Introduction

“In the future attention undoubtedly will be centered on the genome, and with greater appreciation of its significance as a highly sensitive organ of the cell, monitoring genomic activities and correcting common errors, sensing the unusual and unexpected events, and responding to them, often by restructuring the genome”—Barbara McClintock’s Nobel Lecture in 19831.

A lot has been learned since 1983, and numerous genomes have been sized, sequenced and analyzed. Yet, many questions regarding genomes remain unanswered, the most fundamental potentially being: why do eukaryotic genomes vary so much in size? While complexity appears to correlate with minimum taxon genome size, the actual genome sizes bear no straightforward correlation with eukaryotic organismal complexity, even among closely related taxa, but are increasingly investigated as a trait subject to natural selection and consequently of relevance to studies of ecology and evolution2,3,4. Selective pressures in copepods have been posed for age at first reproduction in predation intense environments, resulting in smaller genome sizes5, as well as selection for larger bodies and genome sizes in cold environments6,7. Causal links between the ‘bulk’ DNA amount, cell division rate, and cell volume, as explained by the nucleotypic hypothesis8, may underlie relationships between these cellular parameters and organismal development rates and body size, especially in the copepods which possess eutely9.

While genome size does not appear to govern organismal complexity, some relationships appear to be general: genome size often correlates with the proportion of noncoding, or repetitive, DNA in the genome8,10, cell size2,11 and growth rate12. Furthermore, the evolutionary importance of repetitive elements (mainly transposable elements—TEs) in lateral gene transfer13 and generation of new phenotypes14 is becoming increasingly apparent. This is well illustrated by TEs being responsible for more than 50% of the phenotypes emerging in Drosophila laboratory strains15 and playing a role in adaptive evolution16,17. At the same time, it must be realized that species specific effects may affect genome sizes in ways that appear to be inconsistent with the general trends: for example taxon specific allocation of phosphorus to RNA rather than nonessential non-coding DNA may result in a selection for compact genomes in phosphorus limited environments18,19. As the role(s) of noncoding and repetitive DNA become better understood, the importance of knowing to what extent this component of the genome has been accurately included in the assemblies and annotations becomes increasingly clear. This does not contradict the fact that partial genome assemblies may be both of high quality and immense value.

The salmon louse (Lepeophtheirus salmonis, Krøyer 1837) is a marine parasitic copepod of large economic and ecological importance20,21. It belongs to the order Siphonostomatoida and is found on salmonid fishes in the northern hemisphere. There are two L. salmonis subspecies separated by approximately 5 million years of evolution, L. salmonis salmonis (Krøyer, 1837) inhabiting the North Atlantic and L. salmonis onchorhynchii21 inhabiting the Northern Pacific22,23,24. These parasites alter the physiology, disease susceptibility, growth rates, and behavior of their salmonid hosts25,26,27 and inflict large economic losses28. As the salmonid aquaculture has expanded to the extent that farmed salmonids outnumber wild salmonids by 2–3 orders of magnitude in some regions in the North Atlantic, the salmon lice populations have increased in parallel and currently inflict significant economic and ecological challenges28,29. The combined societal and ecological impacts of L. salmonis have spurred intense research and tool development, including modelling to assess ecological risk30,31, development of methodology for surveillance25,32, studies of population genetics33,34,35, resistance and resilience development against delousing agents36,37 and molecular biology38,39,40,41. As a result, the salmon louse genome has been sequenced several times using various sequencing platforms, and independent genome assemblies have been made41,42—including two chromosome level assemblies.

As DNA sequencing technologies have advanced and improvements in the identification and annotation of noncoding DNA have gradually followed, there is a growing awareness that genome assembly methods sometimes fail to correctly reconstruct repetitive regions and noncoding DNA [43,44]. Traditional quantitative cytogenetic methods such as flow cytometry and Feulgen microdensitometry are recognized as being reliable with respect to estimating the true amount of nuclear DNA contents and providing estimates of total genome size45,46. Thus, genomic assemblies and cytometric methods possess different strengths with respect to the kinds of information they provide. When these two approaches are applied in combination, they are likely to either validate the independent estimates or provide direction as to seeking explanations for the discrepancies. In light of the importance of the L. salmonis genome it seems prudent to use cytometric methods to validate sizes estimated from genome sequencing.

In the present study of the salmon louse, we compare genome size estimates based on two quantitative cytometric methods (flow cytometry and Feulgen image analysis densitometry) performed on multiple samples with unpublished cytometric measurements and estimates based on whole genome sequencing. Where discrepancies are apparent, we propose explanations and a path forward for resolving them. Additionally, as this is the first published estimate of the genome size of a parasitic copepod, the genome size of this siphonostomatoid is discussed in relation to the free-living copepods in the orders Cyclopoida, Calanoida, and Harpacticoida.

Results

Sequencing—and assembly based genome size estimates

The salmon louse genome has been sequenced several times using various sequencing platforms, and six independent genome assemblies have been made—including two chromosome level assemblies (Table 1). The resulting assemblies appear consistent in structure, as revealed by linkage analyses34,41,47, and sizes (Table 1), collectively suggesting the salmon louse genome size to be approximately 600–700 Mbps.

