Two experiments (field and controlled conditions) were conducted during the 2019–2020 growing season. Treatments in each of the experiments consisted of a factorial combination of (i) four near-isogenic lines (NILs) for PPD-H1 alleles (ppd-H1 or Ppd-H1) under two contrasting PHYC genetic backgrounds (PhyC-l and PhyC-e) and (ii) two photoperiod conditions (short and long days). NILs were produced at CSIRO (Canberra, Australia) after five cycles of backcrossing, using different donors of VRN-H1/PHYC and PPD-H1 alleles into the facultative recurrent barley cultivar “WI4441” (Oliver et al., 2013).

The four genotypes were actually aimed to be isogenic for allelic constitution of PPD-H1 and VRN-H1, but as the latter is closely linked to PHYC (Szucs et al., 2006; Nishida et al., 2013; Pankin et al., 2014), the NILs were actually vrn-H1+PhyC-e and Vrn-H1+PhyC-l ( Table 1 ), as demonstrated by Ochagavía et al. (2022) who genotyped these NILs, finding that winter (vrn-H1) lines carried the PhyC-e allele and spring (Vrn-H1) lines the late allele (PhyC-l). Although this linkage prevents a clear separation of the effects VRN-H1 and PHYC genes, in the experiments reported here, plants were vernalised (see below), and therefore there were no effects of VRN-H1 on any developmental attribute. Thus, for simplicity, we considered herein these NILs as the combinations of the two allelic constitutions of PPD-H1 and PHYC ( Table 1 ). All lines had the dominant Vrn-H2 and Ppd-H2 alleles and haplotype II of HvCEN (Oliver et al., 2013; Ochagavía et al., 2022); i.e., all effects on developmental characteristics will be due to the action of PPD-H1 alleles under the particular backgrounds of contrasting alleles of PHYC and contrasting photoperiods.

The field experiment (Exp1) was sown on 03 December 2019 in a facility with photoperiod control available at the campus of the University of Lleida, Spain (41°37′50″N, 0°35′27″E; altitude 180 m) in a fine loamy, mixed (calcareous), thermic soil classified as Typic Xerofluvent, according to the USDA taxonomy (Soil Survey Staff, 1999). Seeds of each material were distributed in strips of biodegradable paper, ensuring a uniform distance between plants within rows as well as a uniform seedling depth.

Plots were maintained throughout the whole cycle under either (i) natural conditions, with an average photoperiod from seedling emergence (SE) to flowering (Fw) of ca. 12 h (11.7 h ± 0.02 h), or (ii) a 24-h daylength, artificially extending the natural photoperiod with low-intensity (60 W) incandescent bulbs positioned on top of the designated plots. The radiation intensity was more than enough to produce the daylength signal, but increased radiation only negligibly (ca. ~ 3.6 μmol m⁻² s⁻¹ PAR at canopy level), below the light compensation point for barley (i.e., irradiance at which photosynthesis equals respiration and net photosynthesis is zero) normally around 10–15 μmol m⁻² s⁻¹ (Arenas-Corraliza et al., 2019; Chen et al., 2021), allowing plants to alter their developmental patterns but not affecting daily growth directly.

Exp1 was drip-irrigated when needed in order to avoid water stress. Weeds, diseases, and insects were controlled or prevented by spraying herbicides, fungicides, and insecticides at doses recommended by their manufacturers.

