Parasites and cultivation

Four different lines of the Leishmania major strain LV561 (MHOM/IL/67/LRC-L137 Jericho-II) were investigated: the virulent line (V) freshly re-isolated from BALB/c mice, the avirulent line (AV) attenuated by 101 in vitro passages in culture, and the lines AVM and AVS that were obtained from the AV line by five subsequently passages in mice (AVM) and sand flies Phlebotomus duboscqi (AVS) [13]. All the lines were tested previously for virulence in BALB/c mice and development in sand flies. Prior to the molecular analysis described here, the lines were maintained in a cryo-bank at Charles University, Prague: AV frozen as p101 in 2014, V frozen as p1 (1 passage in vitro after isolation from mice) in 1996, AVM frozen as p3 (3 passages in vitro after 5 passages in mice) in 1999, and AVS frozen as p5 (5 passages in vitro after 5 passages in sand flies) in 1999. Additional in vitro passaging was avoided. Lines V and AV were retested in 2018 and 2020 and their phenotypes were confirmed. After thawing, parasites were passaged once in M199 medium (Sigma-Aldrich/ Merck, St. Louis, USA) supplemented with 10% heat-inactivated fetal calf serum (Thermo Fisher Scientific, Waltham, USA), 2% sterile urine, 1% Basal Medium Eagle vitamins (MP Biomedicals, Irvine, USA), and 250 μg/ml amikacin (Medochemie, Prague, Czechia) as described previously [26] and analyzed. The identity of species was validated as in [27].

Nucleic acid isolation, library preparation and sequencing

Nucleic acids (DNA and RNA) were isolated as described previously [28] and sequenced at Macrogen Europe (Amsterdam, the Netherlands). The TruSeq DNA PCR-free protocol was used for the genomic DNA library that was sequenced on a NovaSeq 6000 resulting in ~55M 150 bp paired-end reads. Three TruSeq stranded mRNA libraries per line were sequenced on the same platform resulting in ~125M 150 bp paired-end reads each. Raw sequencing reads were deposited in the NCBI database under BioProject accession PRJNA1391897.

Read processing

Sequencing reads were quality-checked with FastQC v. 0.12.0 [29] and trimmed with fastp v. 0.25.0 [30]. Specific flags ‘--trim_poly_g’, ‘--trim_poly_x’, ‘--cut-front’, ‘--cut_tail’, ‘--cut_window_size 4’ and ‘--cut_mean_quality 20’ were used to properly handle 2-dye chemistry sequencing output, ‘--length_required 100’ was employed to preserve information for unique read position assignment and get higher contiguity of de novo assemblies. The same settings were applied to transcriptomic reads, except for the ‘cut_front_mean_quality’, ‘cut_tail_mean_quality’, and ‘average_qual’ set to defaults, ‘length_required’ set to 50.

De novo nuclear genome assembly and variant detection

Sample data were assembled de novo using SPAdes v. 4.2.0 [31]. Assembly completeness was assessed with BUSCO v. 6.0.0 [32]. Sequencing reads were mapped back on each genome’s assembly with Burrows-Wheeler Aligner (BWA) v. 0.7.19 [33] using ‘MEM’ algorithm with default settings. BAM files were processed with SAMtools v. 1.22 [34], alignment files were visualized in IGV v. 2.16.0 [35] to perform SNP quality curation. Variant calling was performed by a pileup method in BCFtools v. 1.22 [34] and FreeBayes v. 1.13.10 [36], both with default settings. BCFtools determined SNPs with ‘QUAL <= 20’ were filtered out and only the intersection of homozygous SNP sets from BCFtools and FreeBayes was further used. Correlations between samples were analyzed and plotted using a custom R script and ggplot2 R library v. 4.0.1 [37] for visualization.

Reference-based variant detection and copy-number variation analysis of the nuclear genome

Trimmed and quality-checked sequencing reads were mapped on the reference genome of Leishmania major Friedlin from the TriTrypDB release 58 [38] with the BWA-MEM algorithm. Alignments were processed with SAMtools and variant calling was performed with FreeBayes. Variant filtering and inter-sample comparisons were done with BCFtools ‘view’ and ‘isec’ commands and the output was processed with custom python scripts into the final tables. Only homozygous (or nearly homozygous with allele frequency over 0.8) SNPs were analyzed. BEDtools v. 2.31.1 [39] were used for various genomic region manipulations and SNP counting. Gene coding regions along with 1,000 nt upstream and downstream flanking sequences were considered when assigning the SNPs to genes. Regions of interest were extracted and inspected in IGV to confirm the presence of SNPs.

