Identification of QTLs for dynamic and steady state photosynthetic traits in a barley mapping population

Enhancing the photosynthetic induction response to fluctuating light has been suggested as a key target for improvement in crop breeding programs, with the potential to substantially increase whole canopy carbon assimilation and contribute to crop yield potential. Rubisco activation may be the main physiological process that will allow us to achieve such a goal. In this study, we phenotypically assessed the rubisco activation rate in a doubled haploid (DH) barley mapping population [131 lines from a Yerong/Franklin (Y/F) cross] after a switch from moderate to saturating light. Rates of rubisco activation were found to be highly variable across the mapping population, with a median activation rate of 0.1 min−1 in the slowest genotype and 0.74 min−1 in the fastest genotype. A QTL for rubisco activation rate was identified on chromosome 7H. This is the first report on the identification of a QTL for rubisco activation rate in planta and the discovery opens the door to marker assisted breeding to improve whole canopy photosynthesis of barley. Further strength is given to this finding as this QTL colocalised with QTLs identified for steady state photosynthesis and stomatal conductance. Several other distinct QTLs were identified for these steady state traits, with a common overlapping QTL on chromosome 2H, and distinct QTLs for photosynthesis and stomatal conductance identified on chromosomes 4H and 5H respectively. Future work should aim to validate these QTLs under field conditions so that they can be used to aid plant breeding efforts. Highlight Significant variation exists in the photosynthetic induction response after a switch from moderate to saturating light across a barley doubled haploid population. A QTL for rubisco activation rate was identified on chromosome 7H, as well as overlapping QTLs for steady state photosynthesis and stomatal conductance.

Introduction single photosynthetic induction curve at a low intracellular CO2 concentration (i.e. < 300 barley cultivars Yerong and Franklin (Y/F). This population contained 177 DH lines and was 154 maintained at the Plant Breeding Institute at The University of Sydney. The Y/F mapping 155 population has been extensively used for QTL mapping for both morphological (Xue et al. 156 2008) and physiological (Zhang et al. 2016) traits, as well as disease resistance (Singh et al. 157 2014;Dracatos et al. 2016). In this study 131 lines from this population were phenotypically 158 assessed for steady state and dynamic photosynthetic traits. Due to the availability of seed 159 and genotypic data, only 127 DH lines were used for QTL analyses. A second DH barley 160 population (from a cross between VB9104 and Dash) was also phenotyped for 161 photosynthetic traits however due to the low number of lines with available genotypic data 162 this population was not included in further analyses (phenotyping results are however 163 presented in Figures S4 and S5). 164 165 Plants were grown in a controlled environment room for approximately five weeks prior to 166 measurement. Day temperature was 25°C during a 14 h light period and night temperature 167 was 17°C during a 10 h dark period. Relative humidity was maintained at 70% while daytime 168 PPFD was approximately 600 μmol m -2 s -1 at the top of the plants. Seeds were planted in 169 potting mix enriched with slow-release fertilizer (Osmocote Exact,Scotts,Sydney,NSW,170 Australia). Six seeds per genotype were sown in 6 L pots and grown for three weeks before 171 being thinned to three plants per pot. Seed was sown sequentially in time to make sure that 172 all measurements were conducted at the same growth stage. Plants were watered daily to 173 field capacity. 174 175

Photosynthetic measurements 176
Plants were moved from the controlled environment room to a temperature-controlled 177 growth cabinet [temperature 25°C; relative humidity 70%]. Two or three of the youngest 178 fully expanded leaves of a single plant were sealed in a 2x6 cm leaf cuvette (Li6400 11; LI-179 COR, Lincoln, NE, USA) fitted to a LI-COR LI-6400XT gas exchange system to fill the cuvette 180 without overlapping. This simulated an instantaneous shift in light intensity from 600 μmol 181 during a sunfleck. Chamber conditions were set to closely match those of the controlled environment room [leaf temperature 25°C; cuvette CO2 (Ca) 400 µmol mol -1 ; relative 184 humidity 70%], with the exception of PPFD which was set to 1300 µmol m -2 s -1 using a red-185 green-blue light source (Li6400 18A; LI-COR) set to 10% blue and 90% red light. 186 Measurements of photosynthetic gas exchange rates (A and gs) were recorded once per 187 minute immediately after the leaf was inserted into the chamber until photosynthesis had 188 reached steady state. Preliminary photosynthetic light response curves were measured with 189 plants grown under the same conditions to ensure that 1300 μmol m -2 s -1 was saturating 190 and that 600 μmol m -2 s -1 was non-saturating (results shown in Figure S1). 191 192 Rubisco activation rate was calculated using a modified method of Soleh et al. (2016). 193 Photosynthetic data was first normalised to an assumed intercellular CO2 concentration (ci) 194 of 300 µmol mol -1 , using the following equation: 195 where A * is the normalised photosynthetic rate, A is the measured photosynthetic rate and 199 ci is the measured intercellular CO2 concentration. This effectively removed the influence of 200 stomatal opening/closure for the induction phase. The initial rubisco activation rate (1/t) 201 was modelled from the plot of the logarithmic difference between A * and its maximum 202 value after induction (A * max) against the time taken for induction (representative data shown 203 in Figure 1). From this plot, the value of 1/t was determined from the slope of the linear 204 regression on data points in the range of 2 to 5 mins after induction, and points after this 205 that aligned well with these initial points (with an R 2 > 90%). 206

