For genomic DNA extraction, four 50-mL Falcon centrifuge tubes were each filled with 10 mL of extraction buffer (2% CTAB, 1.4 M NaCl, 20 mM ethylenediaminetetraacetic acid [EDTA], 100 mM Tris-HCl at pH 8.0, and 0.2% beta-mercaptoethanol. The tubes were warmed to 60 °C in a water bath. Apre-cooled mortar was filled with liquid nitrogen to a depth of 3 cm. Sterile sand was added to a depth of 0.5 cm. Five young iris leaves, approximately 20 cm in length, were cut from the plant and immediately cut into 2-cm lengths and submerged in liquid nitrogen. They were then ground into a fine powder, with additional liquid nitrogen carefully added as needed. Approximately 25% of the iris–sand powder was added to each warm extraction buffer tube and mixed by inversion 3 to 4 times. The mixture was incubated at 60 °C for 30 min and mixed by inversion every 5 to 10 min. Then, 10 mL of phenol:chloroform:isoamyl alcohol at a ratio of 25:24:1, was added and gently mixed by continuous inversion for 1 min. The sample was centrifuged at 6000 × g for 10 min to separate the phases. The upper phase was carefully removed and transferred to a 50-mL phase-lock gel tube (Eppendorf), and 10 mL of chloroform:isoamyl alcohol at a ratio of 24:1, was added and gently mixed by inversion for 1 min. The sample was centrifuged at 6000 × g for 10 min, after which time the aqueous phase above the phase-lock wax was removed and transferred to a fresh tube. An equal volume of isopropanol was added to the tube, and the sample was mixed by inversion until a gelatinous mixture of nucleic acids formed. This nucleic acid mixture was removed with a glass rod and washed three times with ice-cold 70% ethanol. The nucleic acid was allowed to air dry and then dissolved overnight in Tris–EDTA buffer. The sample was assigned the Novartis tracking ID AS_SAM_17_03QT. Library preparation used continuous long-read methods for genomic DNA sequencing for the Sequel 1 instrument, as per the manufacturer (PacBio)’s instructions. For optical mapping, a frozen sample of the Iris pallida Lam. leaf (GSM-AAB282) was shipped on dry ice to Bionano Genomics (San Diego, CA, USA), which sequenced the leaf genome to generate a genomic optical map.

A summary of sequencing data for this study is listed in Table 1. A total of 236 single-molecule real-time (SMRT) cells were used to produce 824,109,057,251 bp of genomic sequence data. The average length and N50 values for the PacBio subreads were 7,422 and 17,373 bp, respectively.

Iris tissue samples (leaf and rhizome) were ground in liquid nitrogen, and RNA was extracted using the RNeasy Plant Mini Kit (Sigma Aldrich) and QIAshredder spin columns (Qiagen). The RNA samples were then treated with the TURBO DNA-free Kit (Thermo Fisher Scientific). Library preparation was carried out according to the procedures for Isoform Sequencing (Iso-Seq) using the Clontech SMARTer PCR cDNA Synthesis Kit with a BluePippin DNA Size Selection system. The rhizome sample of Iris pallida Lam. was sequenced using 10 SMRT cells and generated 74,815,065,674 bp of subread sequences, while the leaf sample was sequenced using 11 SMRT cells and generated 45,352,970,323 bp. Each of these datasets was processed using the PacBio Circular Consensus Sequencing (CCS) algorithm v5.1.0 (RRID:SCR_021174) of SMRT Link using the ccs2 pipeline named sa3_ds_ccs.

Falcon_unzip [16] was used to assemble the PacBio long read dataset. We used the Conda channels defaults, bioconda, and conda-forge to install pb-assembly, pbmm2, and genomicconsensus, as of July 24, 2019. Falcon-kit v1.4.2 (RRID:SCR_016089) and falcon-unzip v1.3.3 modules were present in our execution of Falcon_unzip. Falcon_unzip produced two assembly files – the primary contigs and the haplotigs – which are contigs representing variant genomic sequences that are similar but not identical to the primary contigs.

A second genomic DNA sample was shipped to Bionano Genomics to generate genomic optical maps. The resulting optical maps were used to scaffold genome assembly using the HybridScaffolding pipeline in the Bionano Genomics Solve package v3.2.1_04122018. Then, the PacBio Arrow algorithm implemented in the sl_resequencing2 pipeline present in SMRT-Link-7.0.1 was used to polish the Bionano Genomics hybrid assembly using original PacBio reads. The total elapsed time was 309 h (final output folder named arrow_iris_20191208). Telomeres were predicted using the FindTelomeres Python script from Jana Sperschneider [17] (RRID:SCR_024403).

The lima program from PacBio v.1.6.1 (commit v1.6.1-1-g77bd658) was run to identify all CCS sequences with the expected 3′ and 5′ sequences. Then, the IsoSeq3 program from PacBio (commit v0.4.0-121-g22a3096*; (RRID:SCR_022749) [18] was used with the cluster option to group the CCS sequences into transcripts. The IsoSeq3 polish option was used to improve transcript accuracy. Prior to the availability of IsoSeq3 software, the PacBio RNA sequence data were analyzed using tools from IsoSeq1 and IsoSeq2 software distributions from PacBio, but the analysis was limited by the slow run time of the earlier algorithms. We also included any additional transcripts from these older algorithms in our genome annotation if they improved matches to the genome sequence compared with IsoSeq3 transcripts. Using the transcripts, the NCBI BLASTN algorithm v2.2.8 (RRID:SCR_001598) was used against the Bionano Genomics-scaffolded genome to identify probable locations for the corresponding genomic DNA. Then, Exonerate v2.2.0 (RRID:SCR_016088) [19] with the cdna2genome model was used to find the likely genomic location as well as putative exons and introns for each transcript. A maximum intron size of 30,000 bp was used initially for the search, with a minimum match percentage of 80%. For all transcripts for which no genomic location was found, a second Exonerate run was attempted, with a maximum intron size of 2,000,000 bp. This two-stage process was used to reduce the total computer run time required to search for all transcripts.

