Three sets of introgression lines with a total of 750 lines were provided from a joint IRRI China research project, along with linkage maps. These were planted in the dry season of 2007, and were found to be very mixed. During the growing season, off-type plants were removed in an attempt to achieve uniformity in each plot. Grain was harvested for chalk measurements from several plants in the plot, 10 panicles were randomly selected from the plot for DNA extraction, and a single plant was selected for replanting to ensure the plots were pure in the next season. A plot was generated from each of those single plants in the wet season, panicles were collected to relate panicle architecture to chalk, and grain was collected for measuring chalk. Chalk values vary across the spectrum from quite translucent all the way to very chalky. For some lines, the chalk value was similar for both the wet and dry season; for the rest, the chalk values were higher in the dry season. No lines gave lower chalk in the dry season than in the wet season. When attempting to use the chalk data and the linkage map to find genetic regions that associate with chalk, we discovered that the samples were not just individually mixed, but there were many inconsistencies in the populations. We were therefore unable to associate genotype with phenotype. We have two options: sort out the problem or create our own genetic map. We have opted for the latter. Despite the issues with the labelling of the populations, we have two seasons of data and excellent segregation, independent of grain shape, for both chalk and panicle architecture, and we are creating a minimal genetic map using 40 SSR markers. We intend to use that map and the phenotype data to select a subset of the 750 lines for growing in the glasshouse in the second year and for carrying out detailed phenotyping. SNP markers will be added to the framework SSR map.

Three sets of introgression lines (total of 750) were planted in the dry season of 2008. Ten leaves were randomly selected from each plot for DNA extraction and genotyping to replace the incorrect map from China. The subset of 238 lines was selected based on chalk data and genotype, and a single plant was harvested from each to ensure that plots were pure in the 2008 wet season. Tungro virus was a problem in the 2008 WS (section 8).
For the subset, four pots per line were grown at 29C (day) and 21 C (night) with 70% humidity. After flowering, two pots of each were transferred to one of two either a high or low temperature phytotron. During grain-filling, data was collected for number of primary branches (NPB) and secondary branches (NSB), number of spikelets per panicle (NSP); fertility; grain weight; and chalk. The percentage of grain with chalk (PGWC) of the rice varieties Lemont (LMT) and Teqing (TQ) grown under high temperature were 9% and 45.5%, respectively, which decreased to 0.5% and 20.0% under low temperature. Chalk values of the 238 lines ranged from translucent to very chalky. The average of PGWC of LMT*3/TQ, LMT/TQ, and TQ*3/LMT set that underwent grain-filling in high temperature was 13.23%, 25.15% and 37.25%, respectively; whereas, in low temperature, in the same order, it was 5.13%, 11.42% and 17.09%. Forty-one of the lines had 25% more chalk at high than low temperature, suggesting that high temperature during filling relates to chalk. Fifty-nine lines had similar chalk values in both temperatures, suggesting that chalkiness is also under genetic control. One line gave lower chalk in the high than in the low temperature (P<0.05) and four lines gave chalk values of less than 1% even under high temperature.
Grain weight of seeds from primary branches was higher than from secondary branches, but PGWC was similar. The relationship between NSB, NPB, and PGWC was not significant.
114 ILP (intron length polymorphisms) and 29 SSR markers were used to genotype the 238 lines. The genetic basis of PGWC and chalkiness volume was dissected into main effect QTLs (quantitative trait loci), and the stability of these assessed in LMT*3/TQ and TQ*3/LMT across four seasons (2007DS, 2007WS, 2008DS, and 2008WS). Eighteen chalkiness QTLs were detected on seven chromosomes-several of them were identified in multiple seasons. Eighteen chalk volume QTLs were identified on seven chromosomes. QTLs were in the region of sucrose synthase genes, 1 sucrose transporter gene, 1 pyruvate orthophosphate dikinase gene, 1 heat shock protein gene and 1 glutelin gene; suggesting that these QTLs relate to sucrose transport and metabolism into or within the endosperm. Markers are currently being designed to search for genetic variability in these regions of interest using diverse material.