Table 1 Key statistics and sequencing platforms for salmon louse assemblies available March 2021.
Full size table

To evaluate the congruence of the assemblies, while not requiring them to conserve synteny, we created libraries of 240 bp synthetic reads from each of the assemblies. These synthetic reads were then mapped to each of the published assemblies using BLAST. The results show that the assemblies are close to interchangeable in terms of their sequence composition (Table 2) and that the differences in the sequence captured by the different sequencing technologies are minor.

Table 2 Consistency of assemblies.
Full size table

It is well known that genome assembly sizes may deviate significantly from the actual genome size10 and additional sequence-based genome size estimates were therefore produced. First, k-mer analysis was performed using Jellyfish51. The published GLW4 dataset originating from inbred salmon lice41 and word sizes (k-mer lengths) of 21–31 yielded estimated genome sizes of 976–1017 Mbps (modal k-mer coverages: 20–17). Repeating the analysis using previously published data from wild salmon lice34 and using the word lengths of 29 and 31 yielded estimated genome sizes of 1086 and 1015 Mbps (modal k-mer coverages: 55 and 53).

Second, sequencing reads were mapped to the LSalAtl2s genome using BWA52. For each library, modal coverage (M) was extracted, and assumed to be representative of diploid coverage. All coverages for the genome were then summed and divided by the modal coverage to estimate the genome size. Under the assumption that repeat sequence occurring N times in the genome would have a coverage distribution centered on N*M, each location in the repeat would be counted N times. Seven inbred salmon lice libraries41 were used, with modal coverages of 3–23x, and resulting in genome size estimates of 791–1184 Mbps, with low coverage libraries giving the highest size estimates. The analysis was repeated with six libraries from wild lice34, giving coverages of 5–17×, and size estimates of 845–1073, again with low coverage libraries leading to higher size estimates.

Genome size estimates based on FCM

The nuclear DNA contents of somatic (2C values) and gametic cells measured by flow cytometry (FCM, Runs 1–5) in gametic, naupliar or adult stages of Atlantic L. salmonis salmonis are reported in Table 3. Figure 1 shows representative fluorescence (FL histograms of the above cells analyzed together with reference standards). Overall, fluorescence histograms of propidium iodide (PI) -stained somatic cells or gametes indicated good resolution levels with coefficients of variation (CVs) in the range of 1–5% (Table 3).

Table 3 Nuclear DNA contents of somatic and gametic cells of L. salmonis salmonis as measured using flow cytometry.
Full size table
Figure 1
figure 1

Representative fluorescence histograms of PI-stained somatic and gametic cells of L. salmonis salmonis. (a) Run 1, somatic cells obtained from whole body squash of nauplii; (b) run 3, somatic cells obtained from cephalic regions of adult females; (c) run 4; oocyte and sperm samples; (d) run 5, somatic cells obtained from cephalic regions of adult males and females. Samples analyzed using chicken and human cells as internal reference standards. Fluorescence (FL) peaks values on the X-axis are reported in arbitrary units expressed as Propidium Iodide (PI) fluorescence channel numbers (FL value). Values on the Y-axis (counts) refer to the number of nuclei counted per channel. CEN chicken erythrocytes nuclei, MNCs human mono-nucleated cells.

Full size image

The estimated 2C DNA contents of somatic cells in naupliar stages of Ls Tromsø were similar across replicates (FCM Runs 1 and 2), averaging 3.08 pg DNA per nucleus when using chicken (CEN) as an internal standard (Table 2) and did not vary significantly when both chicken and human (MNCs) standards were used simultaneously (FCM Run 2). Nauplii cannot visually be assigned a sex and thus it cannot be known for certain what sex ratio occurred in these samples although sex is genetically determined with a 50:50 ratio41.

The average nuclear 1C DNA contents of Ls Gulen sperm cells (FCM Run 4) were 1.68 and 1.69 pg DNA per nucleus, and did not differ when estimated using chicken or human standards. The estimated 2C value of oocytes (FCM Run 4) averaged 3.33 pg DNA per nucleus with no significant difference between values based on the two standards. A derived 2C value of sperm cells (twice the 1.69 pg DNA per sperm cell) does not differ significantly from that of the unfertilized oocytes, 3.33 pg DNA per nucleus (Table 2).

The nuclear DNA contents of males and females in both wild caught and laboratory reared Tromsø strains were compared in Run 5. The 2C DNA content of somatic cells averaged 3.01 and 3.19 pg DNA/nucleus in female and male specimens of a laboratory strain, respectively. The same trend was observed when analyzing somatic cells of wild specimens; the 2C values averaging 3.08 and 3.19 pg DNA per nucleus in females and males, respectively. Overall, within this FCM run male genome size estimates were consistently larger (ANOVA, P = 0.0001) than female genome size estimates whereas, the 2C DNA contents of somatic cells within a sex did not differ significantly between salmon lice of different origin and with no interaction between the two factors. Despite the lack of a statistically significant difference between male and female adults of laboratory reared Ls Tromsø salmon lice recorded in FCM Run 3, the DNA content of these somatic cells did not significantly differ from the laboratory and wild caught adults measured in Run 5, when comparisons were made within a sex.

