Combining high-throughput micro-CT-RGB phenotyping and genome-wide association study to dissect the genetic architecture of tiller growth in rice

Abstract

Manual phenotyping of rice tillers is time consuming and labor intensive, and lags behind the rapid development of rice functional genomics. Thus, automated, non-destructive methods of phenotyping rice tiller traits at a high spatial resolution and high throughput for large-scale assessment of rice accessions are urgently needed. In this study, we developed a high-throughput micro-CT-RGB imaging system to non-destructively extract 739 traits from 234 rice accessions at nine time points. We could explain 30% of the grain yield variance from two tiller traits assessed in the early growth stages. A total of 402 significantly associated loci were identified by genome-wide association study, and dynamic and static genetic components were found across the nine time points. A major locus associated with tiller angle was detected at time point 9, which contained a major gene, TAC1. Significant variants associated with tiller angle were enriched in the 3'-untranslated region of TAC1. Three haplotypes for the gene were found, and rice accessions containing haplotype H3 displayed much smaller tiller angles. Further, we found two loci containing associations with both vigor-related traits identified by high-throughput micro-CT-RGB imaging and yield. The superior alleles would be beneficial for breeding for high yield and dense planting.

Fig. 1.

Main components and configuration of HCR. (A) Prototype of the HCR system. PLC, Programmable logic controller. (B) Layout of the inspection unit. (C) The cone angle of the focal spot of the X-ray source is 33º, and the distance between the X-ray source and the detector (D_sd) is 634 mm, thus the projection diameter of the X-ray source (PD_source) is 376 mm. (D) The distance between the focal spot of the X-ray source and the X-ray flat panel detector (D_sd), the distance between the X-ray source and the rotation center (D_sc), and the vertical length of the X-ray flat panel detector (VL) are 634 mm, 484 mm, and 195 mm, respectively, thus the vertical length of the field of view (VL_FOV) is 149 mm. (E) The distance between focal spot of the X-ray source and the X-ray flat panel detector (D_sd), the distance between the X-ray source and the rotation center (D_sc), and the horizontal length of the X-ray flat panel detector (HL) are 634 mm, 484 mm, and 244 mm, respectively, thus the horizontal length of field of view (HL_FOV) is 186 mm. The spatial resolution can be calculated as 97 μm×97 μm. (F) For the RGB camera, the object distance is 1520 mm, the image distance is 8 mm, and the image plane area is 7.09 mm×8.46 mm, thus the field of view is 1607 mm (vertical, SV_V)×1347 mm (horizontal, SV_H).

Fig. 2.

Image analysis pipeline of HCR. (A, B) As the rice sample rotated, 20 color images and 380 X-ray projected images at different angles were acquired synchronously; (C) one row of X-ray projected images at the same height as the 380 X-ray projected images, which formed a sinogram covering 380 orientations, was selected (step 0.6°, total angle 0.6°×380, ~228°); (D) a conventional filtered back-projection (FBP) algorithm was applied to obtain a reconstructed transverse section image of the rice tillers; (E, F) after image segmentation and the removal of small particles, the tiller number, size, and shape were counted; (G) two transverse tiller images were reconstructed at two different heights (row 600 and row 650), and the tiller angle was calculated using the spatial location of the central point of the rice tiller images; (H) 75 phenotypic traits (including plant color, plant height, digital biomass, plant compactness, tiller number, shape, area, and angle) were extracted and stored in a database, which also included RGB images and micro-CT images.

Fig. 3.

Comparison of automatic digital measurements versus manual measurements. (A–C) Scatter plots of manual measurements versus automatic measurements with the micro-CT unit for (A) tiller number, (B) tiller diameter, and (C) stem wall thickness. (D) Absolute percentage error of automatic measurements versus manual measurements of eight round plastic pipes used to represent rice tillers. To evaluate the accuracy and repeatability of the micro-CT unit, eight round plastic pipes with fixed size were measured manually and digitally 10 times by the micro-CT unit.

Fig. 4.