Large structural variants were detected with the DELLY v. 1.5.0 [40] ‘call’ algorithm using the same bam mappings and visualized using a custom R script and gridExtra R library v. 2.3 [41]. All samples were collectively used to detect locus copy number variation with cn.MOPS R package [42] with window length values 300; 1,000; 5,000; and 10,000, and the ‘DNAcopy’ algorithm using the same bam mappings.

Comparison of mitochondrial genomes and their expression

The BLASTN program of NCBI-BLAST suite v. 2.15.0+ [43] was used to search for the known maxicircle coding region (CR) sequence of Leishmania spp. in de novo assembled contigs of the V line. The coding region was extracted, re-oriented to start with the 12S rRNA gene on the forward strand, and annotated by homology with Leishmania spp. The RNA sequencing reads were mapped and processed with T-Aligner and T-Aligner DE pipelines, as described previously [44,45]. Briefly, reads were mapped onto each gene or cryptogene sequence possessing short flanking sequences of ~50 bp. Predicted RNA editing and expression profiles and edited mRNA sequences were obtained with ‘findorfs’ and ‘coverage’ tools of the T-Aligner suite. The differential expression analyses of mitochondrial transcripts were performed on each pair of samples (in triplicates) using EdgeR [46] to detect differentially edited states with adjusted p-value significance threshold 0.001 and editing site support absolute fold change > 2. Visualization was done with T-Aligner scripts for differential editing analyses.

Differential expression analysis of the nuclear genome

RNA sequencing reads were quality-checked with FastQC and MultiQC v. 1.31 [47] and trimmed with fastp. Trimmed reads were mapped onto the reference genome of L. major Friedlin from TriTrypDB with STAR v. 2.7.11b [48]. Produced and sorted bam files were quality-checked with Qualimap v. 2.3 [49]. Read counts were obtained with featureCounts v. 2.0.6 [50]. Differential expression was analyzed with DESeq2 package [51], using the simple ‘control versus experiment’ setup, where sample groups were provided in three biological replicates (3 versus 3). Genes with a fold change of ≥ 2 and an adjusted p-value of ≤ 0.001 were considered differentially expressed. Protein products conceptually translated from the differently expressed genes of all lineages were submitted to STRING v. 12.0 [52] and the network reconstruction was performed as described previously [53]. The same set of proteins was submitted to the eggNOG-mapper v. 2.1.12 [54] to identify potential functions of proteins using the clusters of orthologous genes (COG) categories. In total, 195, 185, and 129 differentially expressed genes were identified in the V vs AV, AVM vs AV, and AVS vs AV comparisons, respectively, of which 152, 144, and 98 genes carried a COG annotation. Differentially expressed genes lacking annotation or assigned to the “function unknown” category (S) constituted ~30–40% of the total across all comparisons, while “general function” category (R) was not part of the reference genome annotation table. Neither group was analyzed further. To assess whether differentially expressed genes were non-randomly distributed across functional categories, we performed one-sided Fisher’s exact tests for enrichment against the 4,754 COG-annotated genes of the L. major Friedlin reference genome, with Benjamini–Hochberg false discovery rate (FDR) correction applied within each comparison (S1 Table). The RRMScorer [55] was used to analyze the probability of RNA binding and preferences of differentially expressed proteins LmjF.18.0170 and LmjF.18.0190, which were also analyzed for nucleic acid and amino acid similarity and identity using the Needleman-Wunsch algorithm through EMBL-EBI [56]. MEME tool from the MEME suite [57] was used to detect potential motifs of sizes from 5 bp to 12 bp for the 217 differentially expressed genes of the V and AVM lines.

Changes in the gene copy number are likely involved in alterations to avirulence

Single nucleotide polymorphisms (SNPs) are an obvious and testable source of phenotypic variation. Therefore, we compared genomic sequences between parasite lines to find homozygous SNPs, using the L. major Friedlin high-quality reference genome [58] for calling. In parallel, SNPs were identified using a de novo assembled V line, as the sequence divergence between Friedlin and LV561 strains could potentially cause mapping artifacts. Hundreds of SNPs between genomes of the LV561-derived lines and L. major Friedlin reference were detected. However, patterns of the variant calls among LV561-derived lines suggested that most were artifacts: most SNPs were found in tandemly repeated or long genes composed of internal repeated sequences, such as proteophosphoglycans. This genomic context is known to trigger technical artifacts during the alignment process [59–61]. Coverage variation results in SNPs called because of reads mapping to multiple locations, for which allele frequency floats across samples. These may be computationally identified as homozygous, heterozygous, or even absent SNPs across samples. Particularly when coverage is low, this can result in the identification of homozygous SNP false positives. Following manual curation to remove these, no cell line-specific homozygous SNPs were identified. Homozygous SNPs of all LV561-derived lines relative to the reference genome were identical, so we concluded that de novo mutations are not responsible for virulence attenuation.

