In recent years, there has been increasing interest in HGT, but most investigations have been performed in prokaryotes or in unicellular eukaryotes. In these organisms, HGT is considered a main driver of innovation, often associated with speciation 
. In multicellular eukaryotes, there has been less emphasis on HGT, partly because it is thought to occur on a much smaller scale 
, and partly because there are fewer well-annotated genome sequences available. Since de novo
whole genome assembly is still a significant challenge for complex organisms, particularly for the bdelloid rotifer with its unusual genome characterised by degenerate tetraploidy, divergence of formerly allelic sequences, and gene conversion between gene copies 
, we chose to assess HGT primarily at the transcriptome level. This study represents the first global analysis of the expressed genes in a bdelloid rotifer, A. ricciae
, and the contribution of horizontally acquired sequences to its transcriptome. The results reveal a remarkable degree of HGT in the bdelloid, with approximately 10% of identifiable, transcribed sequences deriving mainly from prokaryotes, but also from fungi, plants and algae, and protists.
The method for assessing HGT in the bdelloid transcriptome is novel, but follows principles currently recognised as the most rigorous, where sequence matching is coupled with phylogenetics 
. There have been relatively few such global analyses among the Metazoa that test for expression of horizontally acquired sequences, one example being in Hydra magnipapillata
, where seventy-one “gene models” apparently derive from bacteria, 70% of which were shown to be transcribed 
. For the bdelloid work, we introduced the HGT index, h
, which is calculated as the difference in bitscores between best non-metazoan and best metazoan matches in blast alignments, to give a measure of the “foreignness” of any sequence. We preferred the HGT index to the alien index (AI), developed previously for assessing foreign sequences in bdelloid subtelomeric regions 
and also used in the Hydra
, because h
is calculated from bitscores and is therefore not influenced by the sizes of the databases used to perform the blast screen. In contrast, if E-values are used, as for the AI, the score changes with database size. Additionally, an arbitrary constant must be included in the AI formula so that the index does not become infinite with identical matches to database sequences; this adjustment is unnecessary with the HGT index. Although showed that, whatever value of h
is chosen, there is a greater proportion of foreign sequences in the bdelloid than in other invertebrates, it is useful to adopt a threshold value to signify a foreign sequence. In principle, any sequence with h
>0 is more likely be foreign, but there will be uncertainty at values close to zero where non-metazoan and metazoan sequences have similar degrees of divergence from the test sequence. One technique for identifying a reliable threshold value of h
is to normalise the proportion of foreign sequences against the “background” levels found in other invertebrates. The greater proportion of horizontally acquired sequences in the bdelloid then becomes apparent above the minimum threshold level of h
required to confidently identify their foreign nature, as shown in . This was validated by phylogenetics, where possible (i.e. where matching metazoan counterparts exist), which showed that the vast majority of bdelloid transcript contigs with hU
≥30 did not cluster with metazoan sequences.
There are other technical considerations in any assessment of HGT. For example, we classified sequences as either metazoan or non-metazoan, and therefore any HGT from other animals (including other bdelloids) into the A. ricciae genome would be missed. Of course, there is no reason to believe that bdelloids are unable to acquire genes from other metazoans, or indeed from other rotifers; in fact, this might be more efficient than acquisition from microorganisms, since fewer changes to metazoan genes should be required before they become expression competent. Therefore, our approach is likely to give a minimum estimate of the extent of HGT in the bdelloid. Another factor that might influence this estimate is the approximately half of transcript contigs that showed no match with known sequences and therefore had to be excluded from further analysis. If all these sequences originate from vertical transmission into the bdelloid lineage, then this would reduce the estimate of HGT. However, there is no a priori reason to assume this: the proportion of foreign sequences in this non-matched set could be higher, lower or about the same as in the matched set. How the matched and non-matched sequence sets are defined could also potentially influence the proportion defined as HGT. We used 10−5 as a maximum value for a significant match when blast screening the transcript contigs against the databases and this gave 28,922 contigs in the matched set. If 10−10 or 10−15 is used as a cut-off value, the number of matched contigs decreases to 22,719 and 17865, respectively, but the fraction scored as foreign (i.e. with hU≥30) remains high, at 11.5% and 11.7% of matched sequences, respectively. Which database is used for blast matching also does not seem to be a major factor since both UniProtKB and Swiss-Prot gave similar proportions of foreign transcripts at 9.7% and 11.3%, respectively.
