Plant materials were obtained from the Institute of Crop Science, Chinese Academy of Agricultural Science in China. A RIL population comprised of 314 F11 individuals was developed from a cross between the maize inbred lines Qi319 (MRDD resistant) and Ye478 (MRDD susceptible) by single-seed descent. The MRDD-resistant line Qi319, which belongs to the PB heterotic group was originally developed from the US hybrid 78599, and the MRDD-susceptible line Ye478 was developed from the U8112 × 5003 cross. Using the MRDD-resistant line Qi319 as the donor and the MRDD-susceptible line Ye478 as the recurrent parent, 200 CSSLs were developed using a combination of crossing, selfing, and molecular MAS. Detection of SSR markers was performed as described by Wang et al . The lengths of substituted chromosomal segments were assessed using graphical genotypes . Detailed methods and processes we followed for developing CSSLs were described in a previous study in rice . We characterized the 2 to 4 introgression segments in each CSSL in our study according to the physical positions and genotypes of 201 SSRs. The 201 SSRs were evenly distributed along all 10 maize chromosomes at an average marker interval of 9.94 Mb. Both the parental lines and RILs were evaluated for MRDD resistance in two fields including one in Xinxiang (35.05°N, 113.96°E), Henan Province, China in 2016 and one in Xuzhou (34.79°N, 116.57°E), Jiangsu Province, China in 2015 and 2016. The Xuzhou growing area experiences severe outbreaks maize rough dwarf disease. All of the plant materials including the 314 RILs and two parental lines were arranged in the fields in randomized incomplete blocks with two replications per location. In each block, the parents Qi319 and Ye478 were planted as MRDD-resistant and -susceptible controls, respectively. The plots for each line consisted of approximately 17 plants in individual 4-m rows spaced 0.6 m apart at a planting density of 60,000 plants/ha. Standard agricultural management practices for maize were followed during each growing season at each location.
Evaluation of plants for MRDD symptoms
To evaluate the effects of MRDD, plants were grown in three locations including Xinxiang (35.05°N, 113.96°E) in Henan Province, Xuzhou (34.79°N, 116.57°E) in Jiangsu Province, and Jining in Shandong Province, which allowed MRDD infection of plants to occur under natural conditions. All fields were planted on May 16 each year to coincide with SBPH infestations when RBSDV, which causes MRDD, is transmitted. At the R6 maturity stage, we visually assessed the MRDD resistance of all plants and assigned disease scores on a scale from 0 to 4, in which plants scored 0 are highly resistant and those scored 4 are highly susceptible to MRDD . We defined the DSI  as: DSI (%) = S (disease rating score number of plants at each score)/maximum disease rating score total number of plants rated in the line) ´100. The phenotypic data for all plants across different replicates were assessed independently. In order to verify whether plant materials were resistant to virus-free adult planthopper, 100 virus-free adult planthoppers were sealed in boxes with Qi319, Ye478, and F1 plants at the V3 stage, and the survival rate of virus-free adult planthoppers was assessed daily from the first day to the seventh day (with six plants per box and five replications). Multiple comparisons using Student’s t-test were used to determine whether any differences in mean survival rate for virus-free SBPH were significant.
Linkage map construction and QTL detection
The 314 RILs were genotyped using a GBS approach on an Illumina HiSeq2500 platform. A high-density genetic map was constructed from a total of 88,268 high-quality SNPs with 4183 bin markers. The map of the RIL population comprised a total genetic distance of 1545.65 cM covering all 10 maize chromosomes with an average physical distance between adjacent markers of ~0.51 M. This detailed genetic map has been described in a previous study . Phenotypic data for resistance to MRDD were collected from the population of 314 RILs during field experiments conducted from 2015 to 2016 in Xuzhou and Xinxiang. A linkage map was constructed by inclusive composite interval mapping (ICIM) using QTL IciMapping software 4.0  and analyzed together with phenotypic information to identify QTL for resistance to MRDD. The positive and negative signs of the estimates indicated whether resistance effects for QTL with additive effects were inherited from Ye478 or Qi319, respectively. Taking each location as an environment and for each of the datasets (2015, 2016, and BLUP), the significance threshold for identifying a putative QTL was set at a logarithm of odds (LOD) score > 3 with 1000 permutations at P < 0.05 .
Validation of qMrdd2 using the CSSL populations
To validate the effect of qMrdd2, eight CSSLs that covered all of chromosome 2 were developed with Ye478 as the recurrent parent and Qi319 as the donor of the randomized MRDD resistance allele tracked using a combination of crossing, backcrossing and molecular MAS (BC5F2) with average background recovery rates from 91.45 % to 99.62 %. In 2015 and 2016, the MRDD resistance of parental lines and CSSL haplotypes I through VIII were evaluated in Xuzhou and Xinxiang. Field designs were arranged in randomized incomplete blocks with three replications per location and the parents Qi319 and Ye478 were planted in each block as resistant and susceptible controls, respectively. Approximately 17 plants from each line were grown in single 4-m rows spaced 0.25 m apart. Student’s t-test was used to perform multiple comparisons for the MRDD resistance of each genotype.
