Chromosomal-level genome assembly of two dominant desert shrub species in Haloxylon (Amaranthaceae) Authors: Fang Yang, Di Gao, Yifei Wang, Xuelian Ding, Guanghui Lv Published: 2025 ABSTRACT AbstractHaloxylon ammodendron and Haloxylon persicum are ecologically and economically important drought-tolerant plants, often referred to as “desert guardians” owing to their remarkable abilities to withstand conditions of drought and salinity. We conducted a chromosome-level genome assembly using PacBio HiFi long-read sequences and Hi-C technology in order to gain a deeper understanding of their adaptive mechanisms and resource potential. The final assembled genome sizes were 2.32 Gb (contig N50 = 5.11 Mb; Scaffolds N50 = 257.59 Mb) for H. ammodendron and 1.32 Gb (contig N50 = 9.55 Mb; Scaffolds N50 = 143.67 Mb) for H. persicum, with 97.84% and 95.45% of the respective sequences anchored to nine pseudochromosomes. The BUSCO integrity scores were 88.40% and 84.00% for H. ammodendron and H. persicum, respectively. Gene annotation revealed that H. ammodendron contained 69,844 protein-coding genes, while H. persicum had 66,859 protein-coding genes, with repeat elements constituting 57.42% and 52.88% of their genomes, respectively. The reference genomes of Haloxylon serve as invaluable resources for exploring the ecological and economic significance of desert plants.Subject terms: Molecular ecology, Transcriptional regulatory elements Background and Summary The genus Haloxylon, a member of the Amaranthaceae family, comprises approximately 11 species and is distributed from the Mediterranean region to Central Asia. In China, two species, namely, Haloxylon ammodendron (C. A. Mey.) Bunge and Haloxylon persicum Bunge ex Boiss. & Buhse, are native to the northwest region1. These species are dominant shrubs or small trees occurring in desert areas2,3, characterized by opposite leaves that degrade into scales, fused at the base, with blunt or short-pointed tips at the apex (Fig. 1). Both species exhibit strong resistance to drought, salt-alkali conditions, and wind erosion, earning them the title of “Desert Guardians”4. In addition, their wood is strong yet brittle, possessing high combustion power and low smoke emission, hence the nickname “sand coal”5. This material is considered a high-quality fuel source and serves as an excellent building material for livestock shelters6. The tender branches provide excellent fodder for camels during winter and spring and also serve as hosts for the medicinal herb Cistanche deserticola7. Characteristics of the assimilating branches of H. ammodendron and H. persicum. (A) Assimilating branches of H. ammodendron grown in the same year; (B) Assimilating branches of H. persicum grown in the same year. H. ammodendron and H. persicum are C4 perennial succulent shrubs that exhibit drought and salt tolerance, characterized by unique morphological and physiological traits8. The species have become essential for desert ecological restoration through their ability to stabilize sand, regulate the microclimate, and maintain ecosystem homeostasis9,10. Notably, H. ammodendron and H. persicum possess distinct evolutionary adaptations to prolonged drought stress. Transcriptomic analyses indicate that transcription factors, such as MYB and AP2/ERF, are upregulated in H. ammodendron under drought conditions, whereas their inhibition is more pronounced in H. persicum11. However, the absence of a complete reference genome limits the accuracy of transcript splicing and functional annotation, resulting in reduced efficiency in key gene mining and the discovery of new transcripts. Haloxylon plants are facing significant threats due to overdevelopment and climate change in Central Asia12. Population declines and the resulting ecological consequences have generated considerable interest in genetic research on Haloxylon. However, a complete genome of the Haloxylon genus has yet to be released, a situation that severely hinders the utilization of this plant resource. Therefore, there is an urgent need for high-quality assembled genomes of the Haloxylon genus to expedite the development of stress resistance-related markers and facilitate molecular breeding. In this study, we performed a de novo genome assembly and annotation of H. ammodendron and H. persicum at the chromosome level using PacBio HiFi sequencing and High-throughput/Solution Chromosome Conformation Capture (Hi-C). The plants used in the analyses were obtained from the Ebinur Lake Nature Reserve, Xinjiang, China. The assembled genome sizes for H. ammodendron and H. persicum were 2.32 Gb and 1.32 Gb, respectively, with contig N50 values of 5.11 Mb and 9.55 Mb. H. ammodendron and H. persicum exhibited 97.84% and 