A hand-held Agilent Resolve spectrometer (Agilent Technologies, USA) equipped with 830 nm laser was used to obtain spectra from the potatoes. Raman spectra were collected from four tubers per clone (10 clones), per replication (4), and condition (2) (normal vs. heat stress) (2020, 2021). Each tuber was scanned a minimum of three times, resulting in 48 spectra per clone per greenhouse condition (normal vs. heat stress), which resulted in over nine hundred and sixty acquired spectra. The instrument collects three types of spectra – ‘surface’, ‘offset’ and ‘spatially-offset Raman spectra (SORS)’ (Matousek et al., 2005). In 2020, ‘offset’ spectra were obtained directly from the intact tubers (with skin on). We acquired spectra with good signal-to-noise ratios from tubers with thin skin. However, spectra with lower signal-to-noise ratios were obtained from tubers with heavy russet skin. To improve the signal-to-noise ratio in the acquired spectra, in 2021, a very small section (1cm²) of the skin of potato tubers was peeled (2 mm thickness) before taking ‘surface’ spectra. The spectral acquisition time for each scan was 1 s.

Analyses of variance for specific gravity, dry matter, reducing sugars, and protein traits were analyzed using the mixed model approach in JMP ^®16 (SAS Institute Inc, 2021). Clones and growing conditions were considered fixed effects, whereas replications were random. Mean comparisons were made based on Tukey’s HSD. The results ( Supplementary Table 2 ) were graphically represented using MS excel. Variances were calculated for each source of variation from ANOVA table using the sum of squares.

MATLAB equipped with PLS Toolbox (Eigenvector Inc., WA, USA) was used to perform partial least-squares discriminant analysis (PLS-DA) for all collected spectra. Four preprocessing steps were performed before analyzing the spectra for differences in the Normal and heat-stressed samples: (a) MSC Mean, (b) Smoothing (1^st polynomial order and 15 filter width), (c) 2^nd Derivative of the Spectra (3^rd Polynomial order and 51 filter width), (d) Normalization to the Area. Model was built with 70% of the acquired spectra which was then used to test 30% of the spectra to determine the true positive rate (TPR) for each model (normal growth vs. heat stress).

Analyses of variance (ANOVA) for the acquired Raman spectra were performed using MATLAB (MATLAB Co., USA). We focused on vibrational bands with the following peaks: a peak at 479 cm^-1 to assess starch (Morey et al., 2020), a peak at 1527 cm^-1 to inspect carotenoids (Adar, 2017), a peak at 1208 cm^-1 representing phenylpropanoids (Larsen and Barsberg, 2010), and a peak at 1660 cm^-1 for protein (Kurouski et al., 2015).

The highest specific gravity was observed in Atlantic under normal conditions in both years; this was expected since Atlantic is a chipping variety and high specific gravity (high starch content) is a required trait for this market group. Tubers produced under heat stress conditions had significantly lower specific gravities than those harvested from normal growing conditions ( Figures 1 , 2 ). The interaction of clone*condition was significant, indicating a differential response of some clones to the growing conditions ( Figures 1 , 2 ; Table 2 ). Potato tubers harvested from heat-stress greenhouse growing conditions had, on average, 1.2% and 1.7% lower specific gravity than those tubers produced under normal conditions in 2020 and 2021, respectively ( Table 3 ). The least reduction (0.9%) was observed in Russet Norkotah 278, whereas the largest reduction (1.9%) was observed in COTX09022-3RuRE/Y followed by Russet Burbank (1.6%) in 2020. In 2021, tubers of Yukon Gold were the least affected by specific gravity reduction (1.2%).

In contrast, tubers of Russet Burbank were the most affected by specific gravity reduction (2.7%) due to high-temperature stress. Conditions and clones alone explained more than one-third variation associated with specific gravity in both years ( Figure 3 ). The interaction between clone and condition explained from 2.3% to 4% variation in the specific gravity of potato tubers harvested in 2020 and 2021.