In the growth chamber experiment (Exp2), NILs were grown at the relatively low and constant temperature of 12°C (to expose plants to a temperature approaching the average temperature from SE to Fw more realistically than most controlled conditions growing temperate cereals that set growing temperatures at 18°C–25°C, accelerating development to minimise experimental duration). Indeed, the mean temperature from seedling emergence to flowering in the field experiment was 9.2°C. The two different temperature regimes in our experiments—lower average temperatures with natural daily and monthly variations in the field and slightly higher and constant temperatures in the growth chambers—along with other differences in the experimental setups could affect the strength of our conclusions. The conclusions will be more solid if the results are consistent across both experiments and weaker if the results are conflicting. The photoperiod treatments were 12 and 24 h; in the latter, only half of the lights were switched on during the duration of the day to compensate for the difference in daylength, so that in both conditions the daily radiation was the same (5.2 MJ m⁻² day⁻¹). In Exp2, seeds were germinated in 235 cm³ black plastic pots filled with 110 g of a soil mixture (70% w/w peat and 30% w/w organic amendment) freshly prepared before sowing. There was only one seedling per pot, and after being vernalised (see below), we transferred a set of 26 pots per NIL with seedlings at exactly the same stage (see below) to each of the two cabinets, previously configured for temperature and photoperiod conditions. Many of these plants were sampled for periodic dissections and intermediate determinations during the duration of the experiment, but at least three out of the 26 were left intact until flowering. Within each chamber, the pots were distributed randomly on trays and rotated at least twice weekly to avoid any possible positional effect within the chamber. Plants were irrigated daily, and each pot was fertilised with both macro- and micronutrients to avoid nutritional deficiencies.

In both experiments, all plants were vernalised; in Exp1, plants were naturally vernalised when exposed to winter (as they were sown in late fall). From sowing to the end of winter, seedlings were exposed to 41 fully vernalising days (days with mean temperatures with maximum effect on vernalisation, between 0°C and 8°C; Flood and Halloran, 1986; Brooking and Jamieson, 2002; Figure 1 ) plus 26 days with mean temperatures between 8°C and 10°C [that are also strongly vernalising temperatures, considering that vernalisation is produced when temperatures are up to 15°C; Brooking and Jamieson (2002)]. In Exp2, pots were exposed to vernalising temperature (4°C constant during the whole day) for 29 days in a cold room. Firstly, the pots were filled and sown at exactly the same depth with one seed per pot, but with 35% more pots than needed for the experiment (i.e., we sowed and included 70 pots of each individual NIL in the vernalisation pretreatment; in each of the two growth chambers prepared for the experiment, we transferred only 26 pots per NIL). This allowed us to discard not only the few pots in which seedlings did not emerge but also the tails of early- and late-emerging seedlings. As a result, when the experiment started and we transferred the pots from the vernalisation room to the growth chambers at 12 or 24 h photoperiod, all plants were extremely uniform (averaging 1.06 ± 0.02 emerged leaves).

After sowing the pots, before transferring them to the cool room for vernalisation, they were watered and left for 1 day at room temperature to trigger the germination process. Subsequently, all pots were transferred to a cool room. Finally, the 52 pots per NIL selected for having homogeneous seedlings were transferred to the growth chambers, and the experiment started (and for simplicity and using the same terms in both experiments, the date of starting the experiment was identified as “seedling emergence”, which, strictly talking, was slightly later in Exp2).

Treatments in Exp1 were arranged in a split-plot design, where the main plots, allocated to three complete blocks, were assigned to the photoperiod treatments, and the subplots, allocated randomly within the main plots, were assigned to the NILs. Subplots were 3.5 m in length and 1.2 m wide, with six rows (0.2 m apart) and a seedling rate of 200 plants m⁻². In Exp2 within each cabinet, a set of 26 barley plants of each of the four NILs (i.e., 104 plants in each photoperiod condition) were arranged in a complete randomised design.

The duration of both time from seedling emergence to flowering and of the phases composing it (i.e., from seedling emergence to awn initiation [SE-AI], from then to flag leaf [AI-FL], and from then to flowering [FL-Fw]) was expressed in thermal time, using the average temperature recorded at the site in Exp1 (Meteorological station from the Meteorological Service of the Government of Catalonia [Meteocat]) and the temperature of the chamber in Exp2, assuming a base temperature of 0°C, as standardly done (Kirby, 1988). The developmental stages determined (SE, AI, FL, and Fw) were in accordance with the Zadoks’ scale (Z09-10; Z31-33; Z39; Z55; Zadoks et al., 1974). However, for a more accurate determination, AI and Fw were determined, taking into account internal structures not normally visible to the naked eye. Awn initiation was determined microscopically when the tip of the lemma primordium started to grow and curve over the stamen primordia (~ W4.5). Flowering was determined as the time of pollination by regular microscopic dissection of the main spike and determining when it reached stage 10 on the Waddington et al. (1983) scale (i.e., when styles are curved outward with stigmatic branches widely spread and pollen grains visible on stigmatic hairs).