In order to understand how the expression of these genes is influenced by high temperature and how they affect chalkiness, two mini-subsets of 40 lines (with small differences and with big differences in chalk) were selected for further study. Developing grain has been sampled for gene expression and enzyme activity.
Grain quality data including chalkiness were obtained for the twenty varieties in the OryzaSNP project, with clear variation among the lines identified. Moroberekan was found to have an average chalkiness of 50 relative to 8 for IR 64. Grain from an existing RIL (recombinant inbred lines) population at F7, developed for drought QTL mapping, will be phenotyped and genotyped for chalk. Other populations are being created. A panel of 384 SNPs has been selected from verified OryzaSNP alleles for mapping in TQN/LMT and IR64/Moroberekan. These will be run on the Illumina BeadXPress platform during July.
From the OryzaSNP analyses of patterns of shared introgressions, regions in common to 2 to 5 varieties occurred non-randomly (P < 0.025 by permutation test on 1000 resamples) and in clusters (over 70% of common introgressions). 9% of the genome was covered by these clusters. Testing indicated highly significant association with grain quality. These regions appear to comprise signatures of intense selection during domestication and will be explored in detail for their contribution to chalkiness.
Our hypothesis is that QTLs found for the genes relating to starch and protein synthesis are QTLs that respond to a process that regulates source-sink dynamics, and at the time when such a mechanism leads to decreased substrate supply to the panicle, the major genes of starch and protein synthesis respond. We are testing this hypothesis in the coming months.

Two sets of reciprocal introgression lines (IL) derived from Lemont (LMT) and Teqing (TQ) were evaluated for traits of chalkiness such as percent of grain with chalkiness (PGWC), area of chalky endosperm (ACE) and degree of endosperm chalkiness (DEC) using material grown in the field (2007DS, 2007WS, 2008DS) and phytotron (two temperature regimes). Six lines were observed with low PGWC (<10%) and low DEC (<3%) across all 5 environments, indicating high heritability for chalkiness. One stable QTL for ACE was identified on Chr 4 in all 5 seasons in the TQ set, 3 QTLs for PGWC on Chr 1, 4 and 10 and 1 QTL for DEC on Chr 10 were mapped across 4 seasons in the TQ set. A QTL for percentage of grain with less than 10% chalky area (PCAL10), 10-25% chalky area (PCAL25) and 25-50% chalky area (PCAL50) was observed using the the TQ set from 2008DS and phytotron trials. qPCAL10-5 and qPCAL10-10 with positive additive effects were detected on chromosome 5 and 10, while qPCAL50-5 and qPCAL50-10 with negative effects were mapped on the same regions across all environments respectively, that is, the alleles from LMT on two regions are responsible for low areas of chalkiness. Sixteen of 17 epistatic interactions had the same directional effects as the main QTLs, and 64 of 97 (68.04%) QTLs were located on the marker interval of RM104-RI05813, RM252-RM470, RI00399-RI01948, RM163-RI01877, RI03180-RI03489 and RI01015-RI02123 respectively. Several genes located in these regions regulating substrate supply to the endosperm may have important roles in chalkiness formation. Two main QTL regions overlapped with two genes known to down-regulate starch formation, INV2 and PPDKB on Chr4 and Chr5, respectively. Both genes were responsive to high temperature with 0.62- and 0.43-fold decreases in activity. Another region containing 7 main QTLs involved 9 epistatic interactions and the region contains BT1-1, an ADP-Glc translocator on Chr 5. SUT3, a sucrose transporter, is close to the marker interval RI01015-RI02123, which contains stable QTLs for PGWC, PCAL10 and PCAL50. On the end of Chr1, there is a QTL cluster containing 8 QTLs but no genes related to starch synthesis could be identified in this QTL region. QTL analysis from multi-environmental data from the INQR partners detected 5 regions on Chr 01, 04, 2 on 05 and 10, and was consistent with the results of single environment analyses.
