An in situ approach to characterizing photosynthetic gas exchange of rice panicle

Background Photosynthesis of reproductive organs in C3 cereals is generally regarded as important to crop yield. Whereas, photosynthetic characteristics of reproductive organs are much less understood as compared to leaf photosynthesis, mainly due to methodological limitations. To date, many indirect methods have been developed to study photosynthesis of reproductive organs and its contribution to grain yield, such as organ shading, application of herbicides and photosynthetic measurement of excised organs or tissues, which might be intrusive and cause biases. Thus, a robust and in situ approach needs to be developed. Results Here we report the development of a custom-built panicle photosynthesis chamber (P-chamber), which can be connected to standard infrared gas analyzers to study photosynthetic/respiratory rate of a rice panicle. With the P-chamber, we measured panicle photosynthetic characteristics of seven high-yielding elite japonica, japonica-indica hybrid and indica rice cultivars. Results show that, (1) rice panicle is photosynthetically active during grain filling, and there are substantial inter-cultivar variations in panicle photosynthetic and respiratory rates, no matter on a whole panicle basis, on an area basis or on a single spikelet basis; (2) among the seven testing cultivars, whole-panicle gross photosynthetic rates are 17–54 nmol s−1 5 days after heading under photon flux density (PFD) of 2000 μmol (photons) m−2 s−1, which represent some 20–38% of that of the corresponding flag leaves; (3) rice panicle photosynthesis has higher apparent CO2 compensation point, light compensation point and apparent CO2 saturation point, as compared to that of a typical leaf; (4) there is a strong and significant positive correlation between gross photosynthetic rate 5 days after heading on a single spikelet basis and grain setting rate at harvest (Pearson correlation coefficient r = 0.93, p value < 0.0001). Conclusions Rice panicle gross photosynthesis is significant, has great natural variation, and plays an underappreciated role in grain yield formation. The P-Chamber can be used as a tool to study in situ photosynthetic characteristics of irregular non-foliar plant organs, such as ears, culms, leaf sheaths, fruits and branches, which is a relatively less explored area in current cereal breeding community.

