The experiment was conducted at the Koffler Scientific Reserve of the University of Toronto located near King City, Ontario (44°050′N, 79°483′W). A Temperature Free-Air-Controlled Enhancement (T-FACE) system was set up according to Kimball et al. (2008), consisting of 10 experimental plots, each with a diameter of 3 m. Ambient canopy temperature (AT) was recorded using infrared sensors (Model IRT-P5, Apogee Instruments, Logan, UT, USA) in five unheated control plots. For the elevated temperature (ET) treatment, five plots were arranged with six 1000 W infrared heaters (Mor Electric Heating Association, Comstock Park, MI, USA) per plot in a hexagonal array, where leaf temperature was raised by +1.5°C during the day and +3°C during the night, according to Kimball et al. (2008). Ambient air and canopy temperatures were recorded using a CR1000 datalogger (Campbell Scientific Inc., Edmonton, AB, Canada). Precipitation data were obtained from the Buttonville Airport weather station in Newmarket, ON (Environment Canada, 2014), located 25 km from the field site.

Plots were excavated 30 cm deep, filled with a mixture composed of one-third peat, one-third sand and one-third local soil, and tilled prior to planting. Three-year-old (3 + 0) bare-rooted Pinus strobus seedlings were obtained from a local seed orchard (seed zone 37, Somerville Nurseries, Everett, ON, Canada). In early May 2012, 90 seedlings were planted per plot. Gas exchange and fluorescence measurements commenced in mid-August 2012 after seedlings had established, and continued until December 2012. During 2013, measurements were expanded to assess water potential, soil moisture, and freezing tolerance. Measurements in 2013 were taken monthly from August 2013 until November, with final measurements taken in January 2014 (Figure 1). At each time point, three seedlings were randomly selected for measurement from each of five replicate plots per treatment. Soil moisture was measured using a HydroSense ™ soil water content sensor (Campbell Scientific Inc., Edmonton, AB, Canada). Soil moisture, measured as percent volumetric water content, was assessed at a depth of 15, 10 cm from the base of each measured seedling, three times per seedling. Deep frozen soil and ice packs prevented measurements of soil moisture in January 2014. Air humidity and temperature sensors (Hoskin Scientific Limited, Burlington, ON, Canada) were installed in May 2014, in order to assess differences in vapor pressure deficit (VPD), or the difference between actual and saturated air moisture, between heated and unheated plots during July 2014.

Mature current-year needles were collected from measured trees immediately following measurements, flash-frozen in liquid nitrogen, and stored at −80°C until analysis.

Gas exchange and chlorophyll fluorescence measurements were performed simultaneously using a portable photosynthesis system (LI-6400 XT; Li-Cor Biosciences, Lincoln, NE, USA) with attached leaf chamber fluorometer (6400-40). Topmost, south-facing needles of the primary shoot were arranged in a flat single-needle layer and placed into the cuvette. The cuvette was set to maintain a level of 400 ppm CO₂ and ambient temperature, which was selected based on the predicted daily average (Table 1).

Dark-adapted minimum PSII fluorescence (F_o), and dark-adapted maximum PSII fluorescence (F_m) were determined after 40 min of dark adaptation. Subsequently, plants were exposed to 1200 μmol quanta m⁻² s⁻¹ for 7–12 min to obtain measurements of steady-state photosynthesis; this light intensity represents one that is typically observed in boreal environments on clear and sunny days, even during early winter or early spring (Ensminger et al., 2004). Measured parameters included photosynthetic CO₂ assimilation (A), stomatal conductance (g_s), evapotranspiration (E), light-adapted minimum PSII fluorescence (F′_o), light-adapted maximum fluorescence (F′_m), and transient fluorescence (F_t), which were used to calculate gas exchange and fluorescence parameters (Table 2).

The seasonal depression of Fm due to low temperature does not allow for recovery of the maximum fluorescence signal in the dark, and thus limits its use for the calculation of the fluorescence parameter NPQ (Demmig-Adams and Adams, 2006). A good estimation of NPQ requires a dark-adapted control value of F_m that is measured when the photosynthetic apparatus is in a fully relaxed state. During winter, when F_m is depressed and does not relax rapidly in the dark, NPQ will be underestimated (Demmig-Adams et al., 2012). The non-photochemical quenching parameter, NPQ, was therefore calculated as shown in Table 2. The fully recovered maximum fluorescence (F_mrec) was estimated as F_o * 5, according to Schreiber et al. (1995) and Ensminger et al. (2004). This estimation is based on two assumptions: firstly, the ratio of fully recovered F_m/F_o is equal to 5, which has been demonstrated in multiple plant species (Björkman and Demmig, 1987), including conifers (Adams and Demmig-Adams, 1994); and secondly, unlike F_m, F_o shows little seasonal variation (Ottander et al., 1995). However, this approach might occasionally underestimate NPQ when F_o is strongly decreased.