95.45% of sequences anchored to nine chromosomes, corresponding to respective BUSCO scores of 88.40% and 84.00%. Genomic annotation revealed 69,844 and 66,859 protein-coding genes for H. ammodendron and H. persicum, respectively. This high-quality genome offers a valuable resource for further research into the ecological and economic significance of H. ammodendron and H. persicum, as well as the genetic mechanisms underlying long-term stress tolerance. Methods Sample collection and sequencing Plant materials of H. ammodendron (Fig. 1A) and H. persicum (Fig. 1B) were collected from the Ebinur Lake Wetland Nature Reserve in Xinjiang, China. The plant samples were immediately rinsed with ultrapure water, frozen in liquid nitrogen, and the sent to Wuhan Biorun Biotechnology Co., Ltd. (Wuhan, China) for genome sequencing. High-quality genomic DNA was prepared from the annual assimilating branches of H. ammodendron and H. persicum using the CTAB (cetyl trimethylammonium bromide) method13. The quality of genomic DNA was assessed using a NanoDrop™ One UV-Vis spectrophotometer (Thermo Fisher Scientific, USA). The OD260/280 ratio of pure DNA standard was between 1.8 and 2.0, and the OD260/230 ratio was between 2.0 and 2.2. Subsequent DNA quantification was performed using a Qubit® 4.0 fluorometer (Invitrogen, USA). A short-read-length library with an insert size of 350 bp was assembled using a DNA library construction kit (Illumina, San Diego, CA, USA) and then sequenced on the Novaseq 6000 platform (Illumina). H. ammodendron and H. persicum generated approximately 128.70 Gb (~55 × coverage) and 129.91 Gb (~98 × coverage) of short-read sequencing data, respectively. The read data were subsequently used for genomic survey analysis and error correction in the genome assembly (Table 1). The PacBio Sequel II platform (Pacific Biosciences, CA, USA) was used for PacBio HiFi sequencing following the manufacturer’s instructions. The total HiFi reads generated by H. ammodendron and H. persicum comprised 56 Gb and 22 Gb, respectively (Table 1). For the Hi-C experiment, the current-year assimilating branches of H. ammodendron and H. persicum were fixed in a 2% (w/v) formaldehyde solution in NIB buffer, followed by chromatin extraction, DPN II digestion, DNA ligation, purification, and fragmentation. The final libraries were sequenced on the Novaseq 6000 platform (Illumina), generating 237.95 Gb and 151.81 Gb of Hi-C reads, respectively (Table 1). Table 1Genome assembly and assessment of the H. ammodendron and H. persicum genomes.AssemblyH. ammodendronH. persicumIllumina sequencing128.70 Gb129.91 GbPacBio sequencing56 Gb22 GbHi-C237.95 Gb151.81 GbEstimated genome size (Gb)2.001.49Estimated heterozygosity (%)1.37%1.43%Scaffolds N50 (Mb)257.59143.67Longest scaffold (Mb)288.43154.99Number of scaffolds382494Total length of scaffolds (Gb)2.321.32Contigs N50 (Mb)5.119.55Longest contig (Mb)23.4944.54Number of contigs30941453Total length of contigs (Gb)2.321.32GC content (%)35.2835.17Repetitive sequences57.42%52.88%The number of genes69,84466,859Number of functionally annotated genes60,89563,487Average gene CDS length988.28 pb2140.54 pbAverage exon per gene3.943.95BUSCO (% complete)88.40%84.00% Genome assembly and assessment of the H. ammodendron and H. persicum genomes. Transcriptome sequencing was conducted on the assimilating branches of H. ammodendron and H. persicum to assist in gene prediction. Total RNA was extracted using an RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). A cDNA library was constructed according to the manufacturer’s instructions and sequenced on the MGISEQ-2000 platform. The assimilating branch samples of H. ammodendron and H. persicum generated 7.21 and 7.04 Gb of raw data, respectively (Table 2). Table 2RNA sequencing statistics.SampleRaw dataClean dataTotal number of readsTotal number of bases (Gb)Total number of readsTotal number of bases (Gb)H. ammodendron48,073,5847.2147,504,8846.71H. persicum46,908,5407.0446,422,1846.55 RNA sequencing statistics. Genome survey The Novaseq 6000 platform (Illumina) short reads were filtered using Fastp v0.21.014 (with default parameters) to remove adapters, repetitive sequences, and low-quality sequences. K-mer analysis was conducted to estimate the genome size and heterozygosity of H. ammodendron (Fig. 2A) and H. persicum (Fig. 2B). The frequency of 17-bp K-mers was determined using Jellyfish v.2.1.41615. Based on these analyses, and using GenomeScope v.2.016, the genome sizes of H. ammodendron and H. persicum were estimated as 2.00 Gb and 1.49 Gb, respectively. The rates of heterozygosity were 1.37% and 1.43%, respectively, and the respective duplication rates were 61.30% and 61.00%. K-mer (17-mer) frequency