Leaves, especially flag leaves located at the upmost of stems, have been believed to be the dominant contributor of post-anthesis photosynthesis [5,6]. Nevertheless, reproductive organs, including glumes, caryopsis pericarps, and awns, are green before mature and photosynthetically active [7,8], suggesting a possibility of them being a major source of photosynthate for grain filling as well. In fact, reproductive organs have several potential advantages over leaves in terms of photosynthate production and transportation: (1) they generally locate at the upper canopy, where light levels are higher than those at the lower canopy; (2) they are physically closer to the developing grains than the leaves, which shorten the path of assimilates transport to developing grains [9]; (3) they have large surface area, e.g. the ear surface areas were 1.1-5.5 folds greater than that of the corresponding flag leaves in wheat and barley [6]; 4) reproductive organs tend to senesce later than leaves, especially under adverse conditions such as drought [10,11]. Due to these reasons, photosynthesis of reproductive organs gained more and more attraction in recent years (see reviews in [12][13][14][15]).
In wheat and barley, ear photosynthesis is a major source of photosynthate for grain filling; a large photosynthetic area and high photosynthetic activity from both the green glumes and awns are major factors underlying higher ear photosynthesis [6,16,17]. Using ear-shading and/or stem-shading methods, ear photosynthesis was estimated to contribute 10% to 65% of the grain yield in different wheat lines and under different growth conditions [16][17][18][19][20]. By analyzing the carbon isotope composition in different organs, it was estimated that ear photosynthesis contributed 47% to 100% of the grain yield in six high-yielding bread wheat lines under wellmanaged conditions [21,22]. In a retrospective study, by using 15 winter wheat cultivars released from 1980 to 2012 in North China Plain, it was reported that the contribution of ear photosynthesis to grain yield doubled (from around 30% to around 60%) during breeding [18]. Although the exact proportion of ear photosynthesis to total grain yield varies, breeding cultivars with higher ear photosynthesis was recognized as a new mechanism to improve wheat yield potential [14,15,23].
In contrast to the significant importance of ear photosynthesis in wheat and barley, rice panicle has been traditionally regarded as a sink organ with negligible photosynthetic contribution to grain yield [24]. For example, Tsuno et al. (1975) estimated that net panicle photosynthesis is about 5% of that of an active photosynthesizing rice leaf [25]; Takeda and Maruta (1956) estimated that panicle photosynthesis contributes 8% of the rice grain yield [26]. Reflecting the notion of rice ear photosynthesis being of little importance to grain filling, some researchers recommended lowering spike position in a canopy as much as possible to increase canopy photosynthesis by minimizing shading leaves by panicles [27].
Is panicle photosynthesis in rice really of little importance for grain filling? More and more evidences are suggesting the opposite. For example, Imaizumi et al. (1990) found that whole-panicle gross amount of photosynthetically assimilated CO 2 was 30% of that in a flag leaf, and spikelets had a similar photosynthetic capability to a flag leaf on the basis of Chl [8]. It has been shown that rice panicle contains higher amount of chloroplast NAD(P) H dehydrogenase-like complex, and maintains a similar or even higher quantum yield of PSII than that of the flag leaf from heading to harvest [28]. There are also evidences showing a potential important role of rice panicle photosynthesis in maintaining grain filling under moderate stress conditions. For example, under moderate wetting and drying irrigation after anthesis, genes that function in photosynthesis and carbohydrate metabolism were upregulated in inferior spikelets, together with a simultaneous increase of grain filling rate and grain weight of the inferior spikelets [29].
One major factor attributing to the inconsistency regarding the importance of photosynthesis of reproductive organs in grain filling is to the lack of robust methods to measure photosynthesis in organs with a complex three-dimensional geometry [12,22,23]. Many indirect methods have been developed to study photosynthesis of reproductive organs [23], e.g. through ear (awn) removal/ shading or application of herbicides [17,27,30], isotope tracing [21], gas exchange measurement of excised spikelets [6,8] and chlorophyll fluorescence measurement for plants cultured in pots [28]. Among these methods, the shading treatments or applications of herbicides may affect other function of treated organs and/or trigger compensation effects of unaffected organs [20,22,[31][32][33][34][35]. The stable isotope discrimination method calculated contributions of different organs to grain yield based on instant carbon isotope in water-soluble fraction in different tissues at mid-grain filling stage and carbon isotope in mature grains [12,21,22]. The accuracy of this isotope discrimination method might be affected by the dynamic environment (e.g. temperature and air humidity) and physiological processes other than photosynthesis (e.g. assimilates transport and ear night respiration) during the whole grain filling season [12,21,22]. Gas exchange measurement on excised ears or spikelets is an intrusive method and it can cause bias due to an alteration of ear natural architecture and posture [27].
Thus, the main objectives of this study are: (1) to develop a direct and non-intrusive method for gas exchange measurement of rice panicles; (2) to explore panicle photosynthetic characteristics and the intercultivar variations of photosynthetic gas exchange parameters; (3) and to study the relationship between panicle photosynthetic gas exchange parameters and grain yield related agronomic traits.

