Self-seed from two heterozygous BC5S1 individuals and two heterozygous BC5S2 individuals were screened for recombinants via marker-assisted selection . Individuals that contained recombination events within the chromosome 9 fine-mapped stm9 region were selected, grown to maturity, and allowed to self-pollinate to produce seed of fully homozygous individual sub-NILs in the BC5S2 or BC5S3 generation . Recombinant homozygous individual sub-NILs were allowed to self-pollinate to generate ample seeds for replicated experiments. Phenotyping experiments were performed with one representative line from each recombinant class. All plant materials were grown in greenhouses at UC Davis. Seeds were planted in 73-cell flats containing soil media. Flats were watered daily, and plants were fertilized with a 10:30:20 NPK solution once a week. Greenhouses containing plants in flats, pots, and hydroponic tanks were maintained at ambient conditions of 25–37 °C with 55–80 % relative humidity during the day, and 18–25 °C with 20–55 % relative humidity at night. Plants from which seed was to be collected were transferred at the 4th to 5th true leaf stage to individual 8-L pots filled with soil media,vertical farming and grown to maturity to obtain seed.SNP markers were identified specifically for our project or converted from publicly available S. lycopersicum markers . Three markers used by Goodstal et al. were included: two PCR-based Cleaved Amplified Polymorphic markers, T1670 and T0532, and one Sequence Characterized Amplified Region marker, T1673. We used two additional markers from the Tomato-EXPEN 2000 linkage map .

To convert the above-listed markers to SNP markers, DNA from S. lycopersicum cv. T5 and the inter specific F1 hybrid was amplified using Phusion High-Fidelity DNA Polymerase with recommended reagents and cycling conditions . PCR products were sequenced by the UC Davis CBS DNA Sequencing Facility using Sanger sequencing on an ABI 3730 Capillary Electrophoresis Genetic Analyzer. PCR product sequences from S. lycopersicum cv. T5 and the interspecific F1 hybrid were compared for each marker set using CLC Sequence Viewer 6 to identify SNPs . During the course of our marker development, a pre-publication version of the cultivated tomato reference genome sequence became available. Therefore, we designed additional markers from this sequence . The markers obtained from Goodstal et al. and the Tomato EXPEN 2000 linkage map were mapped to the scaffold sequence using the BLAST tool on the Sol Genomics Network website , and new primers were designed in regions between the mapped markers. The new primers were used to amplify and sequence the targeted regions between the BLASTed markers. The sequences obtained were aligned as described previously. Of nine primer sets tested, two amplified consistently and exhibited polymorphisms suitable for the development of SNP markers . As the preliminary genomic annotation provided by the International Tomato Annotation Group became publicly available, predicted gene sequences in our target region were BLASTed against the latest version of the genome sequence to identify single copy genes. Primers were designed for those single copy genes that mapped to scaffolds of the chromosome 9 region containing fine-mapped QTL stm9. Scaffolds containing QTL stm9 were identified using the markers that were already developed for our project. Three gene sequences amplified consistently and had polymorphisms suitable for the development of SNP markers. All three predicted genes mapped to scaffold 06070 of ITAG version 1.00 . The SNP markers developed from these predicted gene sequences were numbered according to the gene model from which they were designed: H348, H358, and H307 . Using the markers we developed for chromosome 9, a multiplexed SNP genotyping assay was designed using Sequenom’s MassARRAY Assay Design 3.1 Software.

Software presets for Single Base Extension High Multiplexing were used, with one SNP per marker for TG18, T1670, and T1673 and two SNPs per marker for T0532, At5g11560, H9, and H14. As additional markers were developed, they were added using the software’s Superplex Replex mode for a multiplex assay with an additional SNP for H358 and two SNPs for H348 .After 2 weeks of growth in flats containing soil media , the roots of two seedlings of each sub-NIL or control were carefully washed free of soil media in deionized water and transferred to a hydroponic growth tank set at 20 °C containing a modified Hoagland solution at 20 % of full strength.Plants were grown in the hydroponic tank for 1 week under ambient illumination in the greenhouse, with constant aeration and circulation of the nutrient solution. Subsequently, the plants were randomized and transferred to a separate refrigerated hydroponic tank containing fresh modified Hoagland solution at 20 % strength. Plants were acclimated overnight at a solution temperature of 20 °C. The following morning supplementary lighting was provided by one 1000 W metal halide lamp starting at 7:00 am and used throughout the experiment to maintain a light level above 1000 µmol m−2 s −2 PAR. The tank solution was maintained at 20 °C for 1 h after the supplementary lighting was turned on, and then the tank temperature was decreased to 6 °C. The tank was held at 6 °C for 2 h prior to phenotyping. Each experiment was conducted as a Randomized Complete Block Design and repeated in two seasons , with days as blocks and two replicate plants of each genotype per block. In addition to the two individual plants per recombinant sub-NIL, each replication of the experiment included controls: four plants of chilling-sensitive S. lycopersicum cv. T5, and two plants of a chilling-tolerant NIL, 03GH1322 that was also used as a tolerant control by Goodstal et al. . Six repetitions of the experiment that were conducted from May 9th to June 13th comprise the Spring data set, and four repetitions of the experiment that were carried out from October 7th to October 20th comprise the Fall data set. Plants were individually phenotyped for shoot turgor maintenance under root chilling according to the rating scale described in Goodstal et al. . Briefly, shoot turgor was scored for each plant on a scale of 0–3, with a stmscore of 0 denoting maintenance of shoot turgor, and a stmscore of 3 denoting severe loss of shoot turgor .The linkage map for the S. habrochaites introgressed chromosome 9 region was constructed with JoinMap 4.0 . The Kosambi function with a 4-LOD significance threshold was used to construct the map; the resulting marker grouping was maintained at LOD 10. The population used for map construction included all 2862 BC5S2 and BC5S3 individuals that were genotyped . Since no recombinants were identified between markers T1670 and TG18, or between markers H348 and At5g11560, markers TG18 and At5g11560 were not included in subsequent analyses.