Feulgen image analysis densitometry (FIAD)—nuclear morphologies of somatic tissues

The squashed somatic tissues possessed a variety of nuclear morphologies, from dense to diffuse (Fig. 2), which yielded a corresponding variety of values of integrated optical density (IOD). Such heterogeneity in nuclear morphology was not observed in either chicken or trout erythrocyte standards, as they were comprised solely of erythrocytes (Fig. 2a,b). We made decisions to select for measurement nuclei with intermediate morphologies that possessed a granular and slightly diffuse appearance (Fig. 2c,d). Nuclei that were very densely staining in appearance, or compact (Fig. 2e), yielded IOD values at the lower end of the range, possibly due to DNA compaction. Nuclei with a very diffuse appearance and sometimes with nuclear membranes possessing uneven edges that might indicate partial degradation (Fig. 2f), tended to possess IOD values at the higher end of the range. The corresponding 2C values were 2.1 pg DNA per nucleus for densely stained nuclei to 3.4 pg DNA per nucleus for diffuse nuclei.

Figure 2
figure 2

Photomicrographs of squash preparations of somatic nuclei of cephalothorax of L. salmonis salmonis and erythrocytes of hen and trout used as standards, stained with Feulgen reaction for DNA. (a) Nuclei of erythrocytes of G. domesticus; (b) nuclei of erythrocytes of O. mykiss; (c) representative nuclei, indicated by arrows, of male L. salmonis salmonis slide 64 whose measurements are given in Table 4; (d) representative nucleus, indicated by arrow, of female L. salmonis salmonis slide 63 whose measurements are given in in Table 4; (e) nuclei of male L. salmonis salmonis slide 64 with high level of DNA compaction; (f) nuclei of male L. salmonis salmonis slide 64 with highly diffuse morphology. Scale bars represent 10 µm.

Full size image

Genome sizes based on FIAD

Based on Feulgen image analysis densitometry of the Ls Gulen laboratory strain and the chicken standard, the average somatic nuclear DNA contents of individual adult males (2.86–2.93 pg DNA per nucleus) were consistently larger than the average nuclear DNA contents of individual females (2.64–2.78 pg DNA per nucleus) (Table 4). Despite the consistency of the trend, the average of three individual females, 2.70 pg DNA per nucleus, did not differ significantly from the average of three individual males, 2.90 pg of DNA per nucleus (Table 4). The two sample t tests were based on 3 individuals per sex, rather than 120 nuclei per sex (Table 4; the N values for LsGulen), to avoid pseudoreplication.

Table 4 Nuclear DNA contents of somatic cells of individual adult L. salmonis as measured using Feulgen image analysis densitometry.
Full size table

A single wild caught adult female collected from a Maine population possessed nuclei in an appendage that were especially well suited for measurement as they were well isolated and lacked any visible background stain. The mean value of these nuclei, 3.07 pg DNA per nucleus, does not differ from the wild female that was caught near Tromsø nor from the laboratory reared Ls Tromsø strain (Tables 3, 4). It should be noted that the Maine specimen was squashed and stained as above, but prior to employing the freeze cracking technique, and thus measurements were restricted to those few nuclei in the spine of a swimming leg that possessed a granular and slightly diffuse appearance, and were also isolated and surrounded by clear background.

Comparison of genome size estimates based on FIAD and FCM

Estimated 2C genome sizes of male and female laboratory reared Ls Gulen adults obtained using FIAD and based on the chicken standard (2.90 and 2.70 pg DNA per nucleus, respectively, Table 4) are within 10% of the estimates obtained using FCM on the Ls Tromsø laboratory strain (Table 3), assuming values of 3.2 and 3.0 pg DNA per nucleus for males and females, respectively, for the Ls Tromsø strain). Each slide containing a population of nuclei from a single adult in the FIAD analyses contained some values that overlapped with estimates obtained using FCM; however, these higher FIAD values did not equal the central tendency of values obtained using FCM. The average nuclear DNA content obtained for the Ls Gulen adult females using FIAD, 2.70 pg DNA per nucleus, is within the 15% range of the value based on oocytes 3.26 pg DNA per nucleus (FCM run 4).

Discussion

L. salmonis assemblies are consistently approximately 700 Mbps suggesting that the L. salmonis 1C genome may be of approximately the same size, or possibly larger if the assemblies include substantial collapsed repeated regions. This approximate size was readily accepted by the salmon louse research community, as the first cytometric based estimate of a North Atlantic population was 0.58 pg (≈567 Mbps)53. Thirteen years later a Bay of Fundy population measured in the same lab was estimated to be 0.96 pg (≈ 939 Mbps)54. Yet in the present study the L. salmonis subsp. salmonis genome size estimates range from 1.3 to 1.6 Gbps when determined by two independent cytometric methods being applied to three laboratory strains and wild salmon lice from two locations. Sequence based extrapolations, in contrast, yield estimates ranging between 0.8 and 1.2 Gbps. In attempting to identify the factors responsible for the different estimates of L. salmonis salmonis genome size, we emphasize the importance of harnessing complementary approaches to estimate genome size and where discrepancies exist, not to disregard potentially ‘missing’ portions of DNA which may play an important role in adaptation. Additionally, while it is a common bias to interpret ‘old measurements’ as wrong when they disagree with new measurements, we remain open to a plethora of explanations to reconcile older and present study findings based on cytometric methods.