Genetic analysis and QTL mapping 207
The genotypic data and genetic linkage map for the Yerong/Franklin DH population used for 208 QTL analysis for rubisco activity and steady state photosynthetic traits in the present study 209 was previously described by Singh et al. (2015). In brief, the Y/F genetic map is comprised of 210 Raleigh, NC, USA), carrying out 1,000 iterations permutation analysis with steps at 1 cM, and 217 with a 0.05 confidence level for all traits. 218

Statistical analyses 220
All other modelling and statistical analyses were performed in R (R Core Team, 2019). 221 and genotypes, general trends were quite clear (representative induction curves from one 227 day of measurements is shown in Figure 2). Net photosynthesis (A) increased immediately 228 after transition from to saturating light for all leaves. Stomatal responses were more 229 variable than those of photosynthesis but there tended to be an initial reduction in gs after 230 transition to saturating light and then a gradual rise towards steady state. By normalising 231 photosynthesis to a constant ci of 300 ppm, we were able to obtain a measure of 232 photosynthesis limited by rubisco carboxylation unobstructed by variation in stomatal 233 kinetics (A*). A* showed a similar trend to A, increasing immediately after the switch from 234 low to high light. 235

236
QTLs for rubisco activation rate 237 Rubisco activation rates of the parental lines Yerong and Franklin were found to differ, with 238 within-genotype medians of 0.38 min -1 and 0.74 min -1 respectively. Wide variation in 1/t 239 was found across the population (Figure 3), with within-genotype medians ranging from 240 0.099 min -1 to 0.74 min -1 . Interestingly, the parental line Franklin was found to have the 241 fastest rate of rubisco activation. A frequency distribution of 1/t was plotted for the 242 population and was found to follow a normal distribution suggesting that rubisco activation 243 rate was under complex genetic control ( Figure S2). CIM analysis revealed the presence of a 244 distinct QTL for rubisco activation rate ( Figure 4; further details in Table 1). Q1/t.sun-7H, 245 was located at 41.67 cM on chromosome 7H (proximal to DarT marker bPb-9601 marker) 246 accounting for 10.48% of the phenotypic variance in this trait. 247 248

Steady state photosynthesis and equilibration time tests 249
Variation was also found in steady state photosynthetic rates across the population ( Figure  250 5). Median rates of A and gs were 17.45 μmol m -2 s -1 and 0.31 mmol m -2 s -1 , respectively. 251 From this phenotyping data, there was no correlation found between steady state A and 1/t 252 (p > 0.05; Figure 6). 253 pronounced the earlier the measurements were recorded after enclosing the leaf in the 257 chamber of the IRGA. Mean values of A and gs were both underestimated by 21% at five 258 minutes compared to steady state. It should be noted that although some of the fastest 259 genotypes reached steady state after five minutes, most of the lines did not. In fact, gs was 260 underestimated by 82% for one of the genotypes and A was underestimated by 54% for 261 another if measurements were recorded after just five minutes. 262