One line, 6461, derived from LMT/TQ4 containing 5 QTLs for chalkiness with negative effects from LMT gives chalk-free grain even when grown in high temperatures. This result suggests markers for mapping these 5 QTLs are useful for MAS.
Two yield trials were conducted during the 2010 DS at IRRI. The yield of 6461 was 27.30% higher than that of IR 64 in one trial, and in the other, yield of 6461 was 6.37 tons/ha, 2.24% higher than a hybrid variety, Mestizo 7. Compared with two other IRRI varieties, IR 123 (high yielding) and IR 60 (high grain quality), the yield of 6461 was higher by 15.50% and 18.82%, respectively. This result implies it is possible to breed high yield and low chalk rice varieties. Line 6461 carries all 5 of the major QTLs. Other low chalk lines carry 1 or 2 of the 5 alleles, and several yielded higher than IR64 and Mestizo. Not all the low-chalk lines were higher yielding that than the checks, but the data suggests that there is no association between yield and low chalk.
QTL NILs have been created, wherein each of the five QTLs has been backcrossed into the same genetic background. These populations will be used to determine the effect of each QTL on chalk, and to fine-map the interesting genes in each region.

Multienvironment test: 240 RILs including 2 parents were grown in 7 countries/regions.
Chalkiness of this population on average was from 14.66% (Colombia) to 33.18%
(Hangzhou, China), which suggests that environment influenced chalkiness dramatically.
We have previously identified lines that show very little chalk in all IRRI seasons, or very
high chalk. These lines showed the same pattern in all the countries. This result implies
that it is possible to develop low chalky varieties via recombination and selection against
high temperature. In this population, we identified five highly significant QTLs, qCH-2,
qCH-4, qCH-5a, qCH-5b, qCH-8, across all the environments.
Single QTL effects: 7 lines were selected and backcrossed 3 times with Teqing, the high
chalky parent to create QTL NILs for each of the five significant QTLs. 818 BC4F1 /
BC5F1 were genotyped and phenotyped. Each plant contains ~3% Lemont . For all the
QTLs , a T-test showed significant differences in chalkiness between the subset with
desirable heterozygous QTL regions and the subset without the desirable region. For
example, the average chalkiness of lines with qCH-5b was 14.578.18, whereas it was
27.217.65 for the other allele at this locus (p<0.001). In the next season, the loci will be
homozygous, and the real effect will be seen.
Other interesting regions: Two additional regions were detected in the QTL NILs that
showed big effects on chalkiness. One is on chromosome 3, close to GS3. Another is on
chromosome 7. One line with the QTL from chr7 and 2, had extremely low chalk.
Gene specific marker development: SSR or Indel markers for 3 candidate genes were
developed. Two markers explain chalkiness variation on BC4F1/BC5F1 population. The
marker on qCH-4 does not show any function in the heterozygous state. All single QTL
effects, other interesting regions and markers will be confirmed in coming seasons with
single substituted segment lines.
Fine mapping populations: Since candidate genes are not available in two QTLs regions,
several lines selected from the BC4F1/BC5F1 were crossed with Teqing to create fine
mapping populations.
Development of lines with low chalk: phenotypic selection was combined with MAS on
QTL regions, of some very low chalky lines with different grain shape were created (Fig
1). It is worthy to note that the percentage of chalky grain of the four lines carrying the
QTLs in different genetic backgrounds is 0.
Yield trial: The yield of low chalk line, 6461, was 7.5 tons/ha, which out-yielded by 10%
IRRI 123, a popular variety at Philippines and a popular check variety in breeding
programs. This was in the 2010 DS. In breeder's trials, the yield of 6461 in 3 seasons was
determined.