Next, we analyzed parasite line-specific locus copy number variation (CNV) [16,62–65]. Only five loci exhibited statistically significant CNV among parasite lines (S1 Table). A chromosome 17 region is absent in the AVS line, eliminating several copies of a gene encoding the eukaryotic elongation factor 1-alpha (eEF1A). Also notable is a 7 kb locus of chromosome 34 with a copy number reduction in the AV relative to the V, AVM, and AVS lines. It contains a single gene (LmjF.34.2380) of unknown function. The protein encoded by LmjF.34.2380 appeared in Trypanosomatidae after the separation of Trypanosoma lineage with several independent losses. It does not contain any predicted domain and it is likely cytosolic (S1 Table, tab LmjF.34.2380).

The largest and, arguably, the most interesting CNV region is overrepresented in the AV line. This 12 kb-long locus on chromosome 35 contains genes encoding 60S ribosomal protein L37 (LmjF.35.5100), RNA pseudouridylate synthase (LmjF.35.5110), biopterin transporter 1 (LmjF.35.5140 and LmjF.35.5150), tRNA (guanine-N(7)-)-methyltransferase (LmjF.35.5230), alpha-1,2-mannosyltransferase (LmjF.35.5250), isopentenyl-diphosphate delta-isomerase (LmjF.35.5330), amino acid permease (LmjF.35.5350 and LmjF.35.5360), as well as several genes encoding proteins of unknown function. Many of these proteins have been implicated in Leishmania spp. virulence [66,67]. Of special interest are the isopentenyl-diphosphate delta-isomerase shown in a GP63^high fraction of L. amazonensis exosomes to be implicated in enhanced cutaneous leishmaniasis development [68], and biopterin transporter 1. The latter was previously identified to be a part of the LD1 (Leishmania DNA 1) locus [69–71] and its overexpression has been linked to adaptations to in vitro cultivation conditions and loss of virulence in different Leishmania spp. [72–75], similar to the AV line. This increase in copy number is correlated with the upregulation of transcripts of nearly all these genes two to four-fold in the AV line relative to the V line (S2 Table). Thus, it is likely that the AV line’s CNV is the mechanism, through which this gene dosage change is affected. However, additional experiments are necessary to determine whether the chromosome 35 CNV results in a changed growth or protein translation rate that impacts the parasite’s ability to grow in host organisms.

In summary, while SNPs in the nuclear genome appear unlikely to drive alterations in parasite virulence in different hosts, some CNV-facilitated gene dosage differences are likely involved in this process.

Passage of the avirulent strain in mice restores gene expression nearly to that of the original virulent strain

Next, we analyzed whether the differential gene expression could account for the changes in virulence or persistence phenotypes across different parasite cell lines. As regulation at the transcriptional level is largely absent in trypanosomatids [76,77], differences in mRNA expression between cell lines likely result from altered mRNA stability. To analyze this, we performed high-throughput RNA sequencing followed by pairwise analyses of mRNA abundance in V, AVS, and AVM lines relative to the AV line (S2 Table). In the V line, 192 transcripts of protein coding loci were at least two-fold differentially expressed in comparison to the AV line (S2A Table). Nearly this many (182) were differentially expressed in the AVM line (S2B Table), whereas only 127 were differentially expressed in the AVS line compared to the AV counterpart (S2C Table). Twenty-three percent of all differentially expressed transcripts (71 out of 311) were annotated as coding for “hypothetical protein” and not studied further.

To globally characterize the outcomes of these three differential gene expression analyses, we generated STRING networks (Figs 1 and S1 for the full network) and categorized genes into clusters of orthologous genes (COG) (Fig 2). While the COG categories revealed differences regardless of directionality of a change, the STRING diagrams showed directionally conserved changes in mRNA abundance between the different comparison sets and relationships between the transcripts involved.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Protein comparison of differently expressed genes in studied cell lines.