Screening the dynamic process of rice growth at the tillering and elongation stages. (A–I) RGB images and reconstructed CT images at nine sequential growth time points of rice accession (C055, Sanbaili). The red circles indicate dynamic tillering and elongation processes. (J) Changes in total tiller area and the first derivative of total tiller area to reflect the dynamic tiller growth. Error bars represent the SD between the accessions. (K) Distribution of the number of samples (accessions) in the initial elongation stage among the nine time points. The blue arrows indicate the time of most active tiller growth.

Fig. 5.

Heatmaps of (A) total tiller area (TTA) and (B) total projected area (TPA) of 234 individual rice plants at nine different time points. (C) Comparison of actual total tiller area (blue line) and predicted total tiller area (red line). (D) Comparison of actual total projected area (blue line) and predicted total projected area (red line). Error bars represent the SE of the TTA or TPA of the 234 samples at each time point.

Fig. 6.

Prediction of grain yield and shoot dry weight. (A) Modeling accuracy change for grain yield at nine time points. (B) Modeling accuracy change for shoot dry weight at nine time points. (C) Scatter plot of tiller number (TN) versus grain yield at the fifth time point. (D) Scatter plot of total tiller area (TTA) versus grain yield at time point 5. (E, F) Scatter plots showing the relationship between the actual and the estimated grain yield using the formula predicted by (E) 2 traits and (F) 10 traits. a, b, c, d, e, f, g, h, i and j represent TTA_5, MEANTA_8, THR_4, FDIC_7, MAXTAPR_7, FDIC_8, SDTTA_5, TN_3, MEANTAPR_2, and MAXTTA_2, respectively (see Table 1 for definitions).

Fig. 7.

GWAS results of traits at nine time points measured by HCR. (A) Venn diagram showing the number of associated loci at time points 1, 5, and 9. (B) Frequency and distribution of loci associated with traits at the nine time points (T1–T9). (C) GWAS plots of mean of tiller angles (MEANTA) at the nine time points. The strongest association signal on chromosome 9 corresponded to the locus with highest detection frequency (indicated with an asterisk in B).

Fig. 8.

Association analyses of TAC1 and MEANTA_3. (A) Local Manhattan plots and heatmap showing the level of linkage disequilibrium of the TAC1 region. (B) Haplotype analyses of TAC1. The P-value was calculated by ANOVA. Multiple-haplotype comparison was conducted using the least significant difference method; different letters above the boxplots indicate significant differences. (C) Images of two representative varieties, Minghui63 (from the H2 haplotype group) and Zhenshan97 (from the H3 haplotype group).

Fig. 9.

Co-localized loci associated with traits measured by HCR and yield. (A) The locus on chromosome 4 associated with AGRTTA_5 measured by micro-CT (upper panel) and yield (lower panel). (B) The locus on chromosome 6 associated with AGRTPA_4 measured by RGB (upper panel) and yield (lower panel).

Fig. 10.

Relationship between tiller angle change at the late tillering stage and yield. (A) Differences in tiller angle change between groups of 50 rice accessions with higher yield and lower yield. Error bars indicate the SD of the mean value of all tiller angles of the 50 accessions per group. (B) Difference of tiller angle change during the late tillering stages (calculated as MEANTA at time point 8 minus MEANTA at time point 5) between the higher-yield and lower-yield groups.

Fig. 11.

The implications of tiller senescence for yield and drought resistance. Effects of higher and lower tiller senescence on (A) grain yield, (B) green projected area ratio (stay-green trait), (C) total projected area/bounding rectangle area ratio (leaf-rolling trait), and (D) leaf water content. (E–G) Reconstructed tiller images of one rice accession (Zhonghua 11) at three time points: (E) day 0, before drought stress; (F) day 10, after drought stress; (G) day 12, after rewatering.

Fig. 12.

Reconstructed transverse section CT image of a rice tiller at high spatial resolution (30 μm) (left) and transverse section photographic image of the rice tiller after sectioning (right).