*2.4. Phenotypic and Genomic Cluster Analysis*

Accessions were clustered based on values of the log−2 transformed response ratios (Figure 4A). Accessions were most strongly divided into different groups based on growth (PLA, leaf size), with Hs-0 being distinct from all other accessions. Another group of accessions, LDV-58, HSm, Gel-1, and Amel-1 could be defined as FL-sensitive in that these accessions showed the strongest reductions in growth and number of leaves, while also showing some reductions in ΦPSII and relatively large increases in NPQ (Figure 4A). When clustered based on their genomic differences, on the other hand, accessions showed an entirely different pattern (Figure 4B). For example, Col-0 was suggested to have the largest genomic distance from all other accessions, while phenotypic clustering suggested its response to FL to be close to that of, e.g., Fei-0 and MNF-Che-2 (Figure 4A). Another example for the incongruence between genomic and phenotypic clustering is a group of closely related accessions that all originated in the Czech Republic: HSm, Ta-0, DraIV6-16, ZdrI2-24 and Da(1)-12 (Figure 4B, Table 1). The phenotypic analysis, on the other hand, did not indicate a close link between these accessions (Figure 4A). Indeed, no significant correlation was found between log−2 transformed response ratios and corresponding genomic distances between all pairwise comparisons among the 36 accessions (36 × 35/2 pairs; Pearson correlation coefficient = −0.02; *p* = 0.56). Together, these results suggest that the genomic distance alone cannot be used to predict behavior under FL compared to U, indicating that genetic variation in specific genes may account for the difference.

## *2.5. Correlation Analysis*

For a more detailed view of the interrelations between chlorophyll *a* fluorescence, growth, and development data, we constructed a correlation matrix between the average values for each accession and the measured variable, and we included latitude of origin and the number of leaves formed until flowering (Table 1). Strong correlations between most chlorophyll *a* fluorescence parameters (except for the relationship between NPQ and Fv/Fm) and all growth and developmental parameters were found (Table 2). Average leaf size correlated strongly and positively with ΦPSII, and negatively with NPQ (Figure 5), suggesting that light use efficiency had positive effects on leaf growth. These correlations are especially apparent for plants grown under U (yellow symbols in Figure 5). Both projected leaf area and the number of visible leaves correlated positively with Fv/Fm (Figure 6A,B). Also, Fv/Fm correlated negatively with the number of leaves formed until flowering (Figure 6C).

**Figure 4.** Clustering of 36 Arabidopsis accessions based on (**A**) differences in phenotypical logarithmic response ratios under fluctuating vs. uniform light (FL/U ratio; plot produced using heatmap.2 function with default settings of R package gplot) and (**B**) genomic distances, based on published SNP data (single linkage).

**Table 2.** Correlation matrix for traits observed in plants grown under uniform and fluctuating light. Blue colored backgrounds indicate a positive correlation, red indicates negative; the more strongly colored the background, the steeper the slope of the correlation. Statistically significant correlations (*p* < 0.05) are marked in bold. Numbers indicate Spearman's ρ, stars indicate the significance of the correlation, as: \*\*\* = *p* < 0.001, \*\* = *p* < 0.01 and \* = *p* < 0.05 (n = 15–72). Lat., latitude of origin (◦), #leaves flowering, number of leaves at flowering.


**Figure 5.** Relationships between average leaf size and (**A**) photosystem II operating efficiency (ΦPSII) and (**B**) non-photochemical quenching (NPQ) in 36 Arabidopsis accessions acclimated to uniform (U) and fluctuating light intensities (FL). Data from plants grown under FL are sorted by experiment 1 (squares) and experiment 2 (triangles). Averages ± SE (n = 5–7). Spearman's ρ and the significance of a linear correlation through all points is shown (\*\*\* = *p* < 0.001).

Next, we tested whether response ratios of parameters derived from FL over U grown plants (FL/U) correlated at the trait level, as well as with latitude of origin and number of leaves until flowering (Table 3). This analysis showed a strong, positive correlation between ΔΦPSII and ΔFv/Fm (Table 3), suggesting that accessions with a strong reduction in ΦPSII also showed a stronger reduction in Fv/Fm under FL. As might be expected, the response ratio of projected leaf area correlated strongly and positively with the response ratios of leaf number and leaf size (Table 3). We tested whether response ratios derived from either FL experiment 1 or FL experiment 2 yielded similar results, by repeating the same correlation analysis as shown in Table 3 for these two subsets of data (Tables S1 and S2). Both subsets yielded highly similar correlation coefficients, which themselves showed a strong linear correlation (Figure S1, *p* < 0.001). Correlation coefficients from each FL experiment subset also correlated strongly with those derived from the total dataset as shown in Table 3 (*p* < 0.001 in both cases, plots not shown). These results strongly suggest that the effects of fluctuating growth light on plants were repeatable within our experimental setup, and that hence similar conclusions can be drawn from both FL experiment 1 and FL experiment 2, further validating our findings.

