Evidence of prevalent heat stress in Yukon River Chinook salmon

Migrating adult Pacific salmon (Oncorhynchus spp.) are sensitive to warm water (>18 °C), with a range of consequences from decreased spawning success to early mortality. We examined the proportion of Yukon River Chinook salmon (O. tshawytscha) exhibiting evidence of heat stress to assess the potential that high temperatures contribute to freshwater adult mortality in a northern Pacific salmon population. Water temperatures greater than 18 °C have occurred almost annually in the Yukon River and correspond with low population abundance since the 1990s. Using gene transcription products and heat shock protein 70 biomarkers validated by field experiment, we identified heat stress in half of Chinook salmon examined (54%, n = 477) across three mainstem locations and three tributaries in 2016–2017. Biomarkers tracked wide variation in water temperature (14–23 °C) within a tributary. The proportion of salmon with heat stress differed between years at four of the six locations, with more prevalent heat stress in the warmer year. This work demonstrates that warming water temperatures are currently affecting northern populations of Pacific salmon.

Chadwick and McCormick 2017). 118 Here, we conducted an experimental temperature manipulation with a subset of wild 119 Yukon River Chinook salmon to validate heat stress biomarkers and subsequently estimated the A standard cDNA synthesis was performed on 2 μg of RNA template from each salmon. 206 Reaction conditions included 4 units reverse transcriptase (Omniscript, Qiagen, Valencia, CA), 1 207 μM random hexamers, 0.5 mM each dNTP, and 10 units RNase inhibitor, in RT buffer (Qiagen,208 Valencia, CA). Reactions were incubated for 60 min at 37 °C, followed by an enzyme 209 inactivation step of 5 minutes at 93 °C, and then stored at -20 °C until further analysis. 210 Briefly, 1 µl of cDNA was added to a mix containing 12.5 μl of QuantiTect Fast SYBR D r a f t 12 250 buffered saline plus 0.1% Triton X-100) for 1 h at room temperature, rinsed in PBST, and 251 exposed to primary polyclonal antibody specific for the inducible form of salmonid HSP70 252 (AS05061; Agrisera, Vannas, Sweden) at 1:25,000 dilution in PBST with 5% non-fat dry milk 253 for 1 h at room temperature. After rinsing in PBST, blots were exposed to goat anti-rabbit IgG 254 conjugated to horseradish peroxidase diluted 1:10 000 in PBST and 5% non-fat dry milk for 1 h 255 at room temperature. Blots were washed in PBST and incubated for 1 min in a 1:1 mixture of 256 enhanced chemiluminescent (ECL) solution A (396 µMcoumaric acid, 2.5 µM luminol, 100 mM 257 Tris, pH 8.5) and ECL B (0.018% H 2 O 2 , 100 mM Tris, pH 8.5), and then digitally imaged and 258 quantified (Syngene PXi, GeneTools, Frederick, MD, USA). All blots were normalized to the 259 internal standard consistent tissue preparation and are represented as a ratio to the mean standard 260 value that we refer to as HSP70 relative abundance.

261
Statistical Analyses of controlled experiment 262 Heat stress was inferred through differences between fish held in the control and elevated 263 temperature treatments that allow for high classification accuracy (correct classification > 75%) 264 for both the gene transcript levels and HSP70 protein abundance. Following separate statistical 265 analysis for the two biomarkers, heat stress classifications based on gene transcription and 266 HSP70 protein were considered jointly with indications of heat stress in both biomarkers 267 interpreted as more severe stress (Lund et al. 2002;Lewis et al. 2016).

268
A linear discriminate analysis (LDA) was used to identify the combination of muscle were not dependent on a single influential point (i.e., outlier).

283
Given the small experiment sample size, a sequential reduction in the number of genes in 284 the LDA was preformed to limit the number of genes used in the model and reduce overfitting.

285
The sequential reduction was based upon gene transcript data that was centered and scaled prior 286 to analysis so that coefficients reflected the influence of each gene in the model. The least 287 influential gene with the coefficient closest to zero (mean of LD1 and LD2 coefficients weighted 288 by the variation attributed to each LD) was dropped from the analysis sequentially until the 289 classification rate fell below the desired threshold of 75%. The model that used the fewest 290 number of genes and still maintained a classification rate of >75% was considered the preferred 291 model. HSP70 protein abundance was plotted and visually examined for the location of a 292 threshold that best separated fish among treatments. The mean HSP70 abundance was compared 293 among treatments using a one-way analysis of variance (ANOVA).    Two Rivers, AK (station number 15493000). Monthly mean water temperatures were compared 387 in each river between years using a Welch's t-test for unequal variances for qualitative 388 comparison to interannual differences in the heat stress proportions. All data and metadata 389 associated with this study is publicly available in von Biela and Donnelly (2020)  categorized as high heat stress fish (e.g., similar to the 21 °C treatment group) ( Figure 4A).

