susceptible [18]; certain bark characteristics were confirmed as potential indirect 140 screening tools for resistance in spruce families [35]. However, stable positive genetic 141 correlations between resistance and height growth would be necessary for an effective 142 multi-trait improvement program [18,36,37]. First assessments of the relationship 143 between height growth (the directly selected trait) and weevil resistance (the indirectly 144 assessed trait by recorded attacks within one OP family progeny test site with 4,330 145 trees from 139 families) were conducted from the late 1980ies to the early 1990ies in 146 juvenile Interior spruce [20,36]. The assessment of attacks took into account the total 147 attacks recorded and three individual years of height assessments during this period.

148
Results indicated overall a strong negative genetic relationship (r = -0.61) between 149 growth and weevil attack. As noticed by the authors, tree heights might also reflect 150 early leader loss and damage for more susceptible families [36]. However, this 151 represents an important confounding factor rendering tree height the result of the level 152 of resistance to pest attack rather than an independent trait. Therefore, the 153 establishment of dedicated progeny trials, where artificial weevil infestations are the 154 tools to screen for weevil resistance and assess whether growth assessed before the 155 attack could be related to subsequent attack severity, becomes crucial [33,35]. 156 In the present study we revisit the entire Interior spruce weevil resistance screening 157 trial encompassing 2,964 individual genotypes (42 families) of varying susceptibility, 158 established in the interior of British Columbia [35]. The same trial was visited 159 previously by the authors [38][39][40] to investigate a two-by-two factorial spruce progeny 160 made up of families widely segregating for resistance (resistant-by-susceptible cross 161 type) and identify individual candidate genes for resistance to the weevil. These studies 162 January 15, 2022 4/19 also pointed at potential pleiotropic relationships between pre-attack growth and pest 163 resistance, thus our interest in revisiting the weevil trial in more detail. Here, we want 164 to clarify the relationships between resistance to the weevil and, for the first time, 165 pre-attack growth within and across all cross types. We also make direct use of The multivariate mixed linear model using MCMC algorithm implemented in the JWAS 205 package [42] was used to obtain the posterior distribution of variance/covariance 206 parameters as follows:  [38]. Forty-two controlled crosses were generated by mating resistant with resistant, resistant with susceptible, and susceptible with susceptible parents. The ranking in terms of weevil resistance of the individual Interior spruce parents is known and was done previously [18]. The evaluation of attacks was done following artificial weevil infestation of the spruce resistance trial. Height data from pre-attack years are shown. The colour code for ascending phenotypic values (that is for classes of attack severity, egg punctures occurrence, heights) is provided to the right of each plot outline. No formal statistical analysis on the spatial distribution of the weevil attack severity was attempted for this experiment designed as family-based plots where each plot contained 25 individuals of the same family.
where Y is a matrix of phenotypes (attack, egg punctures, height at ages 4 yrs and 5 208 yrs), β is a vector of fixed effects including intercept and replication effect, a is the 209 vector of random additive genetic effects following var(a)∼N(0,G1), where G1 is 210 additive genetic variance-covariance structure following trait, σ GxGy is additive genetic covariance between x th and y th trait, is Kronecker 213 product and A is average numerator relationship matrix [43], p is a vector of random 214 plot effects following var(p)∼N(0,G2), where G2 is plot variance-covariance structure Px is plot variance of x th trait and I is 216 the identity matrix, e is a vector of random residual effects following var(e)∼N(0,R), 217 where R is residual variance-covariance structure following σ ExEy is residual covariance between x th and y th trait, is Kronecker product, X and 220 Z are incidence matrices assigning effects in vectors β, a and p to phenotypes of matrix 221 Y . This model was performed for each cross type separately. The model convergence 222 was investigated by the Gelman-Rubin method [44] through merging 5 MCMC runs 223 using 120,000 runs with a burn-in period of 20,000 and thinning of 10 implemented in R 224 package "coda". The posterior estimates of genetic and environmental correlations 225 between traits were estimated as follows: where σ T1T2 is the covariance between first and second trait, σ 2 T1 and σ 2 T2 are the 227 variances for the first and the second trait, respectively. The parameters were extracted 228 Details are taken from [35], and the entire trial's plot outline on a family basis is provided in [38]; since spruce is monoecious, the same tree may serve as the mother or the father in the controlled crosses, as evidenced in the design; R or S indicate the resistance status of the parents with 'S' meaning 'susceptible' and 'R' meaning 'resistant' and the resulting cross types RxR, RxS, SxS are also indicated in the table. The elevation in meters above sea level (a.s.l.) is given in brackets next to the maternal parents' identifier where known.
from the G1 structure for genetic and the R structure for environmental correlations. package "MASS" as follows: where j is the level of an ordered category (attack severity or egg puncture occurrence) 237 and i is the level of the independent variable (elevation at origin or cross type).