The protein network as generated by STRING shows changes in mRNA abundance between (A) V, AVM, and AVS versus AV; (B) V and AVM versus AV; (C) AVS versus AV; and (D) V versus AV. Nodes and edges represent L. major proteins and known or predicted interactions among them. Upward and downward arrows indicate up- or downregulation of the gene. For the full network see S1 Fig.

https://doi.org/10.1371/journal.pntd.0014387.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. COG categories of differently expressed genes in studied cell lines.

Proteins of differently expressed genes were categorized by eggNOG-mapper and their relative abundance is shown as horizontal stacked bar plots. Inside each bar, a total number of differentially expressed genes of that category is given. Note that category S: Function unknown and proteins without assigned COG were omitted for visualization purposes. For details see S2 Table.

https://doi.org/10.1371/journal.pntd.0014387.g002

Over 40% of differentially expressed genes in all three comparisons code for elements of information storage and processing with a larger percentage of these genes being linked to ribosomes and translation as shown in Figs 1 and 2 (COG category J with n = 383 genes; V vs AV: 35 DEGs, FDR = 1.40 × 10 ⁻ ⁷; AVM vs AV: 33 DEGs, FDR = 4.45 × 10 ⁻ ⁷; AVS vs AV: 30 DEGs, FDR = 1.07 × 10 ⁻ ⁹). A completely interconnected cluster of mRNAs all downregulated roughly two-fold in V, AVM, and AVS lines relative to AV was identified (Fig 1A). The tight spatial configuration and extensive associations of these transcripts is due to their identical ontology – they all encode ribosomal proteins. Furthermore, transcripts encoding additional ribosomal proteins, proteins needed for ribosome assembly, and proteins related to translation are among the shared genes exhibiting the same pattern of expression (Fig 1B-1D). Similarly to the universally regulated ribosomal proteins (Fig 1A), these transcripts are also downregulated approximately two-fold in the virulent lines, except for the eEF1A that was downregulated about six-fold in the AVS line. The latter finding is likely linked to the loss of eEF1A copies described in the previous section.

Exponential growth in culture that is unrestricted by nutritional constraints or other challenges of a host environment (for example, host defense mechanisms) naturally increases in rate. It is logical to assume that transition of cellular resources to support enhanced growth would include gene expression adaptations to increase the overall output of the protein translation machinery [78,79]. Conversely, unicellular organisms typically reduce translation in more restrictive circumstances (for example, upon reaching stationary phase in culture or in hosts, where conditions or nutritional resources may be less than ideal) [80,81]. In line with this, the most apparent gene expression feature of L. major cell lines associated with growth in a host versus in culture is a reduction in ribosomal proteins and proteins involved in ribosome functioning.

Changes to transcript abundances appearing after sand fly passage were not as extensive as those appearing after mouse passage, yet they did involve several different biological processes and pathways. In particular, the COG categories of cell motility (category N; 7 of only 13 annotated genes; FDR = 2.37 × 10 ⁻ ⁸) and intracellular trafficking (category U; present but not statistically enriched within AVS relative to AV differentially expressed genes) appeared completely or nearly confined to the AVS versus AV comparison, respectively (Fig 2). Cell motility upregulated transcripts in the AVS line (Fig 2) encoded proteins involved in the formation or function of the flagella. Also, the AVS versus AV changes to metabolic gene expression were centered around individual transcripts relating to carbohydrate transport and metabolism (category G; present but not statistically enriched). The most upregulated individual transcripts in AVS relative to the AV line were those related to synthesis of various phosphoglycans, or GPI-anchored proteins and other surface proteins, including transmembrane transporters (Fig 1C and S2C Table). While other comparisons also exhibited changes in transcripts related to the cell surface proteins, the identities of the individual transcripts involved did not overlap. Finally, a few transcripts involved in glycine biosynthesis were less abundant, while some facilitating its catabolism were more abundant. Additionally, some transcripts for stress response pathway proteins were mildly reduced in abundance, as were transcripts of a family of L. major specific nucleases (Fig 1C and S2C Table).

In contrast to changes resulting from passaging in sand flies, passaging in mice restored transcript abundances remarkably close to those of the original V line. Of all genes found to be upregulated in the V line relative to the AV line, 94% were also upregulated in the AVM line, typically to the same level (S2A and S2B Table). Of the transcripts found upregulated in only one of these comparisons, only two were upregulated more than 2.5-fold. Therefore, the pattern of transcript abundances in V and AVM lines was nearly identical. Compared to changes resulting from sand fly passaging, transcript abundance changes documented in the AVM (or V) line over AV were larger in magnitude. A few transcripts showed increases of eight- or even ten-fold (S2B Table). Finally, more transcripts were upregulated after passaging in mice (106, including hypothetical proteins and counting each copy of multicopy genes separately), relative to the number of transcripts upregulated following sand fly passaging (37) (S2 Table).