263
To assess the importance of equilibration time for accurate identification of steady state 264 QTLs, QTL mapping was first performed for steady state A and gs. Frequency distributions 265 were plotted for both traits and they followed a normal distribution suggesting they are 266 under complex genetic control ( Figure S3). Several QTL were found for both A and gs (Table  267 1). Trait co-location was observed on chromosome 7H whereby the position of the 268 Q1/t.sun-7H QTL was almost identical to QTL for both A and gs. This suggests a region on 269 the short arm of chromosome 7H either carries a single gene or more likely a cluster of 270 genes responsible for the genetic control of photosynthesis, stomatal conductance and 271 rubisco activation. For steady state A and gs, additional overlapping and distinct QTL were 272 identified. A common overlapping QTL for both A and gs was identified, peaking at 27.03 cM 273 on chromosome 2H, whilst distinct QTL were identified on chromosomes 4H (41.67 cM) and 274 5H (53.39 cM) for A and gs respectively. 275 276 QTL mapping was then performed with data collected at five, ten and fifteen minutes after 277 the start of induction for comparison with detected steady state QTLs (coloured traces in 278 Figure 7). Although most QTL were still identified with non-steady state data, the 279 significance these QTL peaks were found to be weakened under non-steady state 280 conditions. This was particularly evident for the gs QTL identified on chromosome 7H ( Figure  281 7h), with the LOD score of this QTL dropping from 6.8 when using steady state data to 4.9, 282 3.9 and 3.3 when using data collected at 15 min, 10 min and 5 min after the start of 283 induction, respectively. 284 As in other crops, we found rubisco activation rate to be a highly variable trait across 288 genotypes of barley, aiding in the discovery of a significant QTL in our doubled haploid 289 population. QTLs were also identified for steady state photosynthetic parameters, including 290 co-localised QTLs for A, gs and 1/t on chromosome 7H. The importance of adequate 291 equilibration time in the measurement of steady state gas exchange was highlighted by 292 comparing these results to those obtained using arbitrary non-steady state rates at 5, 10 293 and 15 min after the start of induction. The significance of QTLs was reduced if steady state 294 conditions had not been reached. It is important now that we explore and exploit natural variation in photosynthetic traits 322 across plant populations (for review see Furbank et al., 2020). As in previous studies with 323 other species, we identified significant variation in rubisco activation rate across barley 324 genotypes. We identified a QTL for rubisco activation rate, as well as several QTLs for steady 325 state A and gs. Q1/t.sun-7H was flanked by the bpb-9601 DArT marker which has previously 326 been associated with both grain yield and crop spike number in the Yerong/Franklin 327 population (Xue et al., 2010). This marker is of particular interest as it also flanks QTLs that 328 we identified for steady state A and gs (QA.sun-7H and Qgs.sun-7H in Table 1) Our study focussed on a step change from moderate (600 µmol m -2 s -1 ) to saturating light 358 (1300 µmol m -2 s -1 ), rather than low to high light as has been reported previously (i.e. 50 -359 1500 µmol m -2 s -1 in Taylor and Long, 2017). We felt this approach would provide more 360 valuable information for plant breeding, as it more accurately represents the light regime 361 experienced by the second youngest leaves in the canopy, which for wheat have been 362 reported to receive between 300 -700 µmol m -2 s -1 PPFD when not in a sunfleck (Townsend 363 et al., 2018). Whilst leaves lower in the canopy receive much less light than this (< 300 µmol 364 m -2 s -1 ), these leaves are also less likely to be exposed to sunflecks and also have a much- Our study has focussed on rubisco activation however this is only one part of the dynamic 377 photosynthesis puzzle, in which all the pieces must be investigated to fully understand 378 potential improvements that could be made to whole canopy photosynthesis. Responses of 379 stomata can also limit photosynthesis in fluctuating light. Faster stomatal opening has now been shown to improve net photosynthesis and biomass production in overexpressing 381 mutants of Arabidopsis thaliana compared to wild type plants (Kimura et al., 2020). And so,382 if improvements are made to rubisco activation rate without also considering rates of 383 stomatal opening/closure, the dominant limitation will likely shift in the direction of the 384 stomata. In effect, this could nullify any improvements made to rubisco activation in terms 385 of net photosynthesis. On a positive note, recent work has highlighted that stomatal traits 386 can be linked to rubisco kinetics during leaf development in some plant species (Conesa et 387 al., 2019), and it has long been realised that stomata respond to photosynthetic activity in 388 the mesophyll (Messinger et al., 2006). It is therefore conceivable that improving rubisco to screen these two populations in the field and validate the QTLs we identified in this 403 study. It is also important that we understand if these QTLS are strong under sub-optimal 404 conditions (i.e. under drought or heat stress), as for most growers such conditions can be 405 common during a growing season. 406 407 A note on gas exchange methodology 408 It is common practice to allow a leaf to stabilise to the chamber conditions of an IRGA, yet 409 the recent push for "high throughput" and "big data" approaches in plant physiology may 410 have made researchers complacent. We hypothesized that this complacency could impact resulted in less accurate detection of QTLs. It is likely that false QTL identifications are 414 worsened by the high variability in photosynthetic induction kinetics that exists across this 415 population (and has also been found in other crop species) and the fact that there is no 416 clear relationship between steady state and dynamic photosynthesis. This result reinforces 417 the importance of good gas exchange technique. The push for high-throughput 418 measurements has resulted in new fast methods, such as the Rapid A/ci method (Stinziano 419 et al., 2017), being developed yet it must be highlighted that most of these methods still 420 rely on the assumption of steady state conditions and these will therefore still be limited by Merr.] genotypes that is not correlated with steady-state photosynthetic capacity. 515 Photosynthesis Research 131, 305-315. 516  regression used to calculate rubisco activation rate (1/t). The orange crosses represent the 573 measured photosynthetic rate A; the grey squares the ci = 300 µmol mol -1 normalised 574 photosynthetic rate A * ; and the blue circles the logarithmic difference between the fully 575 induced photosynthetic rate A * max and A * . Filled circles represent the data points used in the 576 linear regression to estimate 1/t, the rubisco activation rate. The slope of the regression 577 represents 1/t, in this case 0.31 min -1 . 578