238
Elevation was directly used to order the origin of the parents, while the cross type was 239 recoded into 0 for SxS, 1 for RxS and 2 for RxR. The proportion of total variance 240 explained by the model was estimated in R package "pscl" as pseudo R 2 by comparing 241 the log-likelihood for the fitted model against the log-likelihood of a null model with no 242 predictors implemented. Similarly, the linear regression was performed for continuous 243 variables (HT4; HT5) using the "lm" function implemented in the R package "base" 244 [45]. The proportion of total variance explained by the model was reported as R 2 .

246
Previous resistance rankings of parental trees indicated that the weevil attack severity 247 in Interior spruce was lowest for the RxR cross type while moderate for the RxS type showed the statistical significance of cross type on both attack severity and egg puncture 254 occurrence. Additionally, cross type explained 15% and 12% of the total variance in A00 255 and E00, respectively ( Figure 2B). As expected, the predicted effects of cross type for while a decrease in the likelihood of severe attack (A00=2) with increasing level of cross 258 type resistance (from SxS (recoded as 0) to RxR (recoded as 2)). On the other hand, 259 moderate levels of attack severity (classified as A00=1) did not show any pattern in 260 likelihood related to cross type ( Figure S1, Supplemental file 2). For egg puncture 261 occurrences (E00), and with increasing level of cross type resistance, the likelihood for 262 egg puncture level 0 increased while likelihood for all egg puncture levels equal or above 263 2 decreased. For egg puncture level 1, we did not find any pattern in likelihood related 264 to cross type ( Figure S2, Supplemental file 2). Heights at ages 4 yrs and 5 yrs followed a 265 similar pattern (Figure 2A, bottom left and right plots), where families in the RxR class 266 showed the highest median values and which gradually decreased toward SxS class. We 267 noticed that the dispersion of values was highest for RxR class, but lowest for SxS class 268 among all three cross types, potentially reflecting the lowest sample size available for 269 SxS class (Figure 2A). Results from linear regression confirmed the effect of cross type 270 for both height growth years as statistically significant. Yet, the proportion of total 271 variance explained by cross type reached only from 4% (HT5) to 5% (HT4), which was 272 therefore much lower compared to the total variance explained for the resistance traits. 273 The increase in mean height was around 5 cm for HT4 and 6 cm for HT5, respectively, 274 with an increasing level of cross type resistance from SxS to RxR ( Figure 2B). Since Interior spruce genetics reflects the natural hybridization between Engelmann 276 and white spruces along an elevation gradient, we investigated whether the distribution 277 of attack severity in the population can be attributed to parental origins. We found that 278 the mother trees originating from elevations between 600 m and 950 m mostly  Figure 3A, upper row and right column). We tested the effect of parental origin on 283 attack severity and found both maternal and paternal effects as statistically significant. 284 Yet, the model explained only 5% of the total variance ( Figure 3B). When we plotted 285 the predicted effects of parental origin from ordinal logistic regression, we found that 286 the increase in elevation decreased the likelihood of no attack (A00=0) while increased 287 the likelihood of severe attack (A00=2). No pattern was observed for moderate attack 288 severity (A00=1). These patterns were consistent for both parents ( Figure S3 and S4, 289 Supplemental file 2).