Each measurement took approximately 15 min and were taken from 2 h after dawn until 2 h prior to sunset. Measurement campaigns occurred over 2–3 consecutive days. Measurement order was randomized at individual, plot and treatment levels during each campaign in order to minimize confounding diurnal or daily effects.

All measurements were performed on attached needles. Following measurement, needles in the cuvette were harvested to estimate the light-exposed needle surface area using a scanner and the WinSeedle software package (Regent Instruments Inc., Québec, QC, Canada).

Water potential measurements were performed from August to November, 2013. Pre-dawn and midday (noon) water potential (Ψ_w) were assessed on individual current-year needles using a Model 1505D Pressure Chamber Instrument (PMS Instrument Company, Albany, OR, USA). During each campaign, measurements were taken from three needles per seedling on three seedlings per plot, five plots per treatment. During November and January water potential was not assessed because the system did not operate at sub-freezing temperatures.

50–60 mg homogenized frozen needle tissue was extracted in 2 mL methanol buffered with 2% 0.5 M ammonium acetate according to Junker et al. unpublished. Samples were filtered using a 0.45 μm nylon filter prior to HPLC analysis. Chlorophylls and carotenoids were separated on a reverse-phase C30 column (YMC Carotenoid; Chromatographic Specialties Inc., Brockville, ON, Canada). Pigment extracts were analyzed with an Infinity 1200 series high performance liquid chromatography (HPLC) system equipped with a UV-diode array detector (Agilent Technologies, Santa Clara, CA, USA). De-epoxidation state (DEPS) was calculated as (0.5A+Z)/(V+A+Z) where V is violaxanthin, A is antheraxanthin, and Z is zeaxanthin. Total chlorophylls and α-tocopherol were expressed on a per freshweight basis, as the water content of white pine needles fluctuates less than 10% year-round (Verhoeven et al., 2009).

30-40 mg homogenized and lyophilized needle tissue from samples collected in August, October and December of 2012 were extracted in methanol:chloroform:water (12:5:3) according to Park et al. (2009), with the addition of 250 μg galactitol as an internal standard. 2 mL of the soluble sugar extract was vacuum centrifuged and resuspended in 1 mL of nanopure water. The resuspended extract was filtered using a 0.45 μm nylon filter and analyzed using a DX-600 anion-exchange HPLC (Dionex, Sunnyvale, CA, USA) equipped with a Hi-Plex Ca column (Agilent Technologies, Santa Clara, CA, USA) and electrochemical pulse amperometric detector (EC-PAD). Sucrose, fructose, glucose and pinitol were eluted with water at a flow rate of 0.170 mL/min with a column temperature of 70°C. Post-column detection was performed using NaOH at a rate of 100 mM/min. Raffinose was eluted using a Carbo-Pac PA1 column (Dionex, Sunnyvale, CA, USA) with 150 mM NaOH (isocratic) at a flow rate of 1 mL/min with post-column detection using NaOH at a rate of 100 mM/min.

Starch was determined from the residual tissue pellet from the soluble sugar extraction. The pellet was dried overnight at 55°C. 25–50 mg of the dried pellet were resuspended in 5 mL of 4% H₂SO₄, vortexed and autoclaved for 3.5 min. After cooling to room temperature, the extract was spun at 500 rpm for 5 min and the supernatant collected. The supernatant was filtered using a 0.45 μm nylon filter and analyzed using a DX-600 anion-exchange HPLC (Dionex, Sunnyvale, CA, USA) equipped with a Carbo-Pac PA1 column (Dionex, Sunnyvale, CA, USA) and EC-PAD. Glucose was eluted with water at a flow rate of 1 mL/min with a column temperature of 30°C. Post-column detection was performed using NaOH at a rate of 100 mM/min.

Chlorophyll fluorescence was used to assess freezing tolerance in August, September, October, November of 2013 and January 2014, using a modified protocol based on Sutinen et al. (1992). Current-year shoots were dark-adapted for 40 min. Each shoot was excised and F_v/F_m was measured. The shoots were then individually wrapped in moist paper towel and aluminum foil and sealed in a plastic bag prior to transport on ice back to the laboratory.