Performance of the panicle photosynthesis P-chamber
The P-chamber is connected to a gas analyzer Li-6400 (Li-Cor Inc., Lincoln, Nebraska, USA) to measure gas exchange rate of a rice panicle (Fig. 1a, b). The panicle is illuminated with LED light source (Fig. 1c, d). The LED light source emits pure white light (light wavelength information is shown in Fig. 1f ). The photon flux density (PFD) is in a range of ± 10% of set value on 87% of the total 30 × 5 cm light field (Fig. 1e). At a room temperature of 25 °C with a flow rate of 700 μmol s −1 , temperature rise in the chamber at PFD of 2000 μmol (photons) m −2 s −1 is 4.3 ± 0.7 °C.
To test the leakage of the P-chamber, we recorded Δ[CO 2 ] (the difference between CO 2 concentration in the P-chamber and the set reference CO 2 concentration in the reference pipe) at different time after closure of the P-chamber. With an initial Δ[CO 2 ] value of around 500 μmol mol −1 , the Δ[CO 2 ] values were stabilized within 8 min under different flow rates and different CO 2 concentrations (Fig. 1g). The final Δ[CO 2 ] values were ≤ 1.6 μmol mol −1 and ≤ 1.3 μmol mol −1 at a flow rate of 500 and 700 μmol s −1 , respectively (Fig. 1g). During a typical measurement of panicle photosynthetic rate and respiratory rate, the light source was turned on before the rice panicle was enclosed, and the photosynthetic rate was recorded after photosynthetic rate reading was stabilized; then the light source was turned off, and the respiratory rate was recorded after respiratory rate reading was stabilized (Fig. 1h). A comprehensive measurement for photosynthetic and respiratory rates of a rice panicle usually took 10 ~ 20 min after the panicle was enclosed in the P-chamber, depended on size and photosynthetic activity of the panicle.

Quantification of rice panicle photosynthesis
We first used the P-chamber to measure the in situ whole-panicle photosynthetic gas exchange parameters 5 days after heading for different rice cultivars when the panicles were still erect or semi-erect. Specifically, whole-panicle net photosynthetic rate (panicle A net ) and dark respiratory rate (panicle R d ) were measured; and whole-panicle gross photosynthetic rate (panicle A gross ) was calculated as a sum of panicle A net and panicle R d . In addition, average spikelet A gross , A net and R d were calculated by dividing panicle A gross , A net and R d by the spikelet number of a panicle.
Notably, substantial inter-cultivar variations in photosynthetic gas exchange parameters among rice cultivars were found, for both values of a whole panicle and values normalized by spikelet number. For example, the spikelet A net and panicle A net were 0.012 ~ 0.051 nmol s −1 and 2.  (Table 1). Moreover, indica type rice cultivars (YLY900, CY1000, SY63 and 9311) had an overall significantly higher A gross , A net and R d than japonica and japonica-indica hybrid type rice cultivars (XS134, YY538 and YY17), both on whole panicle basis and on single spikelet basis (Fig. 2).
In 2016, we further measured area of panicles and area of the corresponding flag leaves, and then quantified A gross , A net and R d on an area basis by dividing panicle A gross , A net and R d by the area of a panicle (Additional file 1: Figure S1; see detailed procedure used to calculate panicle surface area in Materials and Methods). As a result, we found spikelets contributed the majority of the panicle area, whereas branches only accounted for 24 ~ 33% of the panicle area; moreover, total area of a panicle (single side) was 72 ~ 121 cm 2 , which was 20 ~ 124% higher than that of a flag leaf ( Table 2). On an area basis, A net of panicles were very low, i.e., only − 0.6 ~ 2.2 μmol m −2 s −1 , under a PFD of 2000 μmol (photons) m −2 s −1 ; whereas A gross were 2.1 ~ 7.2 μmol m −2 (See figure on next page.) Fig. 1 A custom-built panicle photosynthesis chamber (P-chamber) for panicle gas exchange measurements. a An illustration of how the P-chamber works together with the gas analyzer Li-6400 to measure the photosynthetic and respiratory rates of a rice panicle. b The setting of a P-chamber in a paddy field. c An illustration of design of the P-chamber. d A top-down view of the inner part of the P-chamber. e The homogeneity of PFD in the chamber. f Spectrum of the LED light source. g The difference between the CO 2 concentration in the P-chamber and the set CO 2 concentration ([CO 2 (Table 3).
Given these different representations of gas exchange rates, i.e., on a panicle basis, on a spikelet basis and on an area basis, what's the relations between them? Positive correlations between values on different bases were found for all the three parameters A gross , A net and R d (the Pearson correlation coefficient r = 0.59 ~ 0.99; Fig. 3a-i).
Interestingly, we found that A net per panicle was strongly positively correlated with A net per unit panicle area (the Pearson correlation coefficient r = 0.97; Fig. 3e), in contrast to that for A gross and R d (the Pearson correlation coefficient r = 0.77 and 0.59, respectively; Fig. 3b, h). In addition, spikelet based A gross , A net and R d were closely related to area based A gross , A net and R d (the Pearson correlation coefficient r = 0.97-0.99; Fig. 3c, f, i), as a result of a tight relationship between panicle area and spikelet number (the Pearson correlation coefficient r = 0.95; Fig. 4).