290
For the distribution of egg puncture occurrences in the population according to 291 parental origin ( Figure 3A, lower row), we found a less clear switch in the pattern above 292 950 m with regards to an impact of parent's origin. This could be explained by a low 293 number of egg punctures that might have already caused damage to the terminal shoots 294 of the offspring, likely due to a lack of resistance mechanisms passed on from parents 295 coming from high elevation ( Figure 3A, lower row). The fact that there was only a 296 small proportion of trees with parental origin at 950 m to 1200 m elevation for which 297 egg punctures did not occur, would exactly reflect such pattern. We also noticed for 298 (especially paternal) parents sourced from 1,220 m elevation a drastic drop in weevil 299 attacks ( Figure 3A). Again, these patterns were tested via ordinal logistic regression 300 and found to be statistically significant. The model explained 4% of the total variance 301 ( Figure 3B). When the predicted effects of parental origin were plotted for each level of 302 E00, we found that the increase in elevation of parental origin decreased the likelihood 303 of no egg puncture occurrence (E00=0), while increased the likelihood at all other levels 304 of E00 (that is E00=1 to 5), although the changes were only marginal. Again, such 305 patterns were consistent for both parents ( Figure S5 and S6, Supplemental file 2). A) The distribution of attack severity (A00, upper row) and egg punctures occurrences (E00, lower row) in the Interior spruce progenies (42 families) according to elevation at mother origin (left column) or at father origin (right column) is shown. These 42 controlled crosses were generated by mating resistant with resistant, resistant with susceptible, and susceptible with susceptible parents. The ranking in terms of weevil resistance of the individual Interior spruce parents is known and was done previously [18]. Attack severity increases from 0 to class 2. Egg puncture occurrence increases from 0 to class 5. The evaluation of attack severity and of egg puncture occurrence was done following artificial weevil infestation of the spruce resistance trial. B) Statistical significance of mother/father origin on attack severity and egg puncture occurrence is obtained from ordinal logistic regression. S.E. represents the standard error and R 2 the proportion of total variance explained by the model as pseudo R 2 .
We assessed heritability, genetic, environmental, and phenotypic correlations across 307 and within three distinct cross types (RxR; RxS; SxS) involving nearly 3k genotypes.

308
Results are summarized in Table 2. The narrow-sense heritability estimated across the 309 entire population reached heritability from 0.14 to 0.15 in traits related to herbivory 310 (A00 and E00) and 0.87 related to productivity traits (HT4 and HT5), respectively.

311
When the population was split into cross types, the highest heritabilities were obtained 312 for resistant RxR and susceptible SxS classes ranging from 0.31 to 0.38 in herbivory 313 resistance traits and from 0.86 to 0.88 in growth traits, respectively. However, the SxS 314 class estimates showed larger standard deviations reflecting its smaller sample size (less 315 crosses were available for this particular class). The lowest heritability estimates were 316 obtained for the RxS class reaching only from 0.14 to 0.16 in herbivory resistance traits 317 and from 0.66 to 0.69 in growth traits, respectively.

318
The highest positive genetic correlations were observed between herbivory traits 319 A00-E00 ranging from 0.53 (RxS class) to 0.83 (RxR class) and between growth traits 320 HT4-HT5 ranging from 0.94 (RxS class) to 0.97 (RxR class). The genetic correlations 321 between herbivory resistance traits and growth traits ranged from moderately to