distribution curve. (A) K-mer frequency distribution curve of H. ammodendron; (B) K-mer frequency distribution curve of H. persicum. De novo assembly without a reference genome SMRT_Link v7.0 (https://www.pacb.com/smrt-link/) was used to filter the raw data from PacBio HiFi sequencing remove low-quality reads and adaptor sequences. The filtered reads were subjected to de novo assembly using HiFiasm v0.18.0-r46517 with default parameters, resulting in an initial draft genome at the contig level. For the Hi-C raw data, Juicer18 aligned the Hi-C sequencing data to the draft genome and generated a Hi-C interaction matrix by incorporating restriction enzyme cleavage site information (Fig. 3). The Hi-C contact map was visualized using Juicebox Assembly Tools (v1.11.08)19, and apparent assembly errors (e.g., mis-splicing and inversions) were corrected manually. After the correction, the script “run-asm-pipeline-post-review.sh” provided by 3D-DNA was implemented to generate the final scaffold assembly at the chromosome level. The final chromosome-scale genomes of H. ammodendron and H. persicum comprised 2.32 Gb and 1.32 Gb, respectively. The N50 values of the contigs were 5.10 Mb and 9.55 Mb, respectively, and the respective N50 values for the scaffolds were 257.59 Mb and 143.67 Mb (Table 1). Total of 2.27 Gb (97.84%) and 1.26 Gb (95.45%) of sequences from H. ammodendron and H. persicum were successfully anchored onto nine chromosomes (Table 3). Hi-C interaction heatmap. (A) Hi-C interaction heatmap of H. ammodendron; (B) Hi-C interaction heatmap of H. persicum. Table 3Summary data for nine pseudochromosomes of H. ammodendron and H. persicum.ChromosomeH. ammodendronH. persicumLength (bp)Length (bp)Chr01288,434,937133,946,977Chr02232,602,162153,633,959Chr03268,553,148152,556,302Chr04210,716,851143,665,069Chr05224,719,551107,041,274Chr06241,870,460132,698,176Chr07257,588,235131,915,868Chr08272,534,332153,299,713Chr09274,684,479154,996,254Total2,271,704,1551,263,753,592 Summary data for nine pseudochromosomes of H. ammodendron and H. persicum. Based on the genome assembly at the chromosomal level, Circlize20 was used for analysis to generate a genome circle diagram and visualize chromosome length, gene density, and the GC content (Fig. 4). We calculated the Long Terminal Repeat Retrotransposons Assembly Index (LAI), a reference-free genome metric used to assess the assembly of repetitive sequences in order to evaluate the continuity of the genome assembly21. The LAI values of the H. ammodendron and H. persicum genome assemblies were 3.52 and 5.11, respectively. Genomes with LAI values below 10 are considered to have poor initial genome assembly quality21. This may be because the plant materials used for genome assembly only included assimilating branches, but not plant materials such as roots, stems, and fruits, resulting in a lower-quality genome assembly. Distributions of genomic features. (A) The distribution of genomic features of H. ammodendron; (B) The distribution of genomic features of H. persicum. First layer: Chromosome length; Second layer: Gene density; Third layer: GC content; Fourth layer: Chromosomal collinearity. Genome annotation and functional prediction A repetitive sequence library of the genome was constructed using EDTA (v2.0.1)22. The library was annotated using the RepeatMasker (v4.1.1) software23, and the repetitive sequences in the genome were masked as N. 57.42% (1.33 Gb) and 52.88% (699.59 Mb) in H. ammodendron and H. persicum, respectively. Among the repetitive sequences, 51.61% and 48.28% were transposons (TE), including long terminal repeat retrotransposons (LTR-RTs) (28.89% and 26.50%), non-LTR-RTs (0.45% and 0.62%), and DNA transposons (22.27% and 21.16%) (Table 4). Table 4Annotation information for repetitive sequences of H. ammodendron and H. persicum.ClassH. ammodendronH. persicumLength (bp)Masked (%)Length (bp)Masked (%)Class I: LTR-RTs————Copia133,227,3485.76%76,648,8035.79%Gypsy370,828,56416.02%200,694,52115.17%unknown164,545,4777.11%73,336,5845.54%Non-LTR-RTs————LINE_element6,877,0070.30%5,263,2090.40%SINE_element126,3570.01%73,3360.01%unknown3,132,9660.14%2,760,8260.21%Class II: DNA Transposon————CACTA61,372,8142.65%28,476,3762.15%Mutator55,775,4042.41%29,515,5832.23%PIF_Harbinger30,814,2181.33%16,169,2641.22%Tc1_Mariner184,651,1397.98%94,424,1017.14%hAT43,376,0261.87%22,585,0971.71%polinton89,8820.00%——helitron139,589,9366.03%88,819,7876.71%repeat_region134,474,3175.81%60,824,0934.60%Total1,328,881,45557.42%699,591,58052.88% Annotation information for repetitive sequences of H. ammodendron and H. persicum. Protein-coding genes were predicted using MAKER v3.01.0324. De novo prediction was performed using GeneMark-ES/ET/EP 4.68_3.60_lic25 