Similar to specific gravity, tubers produced under heat stress conditions had significantly lower tuber dry matter than those harvested from normal growing conditions. Clones showed differential responses for tuber dry matter under different temperature conditions ( Figures 1 , 2 ; Table 2 ). Tuber dry matter is highly correlated with specific gravity, and Atlantic had the highest tuber dry matter. Variation in tuber dry matter was associated with clones, conditions, and interaction between clones and conditions in decreasing order in both years ( Figure 3 ). Though a significant interaction was observed for tuber dry matter, the genetic makeup of the clones had a strong influence, as well as the growing conditions of the plants. The potato clones had an 8.1% (2020) and 14.6% (2021) reduction of tuber dry matter when grown under high-temperature conditions as compared to the normal growing condition ( Table 3 ). The largest reduction in tuber dry matter (16.3%) was observed in Russet Norkotah 296 followed by Russet Norkotah (15.6%) and Russet Burbank (14.3%) whereas tuber dry matter increased by 2.5% and 10.2% in Vanguard Russet and Reveille Russet, respectively in 2020. In 2021, the highest reduction of tuber dry matter was observed in Russet Burbank (24.2%), while the lowest reduction was observed in Russet Norkotah 296 (7.0%) together with Reveille Russet (7.0%).

Reduced sugars (glucose + fructose) in tubers were significantly higher under heat-stress growing conditions than under normal growing conditions. There was an increase of 67.7% (2020) and 143.0% (2021) in reducing sugars in tubers grown under heat stress compared with that of tubers grown under normal temperature conditions ( Table 3 ). Under normal conditions, the lowest reducing sugars were observed in Atlantic followed by Sierra Gold and Yukon Gold in 2020 and Atlantic, Sierra Gold, Yukon Gold and COTX09022-3RuRE/Y in 2021. As in the case of specific gravity and tuber dry matter, reducing sugars also exhibited significant clone-by-condition interaction ( Figures 1 , 2 ), however, the interaction effect (clone*condition) for reducing sugars was lower ( Figure 3 ). Variation in reducing sugars could thus be attributed mainly to the genetic makeup of the clone and the growing condition in which the potatoes were grown ( Figure 3 ).

Protein (% DW) in tubers was significantly increased under heat stress conditions ( Figures 1 , 2 ). Though there was a significant interaction between clones and growing conditions regarding protein in tubers, the effect size of the interaction (7% in 2020 and 2% in 2021) was too small to explain the variation. However, clone (31% in 2020 and 30% in 2021) and condition (19% in 2020 and 35% in 2021) explained most of the variation observed in protein percentage in the tubers from the experiment ( Figure 3 ). Under heat stress conditions, Reveille Russet had the highest protein content in both years. The lowest protein content was observed in Russet Burbank and Sierra Gold under normal conditions. The increase in protein (dry weight basis) was found to be 10.6% in 2020 and 20.2% in 2021 ( Table 3 ).

The averaged Raman spectra acquired from tubers of all potato clones grown under normal and heat-stressed conditions exhibit vibrational bands that could be assigned to starch, protein, carotenoids, cellulose and phenylpropanoids ( Figures 4 , 5 ; Table 4 ). There was a significant decrease in the intensity of most vibrational bands that originated from carbohydrates (477 cm^-1, 865 cm^-1, 940 cm^-1, 1085 cm^-1, 1126 cm^-1, 1261 cm^-1 and 1340 cm^-1) in Raman spectra acquired from tubers harvested from heat-stressed plants compared to the intensities of these bands in the Raman spectra collected from tubers of plants grown under normal conditions ( Figures 4 , 5 , Supplementary Figure 2 ). We also found changes in the intensities of bands that originated from proteins and phenylpropanoids in the spectra acquired from tubers of normal vs. heat-stressed plants. However, no significant changes were observed in the intensities of carotenoid bands ( Figure 6 ). These results suggested that the concentration of carotenoids did not change in potato tubers due to the high-temperature stress. Thus, we can conclude that concentrations of starch, proteins and phenylpropanoids decreased in tubers produced by plants exposed to high-temperature stress. It should be noted that when the yellow flesh clones were compared individually, the carotenoid peaks were significantly more intense in the spectra collected from tubers from normal conditions versus those from heat stress conditions in COTX09022-3RuRE/Y and Yukon Gold, but no significant change in the intensity of carotenoids in the spectra acquired from another fresh yellow flesh clone Sierra Gold™, a heat-tolerant early variety from the Texas A&M Program ( Figure 7 ).

Next, PLS-DA was used to investigate the prediction accuracy of heat stress in potato clones grown in 2020 and 2021 ( Table 5 ). We found that developed models could identify heat stress with accuracies of 63.8% in 2020 and 68.8% in 2021. At the same time, the absence of stress could be identified with accuracies of 73.4% in 2020 and 84.3% in 2021.