From SE to Fw, main stems were monitored once a week to determine the duration of different phenological phases [as delimited by stages determined externally by the scale of Zadoks et al. (1974) and internally by the scale of Waddington et al. (1983)]. In addition, three plants per experimental unit were randomly selected ³ and tagged soon after SE, and the number of leaves that emerged on the main shoot was recorded twice a week following the scale developed by Haun (1973), while simultaneously the number of emerged and living tillers were determined.

From SE onward, representative plants of each NIL (three in each experimental unit of Exp1 and two in each chamber of Exp2) were sampled twice a week, and apical development was observed under the microscope after dissecting the main shoot apex. In addition, a detailed morphological analysis of spikelet and floret development of the main shoot spikes was carried out following the scale described by Waddington et al. (1983). The apices were dissected under a stereomicroscope Leica MZ 80 (Leica Microscopy System Ltd., Heerbrugg, Switzerland) equipped with a digital camera (model DFC420, Leica).

Phyllochron (i.e., the thermal time interval between the appearance of two successive leaves) was calculated as the reciprocal of the rate of leaf appearance (i.e., the slope of the relationship between the cumulative number of leaves on the main shoot and the thermal time). Whenever a linear model did not produce a random distribution of residuals, a bilinear model was fitted (with one phyllochron for the first leaves and another one for the last leaves) and, in these cases, considering the average phyllochron of all leaves as well as those for early- and late-appearing leaves.

Analysis of variance (ANOVA) was used to partition variation into effects of treatments and their interactions using the statistical software JMP^® Pro version 16.0 (SAS Institute Inc., Cary, NC, USA). Differences among means were compared using the least significant difference test (LSD, considered to be statistically significant if p < 0.05). To assess the degree of relationships between variables, linear regression analyses were performed. Polynomial regressions (Loess smooth line) were performed for the numbers of leaves, tillers, and floret dynamics, using an alpha of 0.75 and 95% confidence interval. Graphs were created in R using the package “ggplot2” (Wickham, 2016; R Core Team, 2020).

As expected for a quantitative long-day plant, the overall duration of the cycle from SE to Fw was reduced when plants were grown under long days (cf. right and left panels in Figure 2 ). More relevantly, in the context of the aims of this study, the effect of PPD-H1 gene on time to flowering in each of the contrasting PHYC genetic backgrounds depended on the photoperiod condition. There was an interaction between NILs and photoperiod on time to flowering: at 12 h photoperiod, Fw was delayed by the action of the ppd-H1-insensitive allele (although the effect was a nonsignificant trend when the PhyC-l allele was in the background in Exp1, the direct effect of ppd-H1 was still significant when considered across the two PHYC backgrounds; see boxplots in Figure 2A ), while under 24 h photoperiod, the ppd-H1 allele did not significantly delay Fw ( Figures 2B, D ). The responses of time to Fw caused by PPD-H1 across the two PHYC backgrounds were clearer under controlled conditions, but importantly, we observed the same effects in the field (cf. see boxplots in Figures 2A, C ).

Across all sources of variation, time to Fw was very strongly related (R ² > 0.95; p < 0.001) to the duration of both component phases, from SE to AI ( Figures 3A, B ) and from AI to Fw ( Figures 3C, D ) consistently across the two different experiments (i.e., the effect of all treatments together on time to Fw was due to effects on both phases). However, a major part of the similarly strong relationships of time to Fw with its two component phases was driven by the photoperiod growing condition: the phases of leaf and spikelet initiation and of floret development within spikelets (and then of spikelet survival) were both similarly affected by the photoperiodic condition in both experiments ( Figure 3 ). Focusing on the effects produced by the PPD-H1 alleles, the delay in Fw produced by the insensitivity allele was only significant under short photoperiod conditions in both experiments (see boxplots in Figure 2 ), and this effect was clearer in the duration of the period from SE to AI than in that from AI to Fw (although the latter also showed a consistent, though non-significant, trend to be delayed due to the action of the ppd-H1 allele; open boxplots in Figure 3 ). Thus, under these relatively short photoperiods, there seemed to have been a sort of knock-on effect caused by the ppd-H1 allele, clearly lengthening the duration of the SE-AI phase but also tending to lengthen that of the AI-Fw phase.