and Augustus v3.2.226. Evidence from the transcriptome and protein homology was integrated to enhance prediction accuracy. The transcriptome data were aligned to the genome using HISAT2 v2.2.127. To search for information that could be used in homology prediction, we selected five closely related species (Populus alba, Populus cathayana, Populus qiongdaoensis, Salix purpurea, and Turnera subulata) and obtained their protein sequences. We then used Exonerate (v2.58.3) software to conduct the homology prediction. Finally, EVidenceModeler (v2.1.0)28 was used to integrate the three prediction results. Overall, 69,844 and 66,859 protein-coding genes were predicted in the genomes of H. ammodendron and H. persicum, respectively. For functional annotation of protein-coding genes, the predicted gene sequences were aligned with public functional databases using BLAST v2.11.029 (E-value ≤ 1e−5) and InterproScan30: Kyoto Encyclopedia of Genes and Genomes (KEGG)31, Gene Ontology (GO)32, eukaryotic Ortholog Groups (KOG)33, Pfam34, Swissprot35, NCBI nucleotide sequences (NT), and NCBI non-redundant protein sequences (NR)36. The results showed that 60,895 (87.19%) and 63,487 (94.96%) genes in the H. ammodendron and H. persicum genomes, respectively, were successfully annotated in the functional database (Table 5). Table 5Gene function annotation of H. ammodendron and H. persicum genomes.TypeH. ammodendronH. persicumNumberPercent (%)NumberPercent (%)NR28,48246.773052148.07NT22,18036.422389837.64KEGG10,06916.531072416.89SwissProt20,10433.012129733.54Pfam22,09636.282296736.17GO22,09536.282296636.17KOG58419.5959889.43Annotated in all of the databases35905.8936915.81Annotated in at least one database33,91355.693644957.41Total genes60,89510063487100 Gene function annotation of H. ammodendron and H. persicum genomes. Data Records The original contributions presented in this study are publicly available. The dataset is available at the NCBI database (https://submit.ncbi.nlm.nih.gov/subs/). RNA-Seq data have been saved in the SRA of NCBI with accession SRR2112133537, SRR2112133838, SRR2112134439, SRR2112134040, SRR2112134141, SRR2112134242. The raw sequence data (PacBio HiFi, Hi-C) have been deposited in the NCBI genome database under the BioProject accession numbers PRJNA1221346 and PRJNA1221351, with accession numbers GCA_053755035.143 and GCA_053755055.144. The genome annotation file has been deposited in the Figshare database45. Technical Validation We evaluated the quality of the genome assembly from the following aspects: (1) the use of contig N50, scaffold N50, and BUSCO v5.4.340 to assess genome continuity and integrity. The results indicated that the contig N50 lengths for H. ammodendron and H. persicum were 5.11 Mb and 9.55 Mb, respectively, while the scaffold N50 lengths were 257.59 Mb and 143.67 Mb, respectively, resulting in 88.40% and 84.00% complete BUSCO genes. (2) Mapping short read lengths to the genome assembly yielded localization rates of 99.97% and 99.95%, respectively. These findings suggest that the genome assemblies of H. ammodendron and H. persicum are of high quality. Acknowledgements We thank the associate editor and anonymous reviewers for their helpful comments. This work was supported by the Central Government Guide Local Special Fund Projects for Science and Technology Development (ZYYD2025ZY04), the National Natural Science Foundation of China (Youth Fund) (32301307), and the Natural Science Foundation of Xinjiang Uygur Autonomous Region (2023D01C186). We thank LetPub (www.letpub.com.cn) for linguistic assistance and pre-submission expert review. Author contributions F.Y. conceived the study; F.Y. collected data, analyzed the data, and drafted the text; D.G. collected and analyzed the data; Y.F.W. and X.L.D. collected samples; G.L. revised the manuscript. All of the authors edited the manuscript. Data availability All raw sequencing data can be obtained from NCBI genome database, the genome annotation files can be accessed from the Figshare databases. Code availability All software involved in this research analysis was implemented according to the manuals and protocols provided by the software developers. The software versions were listed in the methods. Custom code was not used in this study. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Citations Data Availability Statement All raw sequencing data can be obtained from NCBI genome database, the genome annotation files can be accessed from the Figshare databases. All software involved in this research analysis was implemented according to the manuals and protocols provided by the software developers. The software versions were listed in the methods. Custom code was not used in this study.