Genome assembly sizes are known to be unreliable predictors for genome size. Highly conserved repeats in combination with read errors can be difficult to resolve, and assemblies can have repeated regions collapsed into one sequence or have multiple copies of the same genomic region. More precise estimates can be achieved by examining the sequence reads. We have used two approaches, one using k-mer statistics and another based on mapping statistics against a reference assembly. These methods point to a genome size of 800–1200 based on mapping, and 1000–1100 Mbp based on k-mers. One partial explanation for the variability can be unmapped reads (approximately 5% of the reads). If they represent sequence not present in the genome assembly, these genome components will be omitted from the mapping estimate, but included in the k-mer estimate. In addition, most of the sequence data is from female salmon lice, and both approaches will count the average of the haploid sex chromosomes, not their sum.

Flow cytometry is a well-established method for nuclear DNA content analysis and characterization in experimental biology, and is increasingly being used due to its rapidity, precision, and reproducibility. Feulgen image analysis densitometry similarly has experienced a resurgence in its use, partially due to its affordability and applicability when the number of available nuclei to measure is small. Interpretations of genome size estimates based on FCM and FIAD require careful consideration of the advantages and disadvantages of each method for the species and specific tissues under consideration. Explanations of differences among cytometric measurements in general, as well as those in the present study with unpublished estimates53,54, performed on L. salmonis could include tissue compaction levels that can cause estimates to vary two-fold, misidentification of haploid and diploid cells, and misidentification of species, the latter of which seems quite unlikely as species identifications were provided by external expert collaborators. Not to be discounted is the possibility of real variation in genome size among populations. Jeffrey’s54 review highlights the issues of concern, particularly the chemistries and tissues measured most applicable to the present study, and are discussed in Supplemental Methods S1.

Measurements of nuclear DNA contents of naupliar and/or adult stages of two laboratory strains and wild caught L. salmonis salmonis based on FCM and two laboratory strains and one wild caught population of adults’ stages based on FIAD estimate 2C somatic nuclear contents to be 3 ± 0.3 Gb. Halving the FCM values to obtain the 1C amount and directly measuring sperm DNA content yielded values ranging from 1.47 to 1.65 Gb (Table 3). Gametic nuclei of L. salmonis salmonis contained one half the DNA of the somatic nuclei, and indicates a lack of chromatin diminution, a phenomenon in which 1C values cannot be estimated by halving somatic genome sizes55.

Male genome size was consistently slightly larger than female genome size, expectedly so due to erosion of the W-chromosome in the heterozygotic female56. There was also no evidence of mitoses in the adult somatic tissues, and therefore most adult somatic cells are suitable for genome size measurement9. Furthermore, we found no evidence of significant differences based on cytometric based comparisons of geographical (Norway, Maine), laboratory (Ls1a, Ls Gulen and Ls Tromsø) or wild caught (Norway, Maine) populations.

Genome size estimates of crustaceans based on FIAD are commonly lower than those based on FCM and estimates within 15% of one another are generally considered reliable54. Accordingly, the FIAD derived estimates of 2.7 and 2.9 pg DNA per nucleus for adult females and males respectively, are within 10% to 15% lower than the FCM based estimates, depending upon the particular comparison. The most likely explanation for the disparity between FCM and FIAD based estimates is that the laser detection of cells in a suspension used in FCM is less sensitive to background noise, DNA compaction and other conformational changes in the chromatin sometimes encountered when measurements are based on the quantitative intercalation of the Schiff reagent among nucleotides as they sometimes are in squashed tissues in FIAD. These differences between FCM and FIAD based estimates of adults correspond to approximately 0.3 pg in a 2C nucleus, or ≈ 150 million base pairs in the 1C genome. We conclude that measurements obtained from FIAD and FCM were internally consistent and that the discrepancy between the results are well within the boundaries expected from earlier studies57. Since cytometric measurements are based on direct observations and the derived estimates are both consistent within the methods and discrepancies in accordance with methodological expectations we regard the cytometric results, collectively indicating a L. salmonis salmonis genome size of approximately 1.5 Gbps, as the most reliable measurements available.