456
Water temperature was also related to the presence of elevated HSP70 protein abundance 457 (logistic regression, deviance = -55.7, df = 1, P < 0.001) ( Figure 4B). Each degree of increase in 458 temperature resulted in a 4.3-fold increase in the odds of elevated HSP70 (log-odds ratio = 1.46).

459
The three-day maximum water temperature associated with a 50% probability of elevated HSP70 fish, many attributes are set prior to river entry (e.g., body length), and not all watersheds have a 635 long, shared migration route like the Yukon River.

636
In addition to thermal stress, heat stress biomarkers can be induced by a variety of HSP70 protein values are just above the threshold for heat stress detection ( Figure 5).

653
The Gisasa     D r a f t Figure 3. Results from a linear discriminate analysis (A) used to reduce mRNA abundance of six genes (HSP70, HSP90, Gata3, IFNa, AHR, and SOD) to two linear discriminate axes (LD1 and LD2) and HSP70 protein abundance (B) in muscle tissue of Chinook salmon from individuals held in three different experimental temperatures: control (blue, river ambient, ~ 15 °C), 18 °C (orange), or 21 °C (red). Shapes indicate the model prediction for each individual with either the control (open circles), 18 °C (filled squares), or 21 °C (filled triangles) treatment group. The dashed line in B is the proposed threshold for elevated HSP70 protein that distinguishes heat stress and points in this panel are spread out (i.e., jittered) so that data points are more visible.
D r a f t Figure 4. The LD1 gene (A) and HSP70 protein abundance (B) plotted against the maximum three day water temperature for Chinook salmon captured at the East Fork Andreafsky River weir during spawning migrations in 2016 and 2017 (n = 86). LD1 gene is calculated based on gene transcript data for each individual fish and gene-specific model coefficients fit based on experiment results ('preferred model'). Symbol color and shape reflects the heat stress classifications based on predictions fit with experiment data.
Field-sampled individuals categorized as unstressed were similar to experimental control fish (blue circles) and individuals classified with heat stress were distinguished between heat stress similar to 18 °C experiment group (orange squares, LD1 genes only) or the 21 °C experiment group (red triangles).
203x254mm (300 x 300 DPI) D r a f t Figure 5. Boxplots of LD1 genes (A) and HSP70 protein abundance (B) for spawning Chinook salmon captures across the Yukon River watershed in either 2016 (navy) or 2017 (purple) in relation to results from an experimental temperature manipulation (green). Dashed lines reflect the threshold for identifying heat stress. Heat stress is indicated by values lower than the LD1 threshold and higher than the HSP70 protein threshold. In each boxplot, the horizontal line is the median, the upper and lower ends of the box are the first and third quartiles, and the whiskers extend to the highest and lowest values that are within the 1.5x inter-quartile range. Outliers are not shown.
233x119mm (96 x 96 DPI) D r a f t Figure 6. A map depicting the percent of Chinook salmon in each capture location (black circle) and year with evidence of heat stress. Paired stacked bar charts reflect the heat stress classifications from gene transcript and HSP70 protein for fish captured in summer 2016 (left) and 2017 (right). Fill color reflects the proportion of individuals in each of the four heat stress categories: no evidence of heat stress in either gene transcription or HSP70 protein, green; heat stress identified only in gene transcription, gold; heat stress identified in gene transcription and elevated HSP70 protein, red; and heat stress identified only by elevated HSP70 protein, purple). Numbers near each bar are the sum across the three categories that identified heat stress presence. The white star is the location of the temperature experiment near Pilot Station, AK, USA. An asterisk (*) prior to the capture location name denotes a significant difference in the heat stress proportion between capture years. Map created in ArcMap 10.7 (ESRI, Redlands, CA, USA) with selected rivers from the National Hydrography Dataset (U.S. Geological Survey 2015) and shorelines from Wessel and Smith (1996).