Topping the list of transcripts upregulated in V or AVM lines were surface proteins. One of them (LmjF.04.0210 encoding a putative GPI-anchored protein) was upregulated 10-fold over the AV line. Altogether, transcripts of 11 proteins likely found on the parasite surface or equipped with a GPI anchor [82,83] were upregulated in the AVM line (S2B Table). They appear confined to the genus Leishmania or their nearest phylogenetic neighbors of the subfamily Leishmaniinae [84]. Many of these proteins possess leucine-rich repeat motifs. Their upregulation is expected, as they are implicated in macrophage infection. For example, the upregulated transcripts for proteophosphoglycans encode proteins PPG1–4 that play a role in macrophage binding preceding cell entry and amastigote replication [85,86]. Similarly, transcripts of eight proteins putatively involved in lipid metabolism or transport, including those encoding an acetylase, fatty acid elongase, and two ABC transporters, were upregulated (S2B Table). This is also expected, as the mechanisms to generate and move more surface proteins into place would also have to be more robust under these conditions. Cytoskeletal genes (Fig 2; category Z) were additionally enriched in both the V vs AV and AVM vs AV comparisons (14 and 13 differentially expressed genes of 149 total; FDR = 2.99 × 10 ⁻ ³ and 6.10 × 10 ⁻ ³, respectively), hinting at a host-specific component to cytoskeletal remodeling. Finally, despite GP63 protein levels and activity being downregulated in the AV line relative to the V, AVS, and AVM lines [16], we see only gp63–4 was modestly upregulated in the AVM line (S1 Fig and S2B Table). Likewise, transcripts encoding LPG proteins that have been previously implicated in Leishmania virulence [17,87–89] were consistently expressed in all lines. At least for GP63, this is consistent with a previous hypothesis that expression and activity of GP63 may be influenced by regulatory mechanisms downstream of mRNA abundance, involving the coordinated interactions of the suite of surface proteins [23,90,91].

Specific regulation of the L. major proteins related to virulence or extended survival in different hosts is, at least partially, affected through modulation of transcript abundance. The logical question then becomes “how can transcripts be stabilized or destabilized under different environmental pressures”? One possibility is that transcripts might be differentially stabilized via ribonucleoprotein complexes (RNPs) [76]. The factors often driving RNP associations are RNA-binding proteins (RBPs) that typically bind 3′ UTRs of the mRNAs they regulate [92]. The related proteins LmjF.18.0170 and LmjF.18.0190 identified here are putative RBPs as they possess RNA-binding domains. The encoding genes are similar size and share 57% nucleotide identity. Their translated products retain a region that is 59% identical and 70% similar with two adjacent RNA recognition motifs identified by RRMScorer. The potential RNA motif for favorable binding was predicted to be U(A/U)UU. The transcript abundance of LmjF.18.0170 is modestly upregulated in both V and AVM lines and LmjF.18.0190 is four-fold upregulated in AVM (S2B Table). We hypothesized that these or other RBPs could be master regulators of the observed expression changes. An initial perusal of the first 100 nt of the 3′ UTRs of differentially expressed genes in the AVM line did not reveal enrichment of the U(A/U)UU motif. Furthermore, no other sequence motifs were determined to be enriched on more than a small (under 10%) subset of the differentially expressed genes in AVM and V lines. Therefore, the mechanism of transcript abundance regulation that is responsible for these changes remains an open question.

Relative mitochondrial minicircle class abundances correlate with degree of virulence

In addition to changes to the nuclear genome, changes to the mitochondrial (kinetoplast) genome resulting from culture growth can impact Leishmania spp. fitness [93,94]. In this work, we found that the kinetoplast DNA region containing the protein-coding and rRNA-coding loci (the maxicircle CR) copy number and sequence is essentially identical in the L. major lines under study. However, the abundance of mature, translatable mitochondrial transcripts could play a role in a successful survival in L. major hosts. While expression changes to some mitochondrial mRNAs can be directly measured, others require specialized analysis. This is because most mitochondrial primary transcripts undergo massive RNA editing with uridine insertions and deletions to generate their translatable open reading frames [77,95–97]. This type of RNA editing reaction results in far more partially- than fully-edited, translatable products [98–100].