The above suggests that sequence-based methods underestimate the genome size by approximately 33%. The mapping approach is sensitive to errors in assembly completeness, uniformity of library and sequencing coverage, mapping accuracy and modal mapping estimation. Alas, our analysis did not reveal which of these factors were the more likely to cause the apparent mapping-based underestimation of genome size. As an alternative to the mapping-based estimates we applied the widely used k-mer approach which similarly appeared to miss approximately 33% of the genome size. The most plausible explanation may be that repetitive elements cause the k-mer approach to underestimate genome size as previously observed by Pflug and co-workers58. Similarly, genome assembly based on k-mer analysis of the lobster Homarus americanus is believed to be missing approximately 28% of the genome59. The salmon louse genome is among the crustaceans with highest occurrence of repetitive elements; ≈60% of the assembly annotated as repeats41, suggesting that such underestimation of size may not be implausible. In the present study sequence-based genome size estimates consistently provided substantially lower genome size estimates than cytometric measurements. Hence estimates should be regarded with caution until they have been confirmed by direct cytological measurements, as conducted here.

While our cytometric based estimates converge on a genome size of approximately 1.5 Gbps, earlier cytometric measurements disagree: a 1C = 0.58 pg DNA per cell (567 Mbp) estimate by Gregory53 was based on material of Norwegian L. salmonis salmonis from a discontinued lab strain supplied by Professor Frank Nilsen of the University of Bergen in the early 2000’s (pers. comm. Frank Nilsen) and a later 1C = 0.96 pg DNA per cell (939 Mbp) estimate by Jeffery54 which was based on material collected in the Bay of Fundy in the 2010’s and supplied by Professor Elizabeth Boulding from the University of Guelph. The most parsimonious explanation of the discrepancy is to consider the earlier measurements erroneous. However, the measurements were made by respected authorities in the field, with whom our measurements on other species have agreed, and the salmon lice were supplied by field experts. We therefore believe their measurements are likely to be correct. Our results show that the DNA content of somatic cells is twice that of gametes; therefore, somatic chromatin diminution, found in some copepods, is an unlikely explanation for the observed differences. We therefore consider the discrepancy in results to indicate that large variations in the salmon louse genome size occasionally arise. Such variability would not be unprecedented in copepods and we suggest that this is addressed in L. salmonis by cytological measurements of genome sizes of both established strains and wild specimens covering their entire geographical range. Based on FCM geographically based intraspecific variations in genome size of magnitudes 1–9 pg (corresponding to a difference of ≈0.5 to 4.5 Gbp in the haploid genome) have been reported in the marine calanoid copepods Calanus glacialis, Calanus hyperboreus, and Paraeuchaeta norvegica populations inhabiting the High Arctic and Southern fjords of Norway6. Based on FIAD a difference of 1 pg between German (Schöhsee “house” lake of current Max Planck Institute for Evolutionary Biology in Ploen) and Lake Baikal populations of the freshwater cyclopoid Mesocyclops leuckarti was reported60,61. Furthermore, the genome size of the North Sea population of the calanoid Pseudocalanus elongatus decreased after being reared in the laboratory for 96 generations62. It would not be surprising to encounter additional examples with intraspecific genome size differences in other copepods.

The somatic 2C nuclear DNA contents of L. salmonis, as estimated using FCM and FIAD in laboratory and wild populations (2.7–3.2 pg DNA per nucleus and corresponding to genome size of 1.33–1.56 Gbps) are at the lower end of the range of all published values of free-living copepods, which vary more than 300-fold from 0.20 to 64.46 pg DNA per nucleus, corresponding to genome sizes ranging from 195 Mbps to 63 Gbps (Fig. 3). Relative to the range of values in free-living cyclopoids, L. salmonis salmonis has an intermediate genome size, although its genome size is larger than the majority of cyclopoid species estimates. Relative to the range of cytometric based values in calanoids, L. salmonis salmonis is comparable to the smaller genomes, with the majority of calanoid species possessing larger or far larger genomes.

Figure 3
figure 3

Somatic nuclear DNA contents of copepods. Each species is represented by a vertical bar. Estimates based on cytometric methods, genome assemblies, and mapping are presented in orange, blue, and dotted black, respectively, for the siphonostomatoid L. salmonis salmonis. The black dotted lines depict a range of values obtained for L. salmonis salmonis using mapping. FIAD and FCM measurements depicted by green vertical lines for L. salmonis salmonis are from Gregory53 and Jeffery54, respectively. FIAD (orange) and genome assembly (blue) estimates are shown for the harpacticoid T. californicus63,64. FIAD (orange) and genome assembly (blue) estimates for the calanoid E. affinis are the two left most vertical lines in the distribution of Calanoida65,66. The adult somatic genome size estimates of Cyclopoida and Calanoida (orange) were mostly acquired using FIAD and are from Gregory53. All estimates are presented as 2C values found in the soma because some cyclopoid species and at least one calanoid species possess embryonic chromatin diminution, in which germline genome sizes (conventionally noted as 1C) substantially exceed the sizes of somatic values, thus violating the appropriateness of halving 2C estimates in the soma to obtain 1C estimates of the germline.