Shoots were exposed to freezing temperatures using a Thermotron SM-16-8200 environmental test chamber (Thermotron Industries, Holland, MI, USA). The maximum cooling rate was 2.5°C h⁻¹, with the 0 to −1°C interval achieved over 1 h. Since freezing resistance varies over the course of the year, preliminary freezing tests were performed throughout the year to identify a range of freezing temperatures suitable to induce freezing damage in white pine seedlings. Target freezing temperatures were then adjusted during each month of the experiment in order to account for the expected change in freezing tolerance, with the aim of selecting a range of freezing temperatures that bracketed the temperature at which 50% of the seedlings were damaged by freezing (LT₅₀). One shoot per tree per freezing temperature was held at the desired temperature for 6–8 h and subsequently thawed in a stepwise manner to room temperature: shoots exposed to ≤ −30°C were kept at −20°C for 24 h, transferred to 4°C for 24 h, and then transferred to room temperature for 24 h recovery. Shoots exposed to ≥ −20°C were transferred directly to 4°C for 24 h and then to room temperature for recovery (Sutinen et al., 1992). Following the 24 h recovery period, shoots were unwrapped and exposed to 1 h light exposure at 800 μmol quanta m⁻² s⁻¹ in order to stimulate PSII, then dark-adapted for 40 min (Burr et al., 2001). F_v/F_m was then assessed. Since we used chlorophyll fluorescence to evaluate freezing injury at PSII, we defined LT₅₀ as the temperature required to reduce maximum F_v/F_m by 50%. Maximum F_v/F_m was assessed by subjecting non-frozen shoots to the same protocol of 24 h recovery period, 1 h of light exposure, 40 min of dark adaptation and measurement. LT₅₀ values were calculated by fitting F_v/F_m values measured from freezing-recovered shoots using a modified Richards curve model (Fircks and Verwijst, 1993):

where K represents the upper asymptote, or pre-freezing F_v/F_m; B represents the maximum slope at LT₅₀ and M represents LT₅₀. Data was tested for normality using the D'Agostino-Pearson omnibus normality test. The curve for each treatment (elevated vs. ambient temperature) was fitted using the least squares method. LT₅₀ values were compared between treatments using an extra sum-of-squares F test with a P-value cutoff of 0.05. Analysis was performed using Graphpad Prism v6.04 (Graphpad Software, Inc., La Jolla, CA, USA).

Two-Way ANCOVA was used to assess the effect of the elevated temperature treatment and time on gas exchange, fluorescence and photosynthetic pigments, while accounting for the effect of seasonal variation introduced by photoperiod and daily temperature. The ANCOVA model used treatment and day of year as categorical fixed factors, photoperiod and minimum daily temperature as continuous numeric covariates, and plot and year as random factors, using the lme4 package in R v3.1.1 (http://www.r-project.org/). Multiple comparisons were used to contrast treatment within each time point, and were performed using the multcomp package in R v3.1.1. P-values for multiple comparisons were adjusted using Bonferroni correction.

Starch and soluble sugars were analyzed using Two-Way ANOVA to identify treatment, time and interaction effects. Tukey's HSD post-hoc test was used to identify significantly different groups. The statistical analyses for sugars were performed using Graphpad Prism v6.04.1.

Treatment responses of A, F_v/F_m, and NPQ_S from both years were pooled, independently plotted against minimum daily temperature and photoperiod, and fitted using the least squares method with a 4-parametric sigmoidal curve function:

where K represents the maximal parameter value; A represents the minimal parameter value; B represents the maximum slope and M represents the midpoint of the curve at which estimated values represent 50% of the maximum value of the parameter. R² and 95% confidence intervals were calculated. Midpoints were compared between treatments using a sum-of-squares F test. Modeling and analyses of the sigmoid curves were performed using Graphpad Prism v6.04.1.

The field site experienced higher amounts of precipitation during the growing season and lower amounts during winter (Figure 1A). 2012 was characterized by a warm early autumn, with daily maximum temperatures remaining above 20°C until the first week of October (Figure 1B). In contrast, during 2013, daily maximum temperatures began to decline below 20°C by the first week of September. Daily mean temperatures remained above 0°C until November in both years. The first night frost was recorded on October 8 in 2012 and on October 27 in 2013. During October 2012, the temperature dropped rapidly until mid-November and remained between a daily minimum of −5°C and a daily maximum of 10°C until mid-December. In contrast, nighttime temperatures during October and early November 2013 were mild, with minimum temperatures only reaching −2°C and daily maximums above 10°C. Minimum temperatures did not reach −20°C during the winter of 2012 until January 1, while minimum temperature reached −20°C on December 12 in 2013 (Figure 1B).