Finally, all predicted exon and intron locations were loaded into a local relational database to facilitate the preparation of a genome submission to the National Center for Biotechnology Information (NCBI), which included coding region predictions based on the transcript RNA sequences.

When preparing the genome submission to NCBI, we aimed to identify likely gene products. We used the BLASTX (RRID:SCR_001653) [20] algorithm to find reasonable matches of the translated transcriptome sequences against three publicly available plant proteomes: Asparagus officinalis L., Oryza sativa L. Japonica Group, and Zea mays L., with an E-score threshold of 0.001.

Because many transcripts aligned to overlapping regions of the genome, we reported only one transcript per region; we were unable to rank multiple aligning transcripts. We used the Exonerate alignments that spanned no more than 1 million base pairs, because a spot check of the very large alignments appeared to be artifactual and obscured smaller groups of alignments. For each region, we chose the transcript whose genomic sequence yielded the longest open reading frame based on the standard genetic code. We reported this coding sequence in the NCBI submission along with the product name from the BLASTX search, and we included a note providing the E-value and bit-score of the BLASTX alignment along with the RefSeq identifier of the plant protein.

It is challenging to generate accurate, relatively complete plant genome assemblies due to their large size, heterozygosity, and high frequency of repeat sequences. For these reasons, PacBio long-read sequencing technology was used to generate the Iris pallida Lam. genome assembly. The total size of the PacBio genome assembly was 10.46 Gbp. To enhance the assembly, we used Bionano Genomics optical mapping, since optical mapping on top of long-read sequencing is beneficial for producing higher quality plant genome assemblies [21, 22]. The total size of the genome assembly after Bionano Genomics scaffolding was 13.49 Gbp (Table 2). The additional size of the scaffolded genome was due to differing haplotigs in the phased assembly from Falcon Unzip.

Because primary contigs and haplotigs were included in the scaffolding process, many regions of this genome sequence are near duplicated. These near duplications are important to the future analysis of this genome because we cannot determine which allele of a heterozygous gene is functional.

Benchmarking Universal Single-Copy Orthologs (BUSCO) v 5.4.7 (BUSCO, RRID:SCR_015008) [23, 24] was run on the Bionano Genomics scaffolded assembly (Table 3) using both Augustus gene modeling software and the maize species parameters provided in Augustus. The lineage dataset was eukaryota_odb10, and the BUSCO mode was set to euk_genome_aug. The GC content of the genome was 41.2%. Compared with other plant genomes, the completeness of our assembly is reasonable with regards to genome–transcript alignment and BUSCO scores [25]. A small number of sequences were omitted from publication by NCBI due to short size or discovery of vector sequence contamination. Thus, the number of scaffolds for this genome reported by NCBI is slightly smaller than the number reported in Table 2.

Rhizome and leaf tissue samples were processed for RNA sequencing data. Statistics are shown in Table 4. For the rhizome sample, 3,032,725 CCS sequences were produced after processing by lima, resulting in 133,484 high-quality transcripts and 6,959 low-quality transcripts. For the leaf sample, 1,910,385 CCS sequences were produced after lima processing, resulting in 91,528 high-quality transcripts and 5,156 low-quality transcripts. Both high- and low-quality transcripts were used in the annotations. The CCS reads after lima processing were deposited into the NCBI Short Read Archive. There were 96,680 transcripts reported for the leaf sample and 140,135 transcripts reported for the rhizome sample. We found a total of 63,944 transcript-identified coding regions.

Out of 236,815 transcripts determined by the PacBio Isoseq3 method, 212,672 were aligned to the genome using Nucleotide BLAST, an alignment percentage of 89.8%. All transcripts that aligned using BLAST were then realigned against the Iris pallida Lam. genome in the vicinity of the genome matched by BLAST. The percentage of successful Exonerate alignments was 88.1%. The quality of alignment of the transcriptome to the genome was very high.

Exonerate computes the number of identical base matches for its alignments, as well as the number of mismatches. For all Exonerate alignments, the ratio of identical base matches to the sum of base matches and mismatches was 98.1%. Given that the individual plant used for RNA isolation was different from the individual plant used for genomic DNA isolation, this result represents an excellent agreement of nucleotide sequences.

After submitting the Iris pallida Lam. transcript sequences to NCBI, NCBI reported that 230 rhizome sample transcripts contained sequences from species other than Iris pallida Lam. Most of these contaminant sequences were from fungal species, an expected result given that the rhizome tissue sample was removed from soil. These sequences were removed from the final submitted transcriptome and were not used in the annotation of the Iris pallida Lam. genome that was submitted to NCBI.

In the Bionano Genomics scaffolded assembly, a total of 26 scaffolds had telomere sequences at their ends. These telomere data were included in the NCBI submission. We did not identify any contigs or scaffolds that had telomeres at both ends.

With 12 unique chromosomes [9], Iris pallida Lam. would be expected to have 24 unique concatenations of telomeres with chromosome end sequences. Given the inclusion of haplotigs into the genome assembly as well as its draft quality, the identification of 26 telomeres in our assembly is consistent with the observed chromosome number.