Full size image

We speculate that the high abundance of repetitive regions in the L. salmonis salmonis genome facilitates the observed variability in genome size by serving as a “size accordion” where the repetitive elements may in- or decrease in copy numbers. This model has earlier been suggested for birds and mammals although the increase in these groups seems to be compensated by a corresponding loss of DNA segments elsewhere, resulting in rather constant genome sizes67. We further suggest that a lower limit to the genome size likely exists, in a manner similar to what is indicated for the rotifer Brachionus asplanchnoidis in which genome size varies by a factor of 1.968. A consequence of a possible variability in genome size is that the mapping and k-mer based size estimates should not be considered conclusively discredited as they are based on sequence information from different wild samples or strains. For the Ls1a strain that was measured both by FIAD, FCM and included in the mapping and k-mer based analysis, the discrepancy may be assumed to be genuine, although the samples for cytometric analyses were sampled more than 10 generations after the samples for sequencing were obtained. Hence the L. salmonis salmonis LSalAtl2s genome assembly41 appears to fail to resolve ≈50% of the genome that is captured in the cytometric measurements. There are several possible sources that may contribute to this discrepancy: failure to isolate some regions of DNA, failure to capture all DNA in sequencing libraries or sequencing reactions, errors introduced during the bioinformatic analyses and specific challenges caused by TE’s and DNA repeats.

The fact that the six independent genome assemblies are congruent in content despite originating from DNA purified from multiple origins of L. salmonis in different laboratories and sequenced using various sequencing platforms (Illumina, 454 pyrosequencing, Oxford nanopore, PacBio and Sanger) may suggest that no parts of the genome are systematically missed by the purification and sequencing protocols applied. However, systematic omission of genetic regions across isolation protocols or biased representations of repetitive regions cannot be excluded and would yield biased or incomplete genome representations. A similar distorting effect may be introduced during downstream bioinformatic analysis, for instance, by collapsing repetitive regions44. Distinguishing between a bias in sequencing representation and bioinformatic artifacts as the cause of underestimating repetitive regions is challenging. PacBio and Nanopore sequencing platforms produce long reads that are commonly suggested as a tool to address repetitive regions. However, the two most recent L. salmonis salmonis assemblies were produced using PacBio (UStir_LSAA, Table 1) and Oxford nanopore combined with Illumina (UVic_Lsal_1.0, Table 1) sequencing. These did not deviate significantly in content or size from earlier assemblies and hence did not resolve the question of the missing DNA. Since there are no indications that specific regions are missed in any of the assemblies (Table 2) it is possible that the majority of the ≈800 Mbp of “missing” DNA (cytometric based genome size minus genome assembly size) in the samples measured by FIAD and FCM in the present study is comprised of TEs and DNA repeats that are not accurately captured in the assemblies. The fact that the long read based assemblies do not, at least partly, resolve the challenge may indicate that mini- and microsatellites may be dominating the missing fraction since these are considered more prone to incorrect rendering by long read sequencing than TE’s44,69. The existing ≈700Mbp salmon louse assemblies consist of ≈60% repetitive regions41 indicating that ≈300 Mbp consists of non-repetitive regions. Hence, the ≈1500 Mbp genomes measured in the present study suggestively consist of ≈80% (≈1200 Mbp) repetitive and/or other kinds of DNA that were otherwise uncaptured regions and ≈20% (≈300 Mbp) non-repetitive regions.

Challenges in capturing repeated regions are likely exacerbated in mid to large size genomes. At the lower range of genome sizes in copepods are the tidepool harpacticoid copepod Tigriopus californicus and the estuarine calanoid copepod Eurytemora affinis whose 2C genome sizes estimated by FIAD are 0.5 and 0.6–0.7 pg DNA per nucleus, respectively63,65. Both genome assemblies were significantly smaller by ≈20% (≈400 Mb for T. californicus, ≈495 Mb for E. affinis for 2C values) which attributed to the inability to sequence all of the repetitive DNA64,66 (Fig. 3). While L. salmonis salmonis has a genome size at the lower range of the distribution of copepods (Fig. 3), it is still two- to three-fold larger than T. californicus and E. affinis. A significant proportion of repetitive regions of the salmon louse genome consists of transposable elements, or of unclassified repeated motifs that may in time be annotated as TEs41. Precisely identifying the portion of the genome that is comprised of TEs and their composition is of interest as TEs are increasingly viewed as drivers of genome plasticity that facilitate the rise of new phenotypes, such as acquiring insecticide resistance in fruitflies70,71. It may be speculated that the high occurrence of repeated regions, including TEs, in the salmon louse genome may have contributed to its documented ability to develop resistance towards new medicinal treatments despite a low diversity of genes typically associated with detoxification and stress response37,41,72,73,74. If this is the case, the use of medicines in salmon farms that harbor the majority of sea lice in parts of the North Atlantic may have positively selected for high numbers of TEs and hence for a larger genome size. More evidence to support this hypothesis may be provided by evaluating the DNA repeat content in specimens of historical populations that existed prior to the introduction of drugs in aquaculture or present-day populations with little or no known exposure to such drugs. The effect of TEs on drug resistance in L. salmonis is an important avenue of study that merits further attention.