Here, we compared the overall transcription levels of all maxicircle-encoded genes and detailed editing maps of transcripts requiring editing. No significant differences between the four LV561-derived parasite lines were observed at the general transcription level (S2 Fig) or in the RNA editing of specific sites, with one exception. The surprising exception is the G4 transcript (S3 Fig), the product of which has not yet been definitively attributed a function [101]. We observed G4 to have a low overall transcription level and a lack of editing in the AV line, thus, the protein product of G4 is unlikely to be generated in these parasites. In contrast, G4 was more abundant and intensively edited in the V and AVM lines, while it was slightly less intensively edited in AVS cell line (S3 Fig). In contrast, extent of editing and the editing patterns of RPS12 and ND3 mRNAs were consistent across the lines. The changes in G4 editing were the only observed cell line-specific differences to kinetoplast mRNA abundances, editing levels, or specific editing events that could impact L. major mitochondrial function in hosts.

In addition to the major protein machinery necessary to execute trypanosomatid mitochondrial RNA editing [95], hundreds of guide (g)RNAs are also needed to provide the specificity of the placement of insertions and deletions [102,103]. These gRNAs are encoded on many small circular mitochondrial DNA molecules that are independent of the larger molecules encoding the traditional elements of the mitochondrial genome. These “minicircles” are multicopy and organized into classes of near-identical molecules [45,104–106]. A potential regulator of gene-specific editing and overall mitochondrial genome informational content are the relative abundances of molecules of each class. We analyzed whether there were differences in the abundances of minicircle classes. Indeed, relative copy numbers of minicircles among classes were changed in the AV line compared to the numbers in the original V line (Fig 3). Notably, passage in mice and sand flies resulted in a close restoration of the relative abundances of the AVM and AVS line minicircles to what was observed in the V line, with some cell line-specific exceptions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Minicircle sequence classes abundance by category.

Heatmaps display copy numbers of minicircle classes measured by PRKM values divided into 7 abundance categories of equal size. Minicircle classes are sorted by decreasing abundance in the V sample. When relevant, we indicate with labels or triangular markers, the mRNA with homology to the gRNA encoded by this minicircle class. Blue, RPS12 gRNA encoded; red, ND8 gRNA encoded. (A) Comparison of V-AV-AVM lines; (B) comparison of V-AV-AVS lines.

https://doi.org/10.1371/journal.pntd.0014387.g003

Notably, the observed lack of differences in editing patterns of cryptogenes make it unlikely that, in this case, minicircle class copy number mechanistically impacts editing in a direct or obvious manner. Nevertheless, we analyzed whether minicircle copy number abundance profiles were co-regulated based on the gRNA they encoded. This analysis requires computational reconstruction of the fully-edited open reading frame sequences of cryptogenes, to identify the edited mRNA corresponding to each minicircle’s encoded gRNAs. RNA-seq files were generated from total mRNA rather than mitochondrial-enriched mRNA. Therefore, mitochondrial read fractions were too low for reconstruction of a complete open reading frame for G4, the only cryptogene with evidence of differential editing among the LV561-derived lines. Reconstructions were possible for ND8 and RPS12, however. We identified 9 and 6 minicircle-encoded gRNAs involved in their editing. As was the case with the total minicircle population, these minicircle class subsets also showed a general trend of minicircle abundance reversion when comparing V, AV, and host-passaged lines. Specifically, 7 out of 15 tracked minicircles returned to their original abundance category in the V–AV–AVM/AVS comparison, while 12 of 15 showed a clear tendency toward reversion (e.g., “high–low–high” or “low–high–low” patterns).

Despite trends for copy numbers of minicircle classes of ND8 and RPS12 editing cascades to reflect parasites’ virulence status, it is not at all clear that the alterations either promote or restrict editing. Not only were ND8 and RPS12 editing patterns and levels among the L. major lines themselves indistinguishable, but minicircles contributing gRNAs to these editing cascades did not exhibit a consistent directionality of change when compared among the parasite lines. Together, these observations suggest that minicircle abundance is not directly linked to mitochondrial gene expression or RNA editing efficiency. It is, thus, difficult to know what the biological outcome of these changes would be, or what might mechanistically drive minicircle class abundances. The mechanism for host passage-dependent maintenance or perturbation of minicircle class copy number warrants further investigation.