332
The phenotypic correlations were all positive ( correlations. An important outcome of our re-assessment of the white pine weevil trial 339 was that for the RxS cross type (Table 2), the sign flipped from moderately negative to 340 slightly positive for genetic versus phenotypic correlations between herbivory (A00 and 341 E00) and height traits. Of note is that the respective environmental correlations 342 between herbivory (A00 and E00) and height traits in the same cross type (RxS) were 343 strong and positive, whereas they were otherwise either moderate and negative (RxR) 344 or insignificant (SxS). The moderate narrow-sense heritability for resistance against the white pine (or spruce 350 shoot) weevil and the relatively high narrow-sense heritability for tree height provides a 351 potential for an effective response to selection for both traits. We stress here that the 352 parental trees deployed in this study were selected solely on the basis of their 353 resistance/susceptibility to white pine weevil attacks and regardless of their origin 354 [18,35], while our results indicated that susceptible parental trees were predominantly 355 originating from a higher elevation and vice versa. The relatively high narrow-sense 356 heritability might result from high genetic diversity captured within the sampled 357 parental trees originating from elevations 610 to 1220 m a.s.l. Since Interior spruce is a 358 natural hybrid between Picea glauca and P. engelmannii which occupy a transitional 359 elevation zone [14], both species with different growing patterns might contribute to 360 the relatively high narrow-sense heritability. While P. engelmannii occupies subalpine 361 environments requiring adaptation for short growing windows and extended winters 362 with deep snow cover, P. glauca occupies a boreal environment that competes for light 363 naturally, which selects for relatively fast-growing genotypes. Individuals within the 364 hybrid zone were found to show advantageous features passed on from parental species, 365 such as autumn cold tolerance (P. engelmannii ) and fast growth (P. glauca) [14]. This 366 trend was also observed in this study, where we confirmed that parents of the RxR class 367 that originated from lower elevations, and thus, trees putatively of high P. glauca 368 ancestral proportion were tallest ( Figure 2). Our hypothesis is based on the observed 369 patterns, while further verification of the population's actual introgression patterns 370 through genomic markers is needed. We also note that this observation is based solely 371 on pre-attack growth rates. Rossi and Bousquet [46] found that under controlled 372 greenhouse conditions, northern provenances of black spruce showed earlier, and faster 373 bud break compared to southern origins. This result can be seen also as a proxy for the 374 differential local adaptation along an elevation gradient that involves P. glauca-P.

375
engelmannii as in the present study, with P. engelmannii exhibiting earlier bud burst 376 and budset [14]. Additionally, Wilkinson [47] found a negative genetic correlation 377 between bud break date and early tree height growth. Although De La Torre et al. [14] 378 stated that P. glauca does not show growth advantage until it reaches ten years of age, 379 our study found superiority already at 4 years of age in the RxR class, which 380 presumably has the highest P. glauca ancestry due to the elevation of origin [16] for the 381 parents involved in RxR crosses (this study). Yet, the differences between cross types 382 explained only ∼4 -5% of the total variance ( Figure 2B). While the potential of the 383 different genomic introgression patterns among P. glauca × P. engelmannii hybrids to 384 explain the differences in the relationship between growth and weevil resistance among 385 individuals has been proposed earlier [40], future research supported by genomic not adapted to this pest. These would thus represent naïve host trees, having never 396 been exposed to the pest. We also found an apparent increase in tree resistance with 397 1,220 m a.s.l. elevation, thus a pattern which deviated from the general trend.

398
Resistance status was tested in seven crosses, five RxR and two RxS families, which 399 confirmed the resistance segregation pattern expected for a resistant genotype [35].

400
Since Pinaceae, including spruces, are outcrossing species as evidenced by high levels of 401 heterozygosity in their genomes [48], and the parental genotypes of the F1 generations 402 tested in the trial had directly been sourced from wild stands, the observed resistance 403 pattern above 1,200 m a.s.l. seems valid. We argue that the spring phenology of the 404 hosts might be the determinant factor here. Alfaro et al. [26] found for Sitka spruce 405 that those families resistant to the white pine weevil showed earlier and quicker 406 budburst compared to susceptible families. Generally, individuals at colder sites exhibit 407 greater temperature sensitivity of spring phenology [49]. Therefore, we could assume 408 that families with parental contributions sourced from 1,220 m a.s.l. could show an 409 increased mismatch with the pest's phenology at the Kalamalka test site, which has a 410 continental climate, and could trigger an accelerated bud flush, since Engelmann spruce 411 and Engelmann spruce-like hybrids require less heat sum [14].