Having high specific gravity is very important in processing potatoes. In these experiments, Russet Burbank was used as a reference for processing French fries, and Atlantic as a reference for processing chippers. Processors are paid by the dry weight of the final product, and oil expenses are directly related to the amount of water replaced in tubers; thus, less water in tubers (higher specific gravity) is desired in processing potatoes. Sensory attributes for French fries and chippers are better in potatoes with high specific gravity (high dry matter, high starch). Heat stress has been reported to decrease the specific gravity of potatoes (Teixeira et al., 2015; Andrade et al., 2021; Fernandes Filho et al., 2021). The decrease in specific gravity is regarded as a loss to the processors. The decrease in specific gravity indicates that there is less dry matter accumulated in the potato tuber. A reduction in dry matter would result in reduced income. The effect of heat stress can be seen as a general decrease in specific gravity in all potato clones tested. The greatest decrease in specific gravity was observed in Russet Burbank, indicating that heat significantly reduced its processing quality and thus can be considered heat sensitive. Reveille Russet is a fresh market potato and has low specific gravity. The reduction of specific gravity in Reveille Russet due to heat stress was proportionally lower than that observed in Russet Burbank. The results of this experiment show that Atlantic when grown under normal (temperature) conditions, gives the highest specific gravity, but its specific gravity decreases when grown under high-temperature conditions indicating its sensitivity towards heat stress.

Tuber dry matter of potatoes is an important quality characteristic for processing that can also be obtained from specific gravity (Nzaramba et al., 2013). Several reports have also indicated that tuber dry matter is drastically reduced when potatoes are grown above 20-25°C (Levy and Veilleux, 2007; Aien et al., 2017; Busse et al., 2018; Kim and Lee, 2019; Obiero et al., 2019). As a result of low dry matter, the quality of the chips and French fries decreases. The chips/fries absorb more oil when potatoes have lower dry matter, as indicated by low specific gravity. Similar to our results with specific gravity, the tuber dry matter also showed a decrease in potatoes grown under high temperatures compared to normal (temperature) conditions. Under similar heat stress temperatures of 35/25°C day/night, it was reported that high temperature caused a reduction in assimilate partitioning (radio labeling of ¹⁴C) to tubers in four clones tested. This reduction in partitioning was evident with a significant decrease in tuber dry matter percentage grown under high temperatures compared to Normal (25/12°C day/night) (Gawronska et al., 1992). The reduction of dry matter concentration under high temperatures has also been associated with the synthesis of high levels of endogenous gibberellins, which reduce the partitioning of assimilates to the tubers and impede the synthesis of starch and tuber-specific proteins (Lovell and Booth, 1967).

Temperature-induced increase in soluble sugars was observed in potatoes grown under moderately elevated temperatures in Chile (Ávila-Valdés et al., 2020). Soil temperatures of 23°C and 29°C during tuber bulking have been reported to increase the reducing sugars more than two folds, even in the Low-temperature sweetening-resistant clone Premier Russet (Zommick et al., 2014). Heat stress of just 14 days has been reported to increase the reducing sugar-glucose in potatoes (Snowden, Lamoka, Megachip and Nicolet), especially in the basal ends of tubers when grown under a high temperature of 35/29°C day/night.

Heat stress increased the protein concentration (DW) in the potatoes based on wet chemistry ( Figures 1 , 2 ). When plants were grown at the high-temperature regime, a greater increase in protein concentration due to heat stress was found compared to the normal temperature regime. A decrease in individual tuber dry mass accompanied a large part of the increase found in protein concentration. Stone and Nicolas (1998) observed a similar case and reported that heat stress increased grain protein percentage even though the overall protein content per grain was reduced by heat.

Dry matter accumulation drastically decreased under heat stress. Since the average tuber dry mass decreased, it can be assumed that the negative effect of heat stress on starch biosynthesis led to a lower level or dilution of tuber protein. The tuber dry matter is mainly starch, and its greater reduction under heat stress would increase protein percent on a dry basis. On the other hand, the protein percentage of tubers, even if significantly higher at heated conditions than potatoes under normal temperature, is very small in comparison to potatoes’ starch. Thus, the observation that the protein content of potatoes under heat stress conditions is lower than that of normal conditions based on the Raman spectrum is contrary to our wet chemistry findings. Protein content in tubers based on wet chemistry was calculated from total nitrogen content. Tuber dry matter and other parameters decreased under heat stress conditions. Thus, it is likely that the absolute protein content per plot/clone did not change or even decrease despite the increase in percentage DW based on a tuber sample.