Materials and methods

Assembly analyses and sequencing-based genome size estimates

To evaluate the congruence of the six assemblies available in public databases (Table 1), while not requiring them to conserve synteny, we made a script that converted the individual assemblies into non-overlapping 240 bp synthetic reads thus generating 6 synthetic read libraries. These synthetic reads were then mapped to each of the published assemblies using BLAST with the following command line: blastn -num_threads 16 -evalue 1e-10 -outfmt 6 -num_alignments 10 -penalty -1 -reward 1 -gapopen 3 -gapextend 2. Since the same fragment may be mapped multiple times we calculated the percentage of synthetic reads that mapped and the fraction of the mapped reads that mapped with > 95% identity (Table 2).

The genome size was estimated from sequencing and assembly data using three different approaches: modal mapping extrapolation, k-mer analysis and single copy gene mapping extrapolation. The assembly based estimates were derived using the LSalAtl2s assembly41. The LSalAtl2s assembly was compared to other available assemblies (Table 1) to reveal potential regions that are missing.

Modal mapping extrapolation is based on the assumptions that populations of non-repetitive DNA sequence reads follow a Poisson distribution and this makes up the majority of DNA. By finding the modal coverage and dividing the total number of sequenced bases mapped by this number, we can estimate the genome size. This is done using the Lande-Waterman formula; G = NL/C, where G is the genome size, N is the number of reads, L is the average read length, and C is the modal coverage. The modal coverage was determined by plotting the number of sites against nucleotide coverage and identifying the peak value. To facilitate this, sequence reads from the GLW4, GLW13 and GLW16 libraries41 previously published were mapped against the LSalAtl2S genome assembly using Samtools52.

K-mer analysis is based on the assumption that the possible words of a certain size in a genome (k-mers) increase with the size of the genome. A genome consisting primarily of non-repetitive DNA regions will generate an approximately random population of k-mers, and the diversity of k-mers in a population of reads can be used to estimate the genome size. In the present study k-mer analyses were performed using Jellyfish51 and sequence reads from the GLW4 library derived from a laboratory strain of L. salmonis salmonis41 and libraries derived from L. salmonis salmonis collected in the field.

Nuclear DNA content analysis by flow-cytometry (FCM)

Field and laboratory populations

Specimens of L. salmonis salmonis were obtained from several sources: (1) Wild adult males and females were collected from naturally infected farmed Atlantic salmon held at the sea cage facilities of the Aquaculture Research Station in Tromsø (FCM Run 5); (2) An outbred laboratory strain, Ls Gulen, was derived from adults collected in Ls Gulen (Norway) and reared at the Salmon Louse Research Centre in Bergen (FCM Run 4); (3) The Ls Tromsø laboratory strain was established by crossing adults from the Ls Gulen strain with a partially outbred strain, Ls Oslofjord, originating from specimens collected in Oslofjord (Norway) and reared at the Aquaculture Research Station in Tromsø (FCM runs 1–3).

Collection of samples and tissue preparation

Newly hatched nauplii were obtained from gravid females, crushed in cold citrate buffer75 containing 5% dimethyl sulfoxide (DMSO), filtered through a 30 µm nylon mesh and deep-frozen until use. Sperm and eggs were collected from the testes and the genital segment prior to fertilization, respectively, and the resulting samples briefly kept on ice prior to analysis. Somatic (cuticular and subcuticular) tissues obtained from the cephalothorax (Supplemental Materials and Methods Fig. S1) of adult wild or laboratory specimens were crushed, and treated in the same way as the newly hatched nauplii. Specimens were squashed onto slides according to Clower and co-workers55 except that a freeze-cracking technique was added.

Flow cytometry analysis

Aliquots of target (sea lice) and internal reference (male human and/or chicken) cells were analyzed using Propidium Iodide (PI) as fluorescent stain following previously reported methods76. The mean DNA content of 5000–10,000 cells per sample was measured with a CyFlow®Ploidy Analyser equipped with a green laser.

Nuclear DNA contents of target species were estimated in relation to an assigned 2C value of 7.00 pgDNA/nucleus for human leukocytes and 2.50 pg DNA/nucleus for chicken erythrocytes53 according to the formula:

$$ {\text{Target species nuclear DNA content }}\left( {{\text{pg}}} \right) \, = \, \left( {{\text{Mean FL value of the sample }}\times{\text{ reference 2}}C{\text{DNA content}}} \right) \times \left( {\text{Mean FL value of the reference}} \right)^{{ - {1}}} $$

Feulgen image analysis densitometry

Field and laboratory populations

Genome size measurements were obtained from each of three adults (females) of the Ls1a laboratory strain described elsewhere77, whose ovaries served as the source material of DNA used in the nanopore DNA sequencing and six adults (three males and three females) of L. salmonis salmonis from Ls Gulen laboratory strain used in the FCM studies. A single adult female was collected from the wild in Copscook Bay, Maine in 2018. Specimens were immediately preserved in undenatured > 95% alcohol.