The variation in weather conditions affected temperature and precipitation during measurement campaigns. Measurements taken during August 2012 occurred after several rainy days, whereas measurements taken in August 2013 were taken after 10 days without rainfall (Figure 1A), resulting in decreased soil water content (Figure 4A). In 2012, we recorded a daily mean temperatures of 22°C during our measurements in August, 15°C in September and 15°C in October. In contrast, during 2013 we recorded daily mean temperatures of 20°C during our measurements in August, 7°C in September and 4°C in October (Figure 1B).

Photosynthetic carbon assimilation (A) remained unchanged from August to the beginning of October, was downregulated during October and November, and eventually ceased in December and January (Figure 2A). This trend was also observed for stomatal conductance (g_s), intrinsic water use efficiency (IWUE) and evapotranspiration (E) (Figures 2B–D). However, we also observed differences between years, e.g. during August and September 2013, when we measured lower rates of A, g_s, and IWUE (Figures 2A–C) compared to 2012. In 2012, photosynthetic gas exchange was fully downregulated by mid-November, while photosynthetic activity was still detectable in November 2013 (Figures 2A–C). Treatment had a significant effect on g_s, IWUE, and E; the interaction of treatment and time significantly affected g_s, IWUE and E (Table 3).

From August to early October, seedlings in heated plots that experienced elevated temperature (ET) exhibited lower A in comparison to seedlings in unheated control treatment (Control) that were exposed to ambient temperature (Figure 2A). However, the timing of the autumn downregulation of photosynthesis was not affected by the elevated temperature treatment, as A began to decrease by the end of October during both years, irrespective of treatment (Figure 2A). Between August and October of both years, g_s was decreased by 20–30% in ET seedlings compared to Control seedlings. Control seedlings maintained g_s between 0.10 and 0.15 mol H₂O m⁻² s⁻¹, while the elevated temperature treatment exhibited values between 0.06 and 0.12 mol H₂O m⁻² s⁻¹ (Figure 2B). g_s was significantly reduced in ET seedlings in heated plots in October 2012 (P = 0.041, Figure 2B). In 2012, IWUE was increased by about 15–20% in seedlings in the heated plots compared to Control seedlings, and was significantly enhanced in October (P = 0.036, Figure 2C). However, in 2013, IWUE was not significantly affected in seedlings in the heated plots (Figure 2C). E was reduced by 20–30% in ET seedlings in heated plots compared to Control seedlings during both years (Figure 2D), particularly during September 2012 (P = 0.009), October 2012 (P = 0.048), and September 2013 (P = 0.003).

The maximum quantum efficiency of PSII (F_v/F_m) was approximately 0.75–0.80 from August to early October, and continuously decreased from late October through January (Figure 3A). The effective quantum yield of PSII (Φ_PSII) was downregulated during the autumn transition and reached minimum values toward the end of November (Figure 3B). Non-photochemical quenching (NPQ) was high during both years; sustained NPQ (NPQ_S) began to develop during late October and comprised nearly 100% of nonphotochemical processes by January (Figure 3C). Excitation pressure (1-qP) increased during October and reached maximum levels in November before relaxing again in December (Figure 3D). In contrast to the substantial interannual variation observed in photosynthetic gas exchange, we did not observe interannual variations in most fluorescence parameters (Figure 3).

Φ_PSII was significantly lower in August 2013 in the elevated temperature treatment (P = 0.015, Figure 3B). NPQ was significantly higher in August (P = 0.014) and September 2013 (P < 0.001) in the elevated temperature treatment, but was not significantly different during 2012; NPQ_S was not significantly different between treatments during either year, although ET seedlings in the heated plots exhibited decreased NPQ_S during November and December (Figure 3C). 1-qP relaxed considerably from November to January under ambient temperature conditions, but did not when exposed to elevated temperature (P = 0.002, Figure 3D). Treatment had a significant effect on NPQ, whereas time had a significant effect on NPQ_S, F_v/F_m, and 1-qP (Table 3).

In 2013, soil moisture was lowest during August (Figure 4A). There was a consistent but non-significant reduction in soil moisture in the heated plots during the growing season (Figure 4A). Leaf water potential (Ψ_w) was generally high, with values consistently higher than −0.2 MPa (Figures 4B,C). There was no significant difference in Ψ_w during pre-dawn or midday (Figures 4B,C).