QTL mapping of stm score was conducted for each season using sub-NIL LSmeans obtained from ANOVA. QTL mapping was performed with WinQTLCartographer2.5 using composite interval mapping Model 6 with forward and backward regression. Due to the relatively small genetic distances between markers, a walk speed of 0.5 cM and a window size of 0.5 cM were used. One thousand permutations were performed to obtain a trait-specific permuted significance threshold at P = 0.05; a significant QTL was declared when the LOD value exceeded the permuted threshold.We used the publicly available S. lycopersicum reference genome sequence version SL2.50 to obtain estimates of physical size and gene content in the QTL stm9 Stm score = Genotype region because assembled S. habrochaites whole genome sequence is not available. The S. habrochaites genome is 1.5 × the size of the S. lycopersicum genome as determined by flow cytometry . The location of the QTL stm9 region in S. lycopersicum was determined on the Sol Genomics Network S. lycopersicum reference genome version SL2.50 browser using BLAST . The BLAST position of the midpoint of each S. lycopersicum cv. T5 sequence that we obtained from PCR product sequencing during SNP marker development was defined as the physical location of that marker in the S. lycopersicum reference genome SL2.50. The physical positions of the markers on SL2.50 were used to compare the S. lycopersicum physical map to the genetic map for QTL stm9. Kilobases per 1 cM were calculated for the QTL stm9 region from the flanking markers H9 to T1673, as well as for each internal marker to-marker interval . Once the physical size of the QTL stm9 region in S. lycopersicum was estimated, the number and identity of annotated genes were obtained from ITAG release 2.40 . Gene name, location, protein sequence analysis,vertical garden hydroponic and classification information InterPro and Gene Ontology annotations were downloaded from the SGN genome browser Gene Track. Genes were considered within the QTL region if any exonic sequence of a given gene fell within the QTL consensus region identified in both the Spring and Fall datasets, and defined by the flanking marker interval H9 to T1673. Genes were categorized according to function and/or type when GO terms and IPRO definitions were available .From a total of 2862 BC5S2 and BCsS3 individuals genotyped, 52 individual recombinant sub-NILs that represented 18 unique recombinant classes were identified, with at least two independent recombinant sub-NILs identified in each recombinant class. Recombinant sub-NILs were subjected to replicated experiments in hydroponic tanks, and stmscore data were obtained for Fall and Spring. A full model ANOVA of stmscore detected a highly significant Genotype × Season interaction , therefore separate ANOVAs for the Spring and Fall 2011 data sets were performed . In both seasons, Levene’s test was significant for Genotype. Consequently, the data were weighted by the reciprocal of the variance for Genotypeto meet the assumptions of homogeneity of variance, and Proc MIXED was used to analyze each data set separately. Genotype was highly significant in both seasons . Within each data set, genotype means were significantly different and several groupings of means were identified . Recombinant sub-NILs were classified into two main groups: susceptible or tolerant, according to a mean stmscore less than 1.0 in the Spring and Fall or mean stmscore greater than or equal to 1.0 , respectively . The Fall data set resulted in distinct groupings between susceptible and tolerant sub-NILs, with no overlap between the two groups . The Spring data set exhibited a more gradual separation of means, with three sub-NILs with a mean stmscore of just under 1.0, and one sub-NIL with a mean stmscore just over 1.0 .

While sub-NIL rank changes within the tolerant and susceptible groups occurred between the two data sets, rank changes did not result in the reassignment of any sub-NIL between the susceptible and tolerant groups, with the exception of sub-NILs C3 and C13. Sub-NILs C3 and C13 were grouped as tolerant in the Spring dataset , and susceptible in the Fall dataset . Not only did these lines score as susceptible in the Fall dataset, they had the highest mean stmscore of any recombinant subNIL in the Fall dataset, and therefore were designated as susceptible .The linkage map of the introgressed S. habrochaites chromosome 9 region included eight polymorphic markers that spanned 1.28 cM . The average distance between markers was 0.18 cM, with the largest interval between markers T1673 and H14, and the smallest between markers H9–H358 and H358–H348. A single significant QTL was detected between markers H9 and T1673 with both the Spring and Fall datasets . The QTL LOD peak was at marker H348 for both datasets. The 1-LOD and 2-LOD intervals for both data sets were defined by markers H358–T1673, with the exception of the 2-LOD interval for Spring. In this case the left-most bracketing marker was H9, not H358. There was no evidence of additional significant QTL or of fractionation of QTL stm9 as a consequence of higher-resolution mapping. The QTL peak marker, H348, was also the only marker with the S. habrochaites allele across all recombinant sub-NILs consistently classified as tolerant . Furthermore, sub-NILs with the S. habrochaites allele at marker H348 and at one of the flanking markers scored as tolerant. Our results strongly support the close linkage of the S. habrochaites gene or polymorphisms responsible for maintenance of shoot turgor under root chilling to marker H348.The chromosomal location of stm9 in our study agrees with Goodstal et al. who fine-mapped stm9 to marker interval T1670–T1673 . We refined the location of stm9 to marker interval H358–T1673, a genetic distance of 0.32 cM. Our data suggests that the gene or polymorphisms controlling the tolerance phenotype are located close to marker H348 and within the marker interval H358–T1673.