Characterization of rice panicle photosynthesis
To further characterize rice panicle photosynthesis, we measured dynamic changes of panicle A gross and R d for seven rice cultivars during grain filling in 2016 (Fig. 5).  Firstly, panicle A gross and R d were highest at early grain filling stage, then slowly decreased over time. Notably, both A gross and R d could still be detected even 40 days after heading, i.e., panicle photosynthesis and respiration were active throughout the grain-filling season. It is worth mentioning here that the panicle A gross should be interpreted as the maximal photosynthetic potential, as the panicle was held upright during measurement, which was not the natural state after early grain filling stage for some cultivars. We further measured the photosynthetic light and CO 2 response patterns of a panicle using rice cultivar YY17. Although the shapes of both response curves are similar to that of a leaf, there are several differences. Firstly, the apparent CO 2 and light compensation points of a panicle were about 180 μmol mol −1 and 390 μmol m −2 s −1 (Fig. 6), much higher than those of a leaf, which are usually < 100 μmol mol −1 and < 50 μmol m −2 s −1 , respectively. Secondly, the panicle photosynthetic rate was still not saturated at very high air CO 2 concentration (1800 μmol mol −1 ; Fig. 6a), which also differed from that of a typical leaf.

Correlation between panicle photosynthetic gas exchange parameters and grain yield related agronomic traits
Finally, we studied the relationship between panicle photosynthetic gas-exchange parameters measured 5 days after heading (Table 1) and six grain yield related agronomic traits collected at harvest across 2 years, i.e., spikelet number per panicle, panicle length, spikelet density (defined as spikelet number per centimeter of panicle length), 1000-grain weight, grain setting rate and panicle weight (Table 4). Firstly, there were positive correlations between different panicle photosynthetic gas-exchange parameters (Fig. 7). For example, significant positive correlations were found between A gross and A net of a whole panicle (Pearson correlation coefficient r = 0.87, p-value < 0.001), between A gross and R d of a whole panicle (r = 0.94, p-value < 0.0001), between A gross and A net per spikelet (r = 0.93, p-value < 0.0001), and between A gross and R d per spikelet (r = 0.97, p-value < 0.0001). In contrast, both positive and negative correlations were found between grain yield related agronomic traits, e.g. significant negative correlations were found between 1000-grain weight and spikelet number per panicle, but significant positive correlations were found between 1000-grain weight and grain setting rate (Fig. 7). Notably, strong positive correlations were found between A gross , A net , R d per spikelet and grain setting rate, 1000grain weight. In particular, the Pearson correlation coefficient r is 0.93 (p-value < 0.0001) between A gross per spikelet and grain setting rate (Fig. 7). Concurrently, we found that A gross , A net and R d of a whole panicle had a moderate positive correlation with panicle dry weight at harvest (r = 0.39 ~ 0.55); whereas A gross , A net and R d per spikelet were not correlated with panicle dry weight at harvest (r = -0.18 ~ 0.12). In addition, all the six  panicle photosynthetic gas exchange parameters were found to be positively correlated with panicle length (r = 0.6 ~ 0.85), whereas they were negatively correlated with spikelet density (r = -0.82 ~ -0.28). Finally, no correlations were found between A gross , A net , R d of a whole panicle and spikelet number per panicle (r = -0.2 ~ 0.1).

