The nematode Halicephalobus mephisto was originally discovered inhabiting a deep terrestrial aquifer 1.3 km underground. H. mephisto can thrive under conditions of abiotic stress including heat and minimal oxygen, where it feeds on a community of both chemolithotrophic and heterotrophic prokaryotes in an unusual ecosystem isolated from the surface biosphere. Here we report the comprehensive genome and transcriptome of this organism, identifying a signature of adaptation: an expanded repertoire of 70 kilodalton heat-shock proteins (Hsp70) and avrRpt2 induced gene 1 (AIG1) proteins. The expanded Hsp70 genes are transcriptionally induced upon growth under heat stress, and we find that positive selection is detectable in several members of this family. We further show that AIG1 may have been acquired by horizontal gene transfer (HGT) from a rhizobial fungus. Over one-third of the genes of H. mephisto are novel, highlighting the divergence of this nematode from other sequenced organisms. This work sheds light on the genomic basis of heat tolerance in a complete subterrestrial eukaryotic genome.

High levels of heterozygosity present a unique genome assembly challenge and can adversely impact downstream analyses, yet is common in sequencing datasets obtained from non-model organisms. Here we show that by re-assembling a heterozygous dataset with variant parameters and different assembly algorithms, we are able to generate assemblies whose protein annotations are statistically enriched for specific gene ontology categories. While total assembly length was not significantly affected by assembly methodologies tested, the assemblies generated varied widely in fragmentation level and we show local assembly collapse or expansion underlying the enrichment or depletion of specific protein functional groups. We show that these statistically significant deviations in gene ontology groups can occur in seemingly high-quality assemblies, and result from difficult-to-detect local sequence expansion or contractions. Given the unpredictable interplay between assembly algorithm, parameter, and biological sequence data heterozygosity, we highlight the need for better measures of assembly quality than N50 value, including methods for assessing local expansion and collapse. In the genomic era, genomes must be reconstructed from fragments using computational methods, or assemblers. How do we know that a new genome assembly is correct? This is important because errors in assembly can lead to downstream problems in gene predictions and these inaccurate results can contaminate databases, affecting later comparative studies. A particular challenge occurs when a diploid organism inherits two highly divergent genome copies from its parents. While it is widely appreciated that this type of data is difficult for assemblers to handle properly, here we show that the process is prone to more errors than previously appreciated. Specifically, we document examples of regional expansion and collapse, affecting downstream gene prediction accuracy, but without changing the overall genome assembly size or other metrics of accuracy. Our results suggest that assembly evaluation methods should be altered to identify whether regional expansions and collapses are present in the genome assembly.

Background: Whole-genome shotgun sequencing, which stitches together millions of short sequencing reads into a single genome, ushered in the era of modern genomics and led to a rapid expansion of the number of genome sequences available. Nevertheless, assembly of short reads remains difficult, resulting in fragmented genome sequences. Ultimately, only a sequencing technology capable of capturing complete chromosomes in a single run could resolve all ambiguities. Even “third generation” sequencing technologies produce reads far shorter than most eukaryotic chromosomes. However, the ciliateOxytricha trifallaxhas a somatic genome with thousands of chromosomes averaging only 3.2 kbp, making it an ideal candidate for exploring the benefits of sequencing whole chromosomes without assembly.Results: We used single-molecule real-time sequencing to capture thousands of complete chromosomes in single reads and to update the publishedOxytricha trifallaxJRB310 genome assembly. In this version, over 50% of the completed chromosomes with two telomeres derive from single reads. The improved assembly includes over 12,000 new chromosome isoforms, and demonstrates that somatic chromosomes derive from variable rearrangements between somatic segments encoded up to 191,000 base pairs away. However, while long reads reduce the need for assembly, a hybrid approach that supplements long-read sequencing with short reads for error correction produced the most complete and accurate assembly, overall.Conclusions: This assembly provides the first example of complete eukaryotic chromosomes captured by single sequencing reads and demonstrates that traditional approaches to genome assembly can mask considerable structural variation.