The leaf appearance rate was constant for the initial ca. six leaves across NILs, as indicated by the linear relationships when plotting leaf number vs. thermal time. However, when the final leaf number (FLN) was clearly higher than this threshold (particularly under short photoperiod), the rate of leaf appearance for the later leaves decreased, exhibiting a bilinear relationship between leaf number and thermal time across NILs (and the higher the FLN, the stronger the increase in phyllochron; Figures 4A, C ).

Time to the appearance of the flag leaf was clearly affected by photoperiod across NILs in both experiments (cf. the pairs of boxplots under short and long photoperiods in Figures 5A, B ), driven by the effects of the photoperiod condition on both FLN and average phyllochron ( Figures 5C, E [Exp1], Figures 5D, F [Exp2]).

The allelic form of the PPD-H1 gene affected phyllochron slightly but consistently across photoperiods and experiments, although the effect was significant only under controlled conditions ( Figures 5E, F ). Under long photoperiods, NILs having Ppd-H1-sensitive and ppd-H1-insensitive alleles had phyllochrons of, on average, 77°C and 82°C day⁻¹ leaf in Exp1 and 68°C and 76°C day⁻¹ leaf in Exp2, respectively ( Figures 5E, F ). Under a short photoperiod, NILs having the Ppd-H1-sensitive allele had on average a consistently shorter phyllochron (95°C and 112°C day⁻¹ leaf) than those carrying the ppd-H1-insensitive allele (98°C and 125°C day⁻¹ leaf, in Exp1 and Exp2, respectively). PPD-H1 alleles did not affect FLN in Exp1 under either of the two photoperiod conditions ( Figure 5C ). However, in Exp2, NILs having ppd-H1-insensitive alleles increased FLN, though rather slightly by less than one leaf, in both photoperiod conditions ( Figure 5D ).

Thermal time to flag leaf was better explained by phyllochron (R ² = 0.94 and R ² = 0.99 for Exp1 and Exp2, respectively; Figures 6A, B ) than by FLN (R ² = 0.87 and R ² = 0.86 for Exp1 and Exp2, respectively) ( Figures 6C, D ).

Tillering dynamics was similar across experiments and NILs with relatively limited tillering and consequently having very little tiller mortality (Figures 7A, C–E, G, H). The effects of PPD-H1 alleles were not large nor consistent for all cases, but when the environmental background was the short photoperiod and the genetic background included the PhyC-l allele, in general, the NIL with the insensitive ppd-H1 allele produced more spikes per plant due to reduced tiller mortality in Exp1 ( Figure 7B ) and maintained tillering a bit longer in Exp2 ( Figure 7F ).

In general, in the central spikelets of the main shoot spike, awn initiation and flag leaf stages coincided with floret developmental stages of W4.75 and W8, respectively, of the scale of Waddington et al. (1983), varying only very slightly across NILs, experiments, and photoperiod conditions ( Supplementary Figure S1 ).

NILs having sensitive Ppd-H1 alleles slightly accelerated flowering by promoting early shoot apex development, whose magnitude depended on the photoperiod condition and PHYC genetic background ( Figure 8 ; Supplementary Figures S2 , S3 ). This effect of PPD-H1 was only slight on long days ( Figures 8C, D, G, H ), when the time to flowering was not significantly delayed (see above). Under short photoperiod conditions, florets in the insensitive ppd-H1 lines showed a much clearer development deceleration ( Figures 8A, E, F ), except for the PhyC-l background under natural photoperiod in Exp1, where the difference was not significant ( Figure 8B ).

This effect of PPD-H1 is reflected in the developmental rates of a particular organ (florets), which has been observed for the phenological effects on flowering time.