Discussion
Reproductive organs of cereal crops have been shown photosynthetically active, and photosynthesis in ear of wheat and barley has attracted great research interest [12][13][14][15]. Current high-yield rice cultivars usually have much bigger ears with many spikelets compared with wheat or barley. For example, the newly bred rice cultivar YLY900, which have reached record-high yield of 15 tons ha −1 [36], and another even higher yielded rice cultivar CY1000, have an extra-large panicle size of exceeding 350 and 400 spikelets per panicle, respectively ( Table 4). As a result, the green area of a rice panicle can even be larger than a flag leaf (Table 2). Furthermore, considering that the panicles locate at the upper part of a canopy, they can intercept a considerably amount of solar radiation. All these suggest a great role of panicle photosynthesis in rice grain filling. However, compared to leaf photosynthesis, panicle photosynthesis is much less studied, largely due to the methodological limitations [12,22,23]. Therefore, new tools and methods are needed to enable better characterization of rice panicle photosynthesis and provide an accurate evaluation of the impact of panicle photosynthesis on rice grain yield.
Here we reported the development of a new panicle photosynthesis P-chamber, which can be connected to a standard infrared gas analyzer to enable direct measurement of gas exchange parameters. Compared to previously reported cuvettes for rice or wheat ear gas exchange measurement [8,17,22,37], the P-chamber has several new features. First, it is made air-tight with very limited leakage under different CO 2 concentrations and flow rates, which enables accurate and fast measurement of gas exchange rate. Secondly, it has a large cuvette volume for gas exchange measurement of a whole rice panicle in situ. The large cuvette volume enables measurement of panicles of irregular shape. We expect that it can also be used to measure inflorescence of Arabidopsis, wheat, rape and soybean and other plant organs such as stem, fruit, and branch in situ. Moreover, the chamber is opaque and the PFD inside is uniformly distributed and can be precisely controlled, which enables studying response of photosynthesis to PFD.
Using the P-chamber, we studied quantitative characteristics of rice panicle photosynthesis. First of all, to facilitate comparison between tissues, gas exchange rates are normalized. There are three commonly adopted approaches to normalize gas exchange rates of foliar and non-foliar organs, i.e., (1) normalization based on organ number, which is widely used for cereal ears [25,38]; (2) normalization based on area, which is commonly used for leaves [37,39]; and (3) normalization based on chlorophyll content, which is applicable for both foliar and non-foliar organs [8,40]. For ears of cereal crops, it is labor-saving and technically easier to measure gas exchange rate on a whole ear basis or on a spikelet basis rather than on an area basis or chlorophyll basis. However, what's the relationship between gas exchange rates