Feulgen staining and scanning microdensitometry

All slides were squashed and stained with Schiff reagent according to previously reported methods46,55, with few modifications. Nuclei were measured using a Zeiss Axioscope A1 equipped with a 63X oil objective and a Qimaging Bioquant PVI CCD camera. Scanning microdensitometric software (Bioquant Image Analysis; Bioquant Life Sciences 2018 program) was used to determine the IODs of the nuclear DNA contents of individual somatic nuclei. We selected for measurement only nuclei that possessed a granular and slightly diffuse appearance and lacked visible pink background; these nuclei were found mostly at the perimeter or outside the carapace (Fig. 2c,d). Nuclei with relatively small areas and dense staining indicating DNA compaction (Fig. 2e) or very diffuse and large areas (Fig. 2f) are less likely to provide accurate measurements. The Bioquant software used to measure IODs has a conservative estimate of resolution of 0.5 pg DNA per nucleus according to the manufacturer. The mean IOD value of the hen was used to convert the IODs of each L. salmonis salmonis specimen to picograms, using the following equation:

$$ {\text{Pg}}_{{\text{c}}} = \, \left( {{\text{pg}}_{{\text{s}}} /{\text{IOD}}_{{\text{s}}} } \right){\text{ x IODc}} $$

where pgc is the unknown amount of pg DNA per nucleus of L. salmonis salmonis pgs, 2.5 pg is the amount of DNA in the standard hen nucleus, IODs is the average IOD value of the hen, and IODc is the IOD value of L. salmonis. Photographs were taken at 100X magnification with a Nikon Eclipse Ti-2 microscope equipped with a PlanApo objective (N.A. 1.45) and QImaging DS RI2 camera.

Reference standards for conversion of integrated optical density (IOD) units to picograms (pg) included mutant white eyed female Drosophila melanogaster (0.40 pg DNA per nucleus), erythrocytes of hen Gallus domesticus (2.5 pg DNA per nucleus) and trout Onycorhynchus mykiss (5.2 pg DNA per nucleus), and leucocytes of male human Homo sapiens (7.0 pg DNA per nucleus) whose values were based on previous works78,79 and the Animal Genome Size Database53. The calibration curve computed for standards in the staining batch containing the Ls1a strain yielded an R2 = 0.997 (Supplementary Methods Fig. S2), indicating quantitative staining over a range of 0.40–7.0 pg DNA per nucleus. Only hen and trout standards (Fig. 2a,b) were used in the staining batch with Ls Gulen and Maine specimens.

The mean nuclear DNA contents are reported as 2C values in picograms (pg) and converted to gigabase (Gb) pairs (1 pg DNA = 0.978 Gb) for both FCM and FIAD derived estimates80.

Statistical analyses

Differences in nuclear DNA content of somatic tissues between LsTromsø nauplii (FCM Run 1 and 2) and Ls Gulen germinal (eggs and sperm) or somatic tissues of LsTromsø and Ls Gulen adult males and females (FCM Run 3–4), as well as those of Ls Gulen adults obtained using FIAD, were analyzed by Students t-test. Analysis of variance (ANOVA) was used to detect significant differences in nuclear DNA content of somatic tissue of adult wild caught and laboratory LsTromsø strain L. salmonis salmonis (FCM Run 5) using fluorescence (FL) PI values as dependent variable and gender and strain (laboratory or wild) as factors. In Exploratory Data Analysis (EDA), Grubbs’ test was used to detect presence of outliers and Levene’s and Shapiro–Wilk tests were used to test homogeneity of variances among groups. Statistical analyses were performed using IBM SPSS Statistics v.25 software. Differences were accepted as significant when P < 0.05. Data are reported as mean ± standard error (SE). The Shapiro-Wilks test applied to nuclei within each of 10 specimens measured using FIAD revealed no departures from normality. Differences between male and female genome sizes based on FIAD were tested using two sample, two tailed Student t-tests.

Ethical standards

The study has been planned an implemented and its results reported in accordance with ARRIVE guidelines (https://arriveguidelines.org). Sampling of parasites from infected fish were carried out in facilities approved by the Norwegian Food Safety Authority (Mattilsynet, FOTS). Fish were handled according to the Norwegian regulations for use of fish as laboratory animals (Norwegian Animal Research Authority) and all operations performed by approved personnel at the Tromsø Aquaculture Research Station (FOTS license nr. 110) and at the Institute of Marine Research in Bergen (permit nr. 2009/186329). The fresh chicken blood samples were provided by a professional veterinarian (D.C.R. Da Rocha Marques, Tromsø, Norway) and the human blood samples (isolated MNCs of anonymous donors) by the University Hospital of North Norway (UNN, Tromsø, Norway). The latter experimental work was done in accordance with relevant national guidelines/regulations and approved by Helse Nord RHF (https://helse-nord.no/ Tromsø, Norway,) via a contract for ‘Use of blood donor blood for purposes other than patient care’ stipulated between the Blood Bank/UNN and S. Peruzzi at UiT (contract nr. SJ1398/V2 of 18 November 2020). The experiments conducted at James Madison University were carried out in accordance with the National Science Foundations’ regulations for use of animals in experiments. Drosophila tissues and blood of chicken, trout and human were provided by research staff of James Madison University These standard preparations were made on a single day in the early 1990’s, stored in the dark at room temperature, and used in all subsequent staining procedures in GAW’s laboratory.