412
A Québec study performed on 45 Norway spruce genotypes originating from various 413 European provenances, however, found no relationship between budburst phenology, 414 traumatic response intensity and P. strobi performance, possibly due to the complete 415 lack of coevolution of this highly sensitive exotic host with the local weevil [50]. The 416 lack of correspondence between the weevil's biological performance and traumatic resin 417 duct formation in Norway spruce was also suggested in a related study [51]. However, 418 on a broader scale and under a warming climate, the narrowing of phenological 419 mismatches in tree hosts and local insect pests are of concern for the boreal forest [52]. 420 Thus, studies on the state of host-insect phenological synchronies and the forecasted 421 climate-induced phenological changes need to gain in relevance (ibidem). For example, 422 while previous defoliation by herbivores induces earlier budburst in the host as an 423 avoidance strategy [53], warming temperatures again reduce such phenological 424 mismatch due to quicker larval phenology [54].

425
For our study, we stress here again that the trees were mainly selected for 426 resistance/susceptibility to white pine weevil attacks, and that the observed pattern 427 might also follow clines in adaptive traits. To better understand the role of the 428 introgression level on the resistance/susceptibility status in Interior spruce, genomic 429 information is needed to trace the contribution of the ancestral species. Moreover, to 430 move forward, the implementation of recently available high-throughput assessment 431 methods for tree phenology monitoring that include remote sensing [55] will be crucial 432 to screen trials more holistically and efficiently along trees' ontogenetic development, 433 thereby offering the possibility to assess phenological changes in situ.

434
Different resistance mechanisms found between low elevation 435 (RxR) versus low with higher elevation (RxS) crosses 436 The positive (unfavourable) genetic correlation between tree height and the severity of 437 white pine weevil attack represents a challenge for simultaneous genetic improvement for 438 both traits and some trade-off should be implemented. This is in opposite to cases where 439 tree height is positively correlated with disease resistance such as white pine weevil in 440 Norway spruce [56] or Leptocybe gall wasp in Eucalyptus grandis [57]. The discrepancy 441 between results from our study compared to those performed in Norway spruce and 442 Eucalyptus grandis might be caused by the fact that our sample is at the early 443 ontogenetic stage and has not passed any previous disease outbreaks while the other 444 studies were performed at the advanced ontogenetic stage of the sample. Therefore, the 445 older plant material could pass previous disease outbreaks and the tree height is the 446 result of the level of resistance to pest attack rather than an independent trait. For 447 example, Dungey et al. [58], and later Klápště et al. [59], investigated Swiss needle 448 cast in Douglas-fir introduced in New Zealand. While the Californian provenances were 449 the most productive on South Island which is pest free, their productivity was severely 450 affected by Swiss needle cast on North Island where the Oregon provenances were the 451 most productive due to a higher level of tolerance to the pest.

452
When each cross type was analyzed separately, opposite patterns in genetic 453 correlations between growth and resistance traits were found. While RxR and SxS 454 classes showed strong or moderate positive correlations, the genetic correlations for the 455 RxS class were always moderately negative. Therefore, fast-growing trees in RxR and 456 SxS were mostly attacked. However, the results of the SxS class should be treated with 457 caution due to the small sample size. The only difference between these two cross types 458 was in the severity of the attack. While the RxR class showed attack severity at level 459 one which indicates that the tree was attacked but that the terminal growth was opposite relationship between growth and resistance traits. Therefore, fast-growing trees 463 were the least affected by weevil attack in the RxS class. However, the opposite pattern 464 was observed for the environmental compared to the genetic correlations, and these 465 environmental correlations were strong for RxS (with a positive sign) and moderate for 466 RxR (with a negative sign). We argue that a common environment was created by the 467 experimental design where each plot represented multiple individuals that belonged to 468 the same family. This might have generated such rather strong environmental 469 correlations, explained by a strong covariance between the permanent environmental 470 effects in the considered traits [60]. For this reason, phenotypic correlations cannot be 471 used in this case as a surrogate of genetic correlations. Where genetic correlations 472 should have the opposite sign, this is mainly assumed for attributes involved in life 473 history evaluation [61,62], and synchronous growth is one of the life-history traits in 474 forest trees [63]. Again, also in these cases, phenotypic correlations cannot substitute 475 genetic correlations [60]. While De La Torre et al. [14] already noticed better growth 476 performance and early frost tolerance based on spruce individuals' hybrid status, our 477 results also indicate that hybrids might indeed be genetically better equipped with an 478 optimized resource allocation between defence and growth. However, our hypothesis 479 should be verified by the re-evaluation of host trees' productivity at the later 480 ontogenetic stage to better understand the impact of weevil attack severity. Also, the 481 hypothesis about superiority of hybrids has to be further supported by future 482 assessment of the trees' hybrid index through genetic markers.