YLY900
356 ± 24b 28.   Table 1 and Table 4  normalized on different bases? Here we show that gas exchange rate on a spikelet basis is highly correlated to gas exchange rate on an area basis (Pearson correlation coefficient r = 0.97−0.99; Fig. 3). This result indicates that we may use spikelet-based spike photosynthesis rate as a surrogate for the area-based spike photosynthesis rate, as the former one is non-intrusive and technically easier. For the seven rice cultivars used in this study, we have the following equation to convert sipkelet number to panicle area (Fig. 4): Strikingly, A net on a panicle basis is also highly correlated to A net on an area basis for the seven rice cultivars used in this study (Pearson correlation coefficient r = 0.97; Fig. 3). This result indicates that panicle A net is largely determined by A net per unit area rather than its spikelet number or green surface area.
Secondly, we found that as a result of high respiratory rate and large panicle area, A gross of a whole panicle 5 days after heading could be as high as 17 to 54 nmol s  [41][42][43], the A gross of whole flag blades for the seven rice cultivars were 82 to 220 nmol s −1 . This estimation resulted in the ratio between panicle A gross and flag leaf A gross to be 20−38% (Table 5). This result is comparable to a previous report, where A gross of a panicle is 30% of a flag leaf [8]. Concurrently, though panicle photosynthetic activity gradually decreased during grain filling, it still maintained CO 2 uptake activity even 40 days after heading, especially for the newly released super high yield hybrid rice cultivars, YLY900 and CY1000 (Fig. 5a). Maintaining (1) Panicle area = 0.266 * Spikelet number + 9.84 photosynthetic activity of a panicle till the end of rice grain filling has been reported earlier as well [28,40]. The contribution of panicle photosynthesis to rice grain yield cannot be precisely calculated in the current study due to the lack of the light environments for panicle throughout the grain filling season. However, the contribution of panicle photosynthesis to grain filling can be estimated based on measured rates of respiration. Specifically, if we assume a constant respiratory rate during the daytime, i.e., from 6:00 AM to 18:00 PM, the total daytimerespired CO 2 during the whole grain filling season for the seven cultivars is 21-40 mmol (an integration of the Tot. R d in Fig. 2), which is equivalent to 0.6-1.2 g CH 2 O and accounts for 12.1-17.0% of the panicle weight at harvest (Table 4). Namely, by refixation of daytime-respired CO 2 , panicle photosynthesis can apparently contribute 12-17% of the final grain yield. Considering that during most of the grain filling season, panicle maintains a positive carbon uptake (Fig. 5), we conclude that the apparent contribution of panicle photosynthesis to grain filling will be higher than 12%. Combination of the photosynthetic light response curve measured by P-chamber with an elaborate three-dimensional modeling of rice panicles and leaves [44], light environments within a canopy can be calculated and the daily and season-long photosynthetic contribution of a panicle can be precisely quantified.
Thirdly, there are substantial inter-cultivar variations in panicle photosynthetic and respiratory rates 5 days after heading, both on a whole panicle basis and on a spikelet basis (Table 1). Indica cultivars have ~ 402% and ~ 693% higher panicle net photosynthetic rates per panicle and per spikelet than japonica and japonicaindica hybrid cultivars, respectively; meanwhile, indica cultivars have ~ 68% and ~ 83% higher panicle respiratory rates per panicle and per spikelet, respectively (Fig. 2). Overall, indica cultivars show ~ 94% and ~ 122% increase in panicle gross photosynthetic rates on the panicle basis and on the spikelet basis, respectively (Fig. 2). One of the factors contributing to these differences might be panicle morphology. For example, the high-yielding japonica rice cultivars usually have erect and densely-packed panicles [45], while the indica rice cultivars typically have drooping and loosely-packed panicles [46]. Indeed, highly negative correlations between spikelet density and respiratory rate per spikelet (Pearson correlation coefficient r = −0.82; Fig. 7), and between spikelet density and net photosynthetic rate per spikelet were found in this study (Pearson correlation coefficient r = −0.66; Fig. 7). The mechanisms by how spikelet density could influence both panicle photosynthesis and respiration are not clear. Intersubspecies differences in photosynthetic apparatus and

Table 5 Estimated ratios of light saturated gross photosynthetic rates between a panicle and a flag leaf 5 days after heading in seven rice cultivars
Data measured in 2016 (n = 5). A gross of a flag leaf, estimated light saturated gross photosynthetic rate of a whole flag leaf by assuming a constant light saturated photosynthetic rate of 25 μmol m −2 s −1 (the leaf area data was taken from Table 2); A gross of a panicle, light saturated gross photosynthetic rate of a whole panicle (data was taken from Table 1); P/FL A gross , the ratio between light saturated gross photosynthetic rate of a panicle and that of a flag leaf

Name
A gross of a flag leaf (nmol s −1 ) A gross of a panicle (nmol s −1 ) metabolism may also exist, which need to be further studied. Similar to this, there is a significant difference between leaf photosynthetic rates between indica and japanica cultivars, the molecular mechanism of which also awaits clarification [47,48]. Fourthly, rice panicles show differences in photosynthetic response pattern to light intensities and CO 2 concentrations compared to the leaves. Specifically, rice panicle has high photosynthetic light compensation point, and has high apparent CO 2 compensation and saturation points. Consistent with our finding here in rice, wheat ear was also found to have higher CO 2 compensation and saturation points than that of a flag leaf, which might be due to its high respiratory rate and different stomatal conductance response pattern [37].