**Figure 6.** Relationships between photosystem II maximum quantum efficiency (Fv/Fm) and (**A**) projected leaf area, (**B**) the number of visible leaves and (**C**) the number of leaves required until flowering, in 36 Arabidopsis accessions acclimated to uniform (U) and fluctuating light intensities (FL). Data from plants grown under FL are sorted by experiment 1 (squares) and experiment 2 (triangles). Averages ± SE (n = 5–7). Spearman's ρ and the significance of a linear correlation through all points is shown (\*\* = *p* < 0.01, \* = *p* < 0.05).

**Table 3.** Correlation matrix for response ratio in traits under fluctuating light divided by those under uniform light (Δ = FL/U). Blue colored backgrounds indicate a positive correlation, red indicates negative; the more strongly colored the background, the steeper the slope of the correlation. Statistically significant correlations (*p* < 0.05) are marked in bold. Numbers indicate Spearman's ρ, stars indicate the significance of the correlation, as: \*\*\* = *p* < 0.001, \*\* = *p* < 0.01 and \* = *p* < 0.05 (n = 15–36). Lat., latitude of origin (◦), #leaves flowering, number of leaves at flowering.


The response ratio (FL/U) of PLA correlated strongly and negatively with PLA of plants grown under U (Figure 7A), suggesting that the reduction in PLA under FL was strongest in plants that showed high growth under U. The response ratio of average leaf size correlated positively with ΔFv/Fm and negatively with ΔNPQ (Figure 7B,C), again suggesting that leaf growth was directly related to rates of photoprotection and photoinhibition. Interestingly, the latitude of origin correlated negatively with ΔΦPSII (Figure 8), revealing a trend for photosynthesis of accessions collected further north on the globe to be more negatively affected by FL. Lastly, both the response ratio of projected leaf area and of number of visible leaves correlated positively with ΔFv/Fm, but this correlation was less meaningful given the large uncertainty around the mean for values of single accessions (Figure S2).

**Figure 7.** Changes in growth and chlorophyll *a* fluorescence in 36 Arabidopsis accessions grown under fluctuating compared to uniform light. (**A**) Relationship between projected leaf area (PLA) under uniform light and the response ratio of PLA under fluctuating light divided by PLA under uniform light (Δ = FL/U), (**B**) relationship between the response ratio of average leaf size and the response ratio of photosystem II maximum quantum efficiency (ΔFv/Fm), and (**C**) relationship between the response ratio of average leaf size and the response ratio of non-photochemical quenching (ΔNPQ). Data are sorted by FL experiment 1 (squares) and FL experiment 2 (triangles). Averages ± SE (n = 5–7). Spearman's ρ and the significance of a linear correlation through all points is shown (\*\*\* = *p* <0.001, \* = *p*< 0.05).

**Figure 8.** Relationship between latitude of origin of 36 Arabidopsis accessions and the response ratio in photosystem II operating efficiency (ΔΦPSII) between plants grown under fluctuating light intensities (FL) divided by values from plants grown under uniform light intensities (Δ = FL/U). Data are sorted by FL experiment 1 (squares) and FL experiment 2 (triangles). Averages ± SE (n = 5–7). Spearman's ρ and the significance of a linear correlation through all points is shown (*p* < 0.05).

## **3. Discussion**

As photoautotrophs, plants interact with light in a direct manner. While our knowledge on rapid responses to changes in light intensity in the range of seconds to minutes is well advanced of plants grown under uniform light regimes (U; reviewed in [12,21–23]), less is known about the long-term response, i.e., acclimation occurring within days to fluctuating light (FL; [1,2]), and how this affects short-term responses to FL [2]. Here, we found that PLA and ΦPSII generally decrease, NPQ generally increases, and Fv/Fm as well as number of leaves per plant generally remain unchanged, when plants are grown under FL compared to U. This, together with the large phenotypic variation observed, broadly confirms our hypotheses. However, we acknowledge that because (i) plants grown under U and FL received different light sums and day lengths during the first 14 days after sowing and (ii) the FL experiment was run twice, with different groups of accessions each, further experimental work should be conducted to confirm the robustness of these results. Additionally, it will be desirable to increase the number of accessions for future analyses.