Genotyping and marker development
Genomic DNA was extracted from leaves of plants at the five-leaf stage using a CTAB procedure following the protocol of Murray and Thompson  with modifications. The quality and quantity of DNA samples used for marker genotyping was assessed by evaluating DNA samples on 1.0% agarose gels and by measuring absorbances using a spectrophotometer (Nanodrop 2000, Thermo Scientific, US). We obtained the SSR primer sequences for our study from the MaizeGDB (http://www.maizegdb.org/). InDel markers were developed from 30×genome sequence data for the resistant parent Qi319 and susceptible parent Ye478 . These InDels were designated with the prefix RD and screened for polymorphisms between Qi319 and Ye478. SSR or InDel primers used for genotyping of plants were synthesized by AuGCT Biotechnology Co. Ltd., China. Each PCR reaction mixture contained 6.8 µL double-distilled water, 1.2 µL 10× Buffer, 0.5 µL dNTPs (2.5 mM), 0.15 µL each primer (0.01 nmol/µL), 0.2 µL Taq DNA polymerase (5 U/µL), and 1 µL template DNA in a 10-µL total volume. The touchdown PCR program for amplifying these markers included an initial denaturing step at 94 °C for 4 min, followed by 10 cycles of 30 s at 95 °C, 30 s at 65 °C, and 30 s at 72 °C, with the annealing temperature decreasing by 1 °C per cycle; followed by 30 cycles of 30 s at 94 °C, 30 s at 55 °C, and 30 s at 72 °C; and ending by extending for 5 min at 72 °C. The PCR products were then electrophoretically separated on 8% polyacrylamide gels in 1× TBE buffer that were then silver stained for visualization of PCR products.
Fine-mapping of qMrdd2
We carried out the recombinant-derived progeny tests to fine-map qMrdd2 (Fig. 1) . Based on the QTL region mapped using RILs and CSSLs, CSSL-31 (haplotype II) was crossed as a male parent with Ye478 to produce an F1 in Yunnan in 2015. The 25 F1 progeny were then self-pollinated to produce the F2 and crossed with Ye478 to produce the BC1F1 in a winter nursery in Hainan in 2015. In the summer of 2016, 850 F2 and 1000 BC1F1 were genotyped, and new recombinants mapped within the region containing the QTL were self-pollinated to produce F3 and BC1F2 progenies at Jining and Xinxiang under natural infection conditions. In the winter of 2016, 6000 F3 and 6000 BC1F2 progeny were planted in the winter nursery and screened for new recombinants. F3 and BC1F2 recombinants with a homozygous Qi319 genotype at the flanking marker on one side and a homozygous Ye478 genotype of on the other side were then selfed. At the same time, we selfed the heterozygous recombinant BC1F2 progenies to produce segregating BC1F3 populations heterozygous at the flanking marker on one side and homozygous at the flanking marker on the other side. BC1F3 recombinants and F3 recombinants were then classified into different haplotypes by developing markers. These segregating populations and homozygous recombinants were then used to fine map qMrdd2.
In the summer of 2017, we detected differences in DSI under natural infection in Jining, Xinxiang, and Xuzhou between the homozygous families derived from recombinants and Ye478 using Student’s t-test in SAS version 9.2. DSIs differing significantly (P < 0.05) between homozygous recombinant-derived families and Ye478 indicated that qMrdd2 was located within a homozygous Qi319/Qi319 segment, whereas DSIs differing insignificantly (P ≥ 0.05) indicated that qMrdd2 was located within a homozygous Ye478/Ye478 segment. Each homozygous recombinant and Ye478 were grown in plots of approximately 17 plants in 4-m rows spaced 0.6 m apart with three replications per location. At the same time, 150-300 kernels randomly selected from each plot representing the progeny of diverse types of BC1F2 recombinants were planted to evaluate for MRDD resistance under natural inoculation conditions in Jining, Xinxiang, and Xuzhou. Each BC1F2 recombinant could be categorized as carrying one of two possible segments, heterozygous Qi319/Ye478 or homozygous Qi319/Qi319, flanking the recombination breakpoint. Individuals from the selfed BC1F2 recombinant progeny were categorized into one of three possible genotypes in the qMrdd2 region: homozygous Qi319/Qi319, homozygous Ye478/Ye478, or heterozygous Qi319/Ye478. One-way ANOVA was used to compare the DSIs of these three genotypic classes in SAS version 9.2 (SAS Inc., Cary, NC, US, 2009). The DSIs of the three genotypic classes differing significantly (P < 0.05) indicated that the MRDD resistance gene was located within a heterozygous region, whereas the DSIs of three genotypic classes differing insignificantly (P ≥ 0.05) indicated that the MRDD resistance gene was located within a homozygous segment.
Analysis of phenotypic data
As described above, the disease response phenotypes of all recombinant-derived progenies were assessed in terms of DSI (See above for calculation of DSI). All of the genotypic and phenotypic datasets were calculated using Microsoft Excel 2010 software. We estimated the broad-sense heritability (H2) of MRDD resistance across three environments according to Knapp et al. . We calculated heritability as: H2 = δ2g /(δ2g + δ2ge /e + δ2/er), where δ2g is the genetic variance, δ2ge is the genotype × environment interaction, δ2 is error variance, e is the number of environments, and r is the number of replications per environment. The estimates for δ2g, δ2ge, and δ2 were calculated with analysis of variance (ANOVA) using PROC MLM, the Mixed Linear Model procedure, in Statistical Analysis System (SAS) software version 9.2 [SAS Inc., Cary, NC, US, 2009].