P/FL
Correlation study between panicle photosynthetic gas exchange parameters and grain yield related agronomic traits further shows a significant role of panicle photosynthesis in rice grain filling. Specifically, we found spikelet gross photosynthetic rates at early grain filling stage were strongly and significantly correlated with grain setting rate (r = 0.93) and 1000-grain weight (r = 0.84; Fig. 7). These results indicate that activity of the spikelet as a source and as a sink may closely linked. It is likely that the activity of the spikelet as a photosynthetic source promotes the activity of the grain as a sink. The detailed mechanism of this "source-sink" correlation is still not clear. Earlier studies on wheat offer a few possibilities: (1) panicle photosynthesis may supply oxygen to the hypoxic regions deep within the developing seed to increase its metabolic activity [49]; (2) the transpiration accompanying panicle photosynthesis not only supports the transfer of nutrients and signal molecules from xylem to spikelets, but also regulates panicle temperature to avoid over-heating [50,51]; (3) panicle photosynthesis can generate assimilatory power, i.e., ATP and reduced ferredoxin, which are required to support glume nitrogen assimilation, a major source of nitrogen in wheat grain [52].
Finally, the method of directly measuring reproductive organ gas exchange rate has a range of potential applications. Firstly, it can help accurately simulate photosynthesis from non-foliar organs and evaluate their role in canopy photosynthesis based on three-dimensional canopy photosynthesis modeling [44,53]. Secondly, as large natural variations of panicle photosynthetic and respiratory rates exist between cultivars, the genetic basis can be explored by measuring panicle gas exchange parameters in rice genetic populations. Thirdly, the diurnal and seasonal panicle respiratory rate change can be measured to trace grain filling dynamics to support identification of new options to improve panicle grain filling patterns [43,54].

Conclusions
Here we report the design and application of a new device, the P-chamber. The features of large cuvette volume, air tight, and programmable light intensity inside the P-chamber enable in situ characterization of photosynthetic gas exchange of a whole plant organ. With the P-chamber, for the first time, we characterized wholepanicle photosynthetic light and CO 2 response patterns, quantified variations of panicle photosynthesis between cultivars and identified correlations between panicle photosynthetic gas exchange parameters and agronomic traits in rice. Further application of the P-chamber will facilitate study on photosynthetic characteristics and contribution of irregular photosynthetic organs, especially the non-foliar photosynthetic organs, e.g. inflorescence of cereals and other plant organs such as stem and branch.

Construction of the custom-designed panicle chamber
To enable direct and non-intrusive measurement of panicle photosynthesis, we built a panicle photosynthesis measurement chamber (P-chamber), which can be used together with infrared gas analyzers to measure photosynthetic and respiratory rates of a whole rice panicle. The chamber body is made of aluminum, with cooling fins and cooling fans on both sides (Fig. 1c). The chamber body has a dimension of 30 × 5 × 5 cm. The air inside the chamber was mixed with two air-mixing fans inside the chamber (F2008ES-05WAV, SHICOH IC FAN, Japan), light source and a PFD control system (Fig. 1c, d). The inner space is uniformly illuminated with 192 light-emitting diodes (LEDs; CREE XML T6, Guangdong Benbon Electrical Co., Ltd) on an 30 × 5 cm Aluminum substrate (Dongxing electronic technology Co., Ltd) for which the PFD can be precisely controlled to span from 0 to 2000 μmol (photons) m −2 s −1 (Fig. 1d). The PFDs in the P-chamber under different input voltages on LEDs were measured with an irradiance sensor (LightScout Light Sensor Reader, LightScout, Spectrum Technologies, Inc. USA). During measurement, the P-chamber is connected to a standard Infra-red gas analyzer Li-6400, and the rice panicle is held upright and fully enclosed in the chamber (Fig. 1a, b). Flow rate and CO 2 concentration of influx air are controlled by LI-6400, and data are recorded and stored in LI-6400. Temperature inside the P-chamber was not controlled during measurement. The ambient air temperature for each measurement was recorded (see in Additional files 1: Table S2 and Table S3).