The extent of vapor pressure deficit (VPD) imposed by our heating treatment was assessed from July 18 to August 12, 2014 (Figure S1). The difference in VPD between heated and control plots varied depending on air temperature (Figures S1A,B). During the night, an increment of +3°C induced a 20% increase in VPD (0.351 ± 0.089 kPa) in the heated plots, while during the day an increment of +1.5°C induced a 6% increase in VPD (0.179 ± 0.089 kPa) in the heated plots (Figure S1).

Total chlorophylls, measured on a fresh-weight basis, increased from August to September, decreased from October to November, and remained stable in December and January (Figure 5A). Chlorophyll a/b decreased, albeit not significantly, from August to December (Figure 5B), whereas total carotenoids increased from October to November and stabilized in December (Figure 5C). β-carotene showed large variations during the autumn and between the treatments (Figure 5D). β-carotene levels increased over October and November, while α-carotene decreased over the same period (Figure 5D). Inter-annual variations were observed in chlorophyll and carotenoid pools, with chlorophyll a/b decreasing earlier in 2013 than in 2012 (Figure 5B), and slightly higher carotenoid levels during August 2013 compared to August 2012 (Figure 5C), though these differences were not significant.

Photoprotective pigments and metabolites also showed distinct changes during the autumn. Lutein (Figure 6A), neoxanthin (Figure 6B), and xanthophyll cycle pigments (Figure 6C) accumulated through October and November to maximal levels in December. The de-epoxidation status of the xanthophyll-cycle pigments (DEPS) (Figure 6D) transiently increased from October through January. α-tocopherol increased from August to November, followed by stabilization in December (Figure 6E).

Accumulation of lutein (Figure 6A), neoxanthin (Figure 6B), total xanthophylls (Figure 6C) and DEPS (Figure 6D) varied between years. In August 2012, seedlings exhibited low levels of lutein (Figure 6A) and total xanthophylls (Figure 6C) compared to August 2013. DEPS was increased in August 2012 compared to August 2013 (Figure 6D). Increases in lutein (Figure 6A), neoxanthin (Figure 6B), and xanthophylls (Figure 6C) occurred during August and September 2013 which were not observed in 2012. Treatment did not have a significant effect on any of the photosynthetic pigments, whereas time had a significant effect on chlorophyll a/b, α-carotene, β-carotene, DEPS, and neoxanthin (Table 3).

In response to the decrease in daily minimum temperature from 10 to −2°C during autumn and winter, we observed a transient decrease in A, F_v/F_m and NPQ_S (Figures 7A,C,E, Table 4). A also showed a response to decreasing photoperiod over a range of 11–9 h (Figure 7B, Table 4). In contrast, rapid downregulation of F_v/F_m and rapid induction of sustained NPQ were observed when photoperiod reached a threshold value of approximately 9.6 h (Figures 7D,F, Table 4). Elevated temperature did not significantly affect the response of assimilation to minimum temperature, but did shift the response of F_v/F_m and NPQ_S (Table 4).

In both treatments, leaf starch levels remained unchanged from August through December (Figure 8A), while the amount of total soluble sugars decreased slightly from August to October and doubled in December (Figure 8B). Raffinose was absent in August, present in minute quantities in October, and present in large quantities in December (Figure 8C). Sucrose levels were constant during August and October but increased in December (Figure 8D). Fructose levels remained constant from August through December (Figure 8E), while glucose levels decreased by more than 50% from August to October, but tripled from October to December (Figure 8F). Pinitol mirrored the glucose levels, and was reduced by 50% from August to October, but doubled in December (Figure 8G). Seedlings in the heated plots exhibited significantly higher amounts of leaf starch (P = 0.026, Table S1); however, this did not significantly affect the accumulation of soluble sugars (Table S1).

In 2013, seedlings exhibited tolerance to freezing exposure of −10°C during August (Figure 9A), −16°C during September (Figure 9B), −30°C in October (Figure 9C) and were fully cold hardy below −60°C by November (Figure 9D). The cold hardiness of ET seedlings from heated plots did not differ from that of Control seedlings from unheated plots during August (Figure 9A) or September (Figure 9B). In October, seedlings from heated plots exhibited significantly greater freezing tolerance in comparison with freezing tolerance of seedlings from unheated plots (P = 0.027, Figure 9C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.