483
As one of the defense mechanisms against bark invading insects, conifer trees are 484 releasing oleoresin around the wound to flood deposited eggs, preventing them from 485 further development into larvae and from causing irreversible terminal shoot damage. A 486 significantly higher density of cortical resin canals was found in resistant compared to 487 susceptible genotypes in Sitka spruce [64] and in white spruce [30]. The same study 488 found a positive relationship between tree growth rate and resin canals density. Porth et 489 al. [40] also found that, overall, genetic correlations between bark histology and growth 490 were all positive. For the RxS class (1,403 genotypes tested), we found indeed those 491 individuals that were fast growing the least attacked. Some families may also display 492 higher tolerance to herbivory, where higher egg puncture occurrence does not 493 automatically result in severe damage outcome for the terminal leader. We found such 494 result, when we regrouped all individual families with strictly 40-80% top-kill outcome 495 [35], where, overall, fast growing trees were less severely attacked and for which genetic 496 correlation between A00 and E00 was extremely low (0.16). However, as stated by chemical defenses towards herbivory might be mutually exclusive, such that we would 499 not expect to observe tolerance in the RxR class.

500
Conclusions 501 Natural disturbances affecting forests are cumulative, highly interdependent and involve 502 biotic stressors, pests, pathogens individually or in association, and environmental 503 factors, drought, fire alike, all of which depend on climate, conditions, which in turn 504 affect phenology or life cycle of these organisms that is trees, pests, pathogens. Climate 505 change along with a warming climate, therefore, presents an increase in the stress on 506 forests due to the expansion of geographic ranges of pests and pathogens to now 507 suitable climates, their increased pathogenicity, enhanced population growth, and 508 potential host jumps [52]. Added to this are increasingly drought stressed and 509 otherwise weakened trees [65]. Therefore, multi-trait selection to build resilience in 510 conifer forests becomes relevant. Moreover, tree planting programs need to increasingly 511 consider the consequences of a warming climate on organisms' phenologies. The 512 resistance to pests and pathogens plays an essential role in enabling tree adaptation to 513 new ranges through tree population movement. This corroborates that resistance 514 breeding programs are continuously important for managing insect pests and 515 pathogen-related diseases. For example, stable positive genetic correlations between 516 height growth and resistance to a local pest would be necessary for an effective 517 multi-trait tree improvement program. Our study on the widely outplanted Interior 518 spruce concluded that selection of fast-growing hybrid genotypes might be the solution 519 for concurrent improvement in resistance against P. strobi attacks, whilst maintaining 520 tree productivity. We hypothesize that the RxS hybrid crosses might have inherited 521 high growth rates and dense resin canals from P. glauca but also some other features 522 involved in resistance (or avoidance) mechanisms from P. engelmannii preventing severe 523 attack of fast-growing genotypes. Such a feature might be a faster rate of bud flush.

524
This is the first study that clearly highlights that hybrid status might be a good 525 predictor for weevil attacks and that also delineates the boundaries of successful spruce 526 population movement. These results that RxS hybrids might be genetically better 527 equipped with an optimized resource allocation between defence and growth has 528 important implications considering the future management of Interior spruce breeding 529 programs and subsequent spruce plantation programs, which need to ensure that 530 adaptation can be maintained across the environment of deployment. Further work is 531 needed to employ genetic markers in efficient and early tree selection.