Measurement of panicle photosynthetic gas exchange parameters
Twenty panicles were randomly selected and labeled at heading for each rice cultivar, which were used to measure photosynthetic gas exchange parameters and yield related agronomic traits. Three and five panicles among these twenty were used to measure whole-panicle net photosynthetic rate and dark respiratory rate in 2015 and 2016, respectively. During each measurement, the CO 2 concentration was set to 400 μmol mol −1 , the air flow rate was set to 700 μmol s −1 . For net photosynthetic rate measurement, the PFD was set to 2000 μmol (photons) m −2 s −1 ; for dark respiratory rate measurement, the PFD was set to 0 μmol m −2 s −1 . Each time when the gas exchange rate readings were stabilized, data were logged after matching the two infrared gas analyzers (IRGAs).
The rice cultivar YY17 was used for photosynthetic light and CO 2 response curve measurements 10 days after heading in 2016. For photosynthetic light response curve measurement, the CO 2 concentration was set to 400 μmol mol −1 , the flow rate was set to 700 μmol s −1 ; and the light levels were manually changed in the following sequence: 2000, 1500, 1000, 700, 500, 300, 200, 100, 50, 25 and 0 μmol (photons) m −2 s −1− during measurement. Panicles were maintained in the cuvette at each light level for about 4 min before photosynthetic rate was stabilized and recorded. For photosynthetic CO 2 response curve measurement, the PFD was set to 2000 μmol (photons) m −2 s −1 , the flow rate was set to 700 μmol s −1 , and the reference CO 2 concentration was set to 400 μmol mol −1 . The panicles were maintained under the above described condition for 15 min. Then, the CO 2 concentrations were manually changed in the following sequence: 400, 300, 150, 100 and 50 μmol mol −1 during measurement. Panicles were maintained in the cuvette at each CO 2 concentration for about 4 min before photosynthetic rate was stabilized and recorded. Next, the CO 2 concentration was set back to 400 μmol m −2 s −1 and maintained at that concentration for 15 min. Finally, the CO 2 concentrations were manually changed in the following sequence: 600, 900,1200, 1500 and 1800 μmol mol −1 . Again, panicles were maintained in the cuvette at each CO 2 concentration for about 4 min before photosynthetic rate was stabilized and recorded.

Calculation of panicle area and flag area
Areas of spikelets and branches in a panicle were calculated to obtain the total area of a panicle. Specifically, the non-degraded green spikelets were firstly detached from the panicle 5 days after heading; then, these spikelets were scanned with an Epson Perfection V300 Photo scanner (Epson, Tokyo, Japan; Additional file 1: Figure  S1a). The areas of spikelets in the scanned images were obtained by image processing using ImageJ (US National Institutes of Health, Bethesda, USA). The flag leaf area was obtained in the same way (Additional file 1: Figure  S1c).
The spikelet-less panicles were then spread out, scanned with an Epson Perfection V300 Photo scanner as well (Epson, Tokyo, Japan; Additional file 1: Figure S1b). The projected area of branches was obtained through image processing using ImageJ (US National Institutes of Health, Bethesda, USA). Branches were assumed as cylinders, and single-sided surface area of all branches on a panicle S was then calculated as: in which s bp is the projected area of branches. Singlesided surface area of a whole panicle was calculated as a sum of the area of branches and the area of all spikelets.