Clones varying in temperature tolerance showed differential responses with their root proteins. In a recent study, the sensitive clone showed a twofold increase in defense and detoxification-related proteins. In contrast, the tolerant clone showed more increase than decrease in proteins related to energy and carbohydrate metabolism (Boguszewska-Mańkowska et al., 2020). Stress-responsive proteins like HSP17, 6-CI, HSP101, and eEF1A are associated with microtubers under high-temperature stress. A higher level of eEF1A in the experiment was putatively marked as lowering the negative effects of heat in potato tuberization (Pantelić et al., 2018). It is plausible to think that the protein content of potatoes would increase at the expense of starch under stress. However, total soluble protein content (DW) was found not to be significantly different in potatoes grown under different temperature conditions (Ávila-Valdés et al., 2020). Also, the total protein content of strawberry plants decreased under heat stress conditions (Gulen and Eris, 2004).

RS determined the effect of heat stress on different cultivars of potatoes. In most potatoes, the carbohydrate levels decreased when plants were grown under heat stress. Phenylpropanoids were significantly reduced in tubers produced under heat-stress conditions, but there was no significant reduction in carotenoids. The strongest Raman signal for starch around 477-479 cm^-1 was used as a marker for the quantification of total cellular starch (Kizil et al., 2002; Ji et al., 2014; Morey et al., 2020). This main starch peak was also found in this experiment to be expressed at a significantly higher intensity in potatoes grown under normal conditions compared with the potatoes grown under higher temperatures. Raman spectra can reveal differences in the starch content of potatoes grown under different conditions, and Raman spectroscopy has the potential to estimate starch content in potatoes. In a previous study, Morey et al. (2020) verified that Raman spectroscopy can predict the starch content of potato samples based on the intensity of the 479 cm^-1 band. Quantification of amylopectin and amylose, as well as proteins, have been achieved in rice (Pezzotti et al., 2021), protein, and oil in soybean (Singh et al., 2019) through the use of Raman spectroscopy. Tuber carotenoids do not seem to change in response to heat stress. Similar to our results with Raman spectra, Fogelman et al. (2019) reported that carotenoid levels did not change with the heat treatment (33-35°C) of tubers for one week before harvest. Also, no significant carotenoid changes were observed in heat stress-treated potato plants compared to normal by Shah et al. (2020). Thus, several studies showed evidence to conclude carotenoids are not being changed with heat stress.

On the other hand, phenylpropanoids showed a significant reduction under heat stress compared to normal (temperature) growing conditions. Secondary metabolites are affected by high-temperature stress. Their lower concentration in potatoes in the heated conditions means that either phenylpropanoids were used to protect other biochemical compounds like proteins and allow normal functioning of the tuber, or their biosynthetic pathway was disrupted by excess temperature. In both scenarios, phenylpropanoids seem to be lower under heat stress. A more detailed investigation would be required to pin down the types of phenylpropanoids significantly affected by growing conditions. Although several factors contribute to a location’s effect on a crop metabolite profile, the temperature seemed to be one of the factors to lower the total phenylpropanoids measured in Magic Molly potatoes across different sites in Alaska, Texas, and Florida (Payyavula et al., 2012). One of the most abundant phenylpropanoids, chlorogenic acid (CGA), decreased from locations in Alaska to Texas and Florida. Anthocyanin was found to decrease with increasing temperature. A higher level of total phenolics in potatoes was reported at a location with lower temperatures (Reddivari et al., 2007).

Also, RS predicted the tubers’ growing condition based on the model built with PLSDA. Since fewer good scans were obtained in 2020, a comparison of Raman spectra of individual cultivars and construction of confusion tables were generated for 2021 datasets only ( Table 6 ). The percent accuracies in determining the presence of heat stress in several cultivars (Reveille Russet and Vanguard Russet) was low, indicating that these cultivars were not predicted as heat-sensitive cultivars (thus, declaring them as heat tolerant).

The difference in TPR for different cultivars could be inspected to see that some potato clones were more tolerant to heat stress than others. Similar TPR % for both growing conditions when considering a clone indicates that the model could distinguish between the treatments; however, a large difference in the TPR% between the conditions indicates that the spectrum of one condition overlapped with the other. The lower TPR % for clones like Reveille Russet and Vanguard Russet indicated that their spectra of potatoes grown under heated conditions overlapped with that of potatoes under normal conditions, indicating their chemical signatures are similar. This spectrum similarity under different conditions means that the potatoes of these cultivars behaved similarly to the conditions. Thus, we could assign them to be more heat tolerant than others of heat stress. Similarly, heat-susceptible clones like Atlantic and Russet Burbank could be differentiated with greater confidence for their growth condition. They exhibited different signature spectrums under different conditions when scanned with Raman.