A final technical consideration might be to ask whether the HGT resulting from the endosymbiosis of the mitochondrial precursor affects our results. Endosymbiosis was a primordial event for eukaryotes, with acquisition of mitochondrial precursors taking place in the earliest eukaryotic cells, perhaps two billion years ago 
. The horizontal gene transfer we describe is very unlikely to have occurred before the divergence of bdelloids from monogonont rotifers (or B. plicatilis
would share similarly high levels of foreign genes), and therefore probably took place at most 100, more likely 65–80, million years ago 
. If horizontal gene transfer has continued throughout bdelloid evolution, many events will be more recent. Consequently, most, perhaps all, gene flux from mitochondrial precursor to nucleus would have occurred before bdelloids arose. Thus, we would not expect significant differences in numbers of nuclear mitochondrial genes between bdelloids and the other major class of rotifers, the monogononts, as exemplified by B. plicatilis
in our study. To test what proportion of foreign genes apparently derive from mitochondrial nuclear genes, we blast aligned sequences of 1,098 known nuclear mitochondrial genes from MitoCarta (www.broadinstitute.org/pubs/MitoCarta
) against our transcripts. Using a cut-off of 10−5
, only 0.7% of transcripts of foreign origin (hU
≥30) matched mitochondrial nuclear genes, whereas 2.9% of those of metazoan origin (hU
≤0) gave matches. If we adjust the blast cut-off to 10−10
, these proportions are approximately the same: 0.7% vs. 3.3%, and 0.8% vs. 3.6%, respectively. This shows that transcripts for nuclear mitochondrial genes are less likely to be found in the foreign sequence set than among metazoan transcripts and therefore will not cause an overestimate of HGT.
The complexity of foreign gene expression observed in the bdelloid rotifer A. ricciae
is comparable to that in prokaryotes 
and is far greater than in other animals where relatively few genes are involved 
. For example, while in Hydra
perhaps 50 foreign genes are active 
, in Drosophila ananassae
, which has acquired most of the genome of its endosymbiont, Wolbachia
, by HGT, only 28 genes are transcribed; the model fly, D. melanogaster
, has not acquired the Wolbachia
. In pea aphids, red body colour results from the expression of carotenoid genes acquired and diversified from fungal counterparts 
. In the sea slug, Elysia chlorotica
, HGT and expression of the algal psbO
gene allows photosynthesis in plastids also acquired from the alga 
. However, there is a need for more animal studies at the whole transcriptome level. It is surprising, for example, that there are no comprehensive global studies of HGT in C. elegans
in the literature 
, as our analysis suggests there are approximately 200 foreign transcripts in the model nematode. The software pipeline developed for this study has the potential to be used more widely where expression data are available to gain a more complete picture of HGT in metazoans.
Nevertheless, the scale of HGT in the bdelloid seems to be unusual among animals and it would be interesting to address the importance of asexuality and desiccation tolerance in this phenomenon. For example, transcriptome data from the nematode Panagrolaimus superbus
, which is anhydrobiotic, but gonochoristic (i.e. reproduces only sexually), has recently been published 
. The authors highlighted one foreign sequence in the P. superbus
transcriptome, but did not perform a global analysis for HGT. If this nematode contains low numbers of foreign sequences, it would rule out that desiccation tolerance per se
, without asexuality, is associated with extensive HGT. Another characteristic of HGT in A. ricciae
is that the source organisms are extremely diverse and include examples that are unlikely to be symbionts or even in the bdelloid's immediate habitat, such as the trypanosome relative from which trypanothione biosynthesis genes derive. Therefore, bdelloids are likely able to readily scavenge and incorporate DNA from the environment, and desiccation, which could lead to both leakiness in cell membranes and double-strand breaks in rotifer chromosomes, might facilitate this.
HGT in the bdelloid has the potential to radically extend and complement metazoan biochemistry, since approximately 80% of foreign sequences are concerned with enzyme activity, much of which is novel in animals. This supports the complexity hypothesis, which states that genes whose products are involved in relatively few protein-protein interactions, such as those specifying enzymes, are more likely to be horizontally transferred than those with a higher degree of connectivity, like transcription factor genes 
. Thus, although the complexity hypothesis was developed to explain observations in prokaryotes, it also seems to apply to the large scale HGT observed in the bdelloid. It would be interesting to investigate in the bdelloid a more recent suggestion from a study in prokaryotes that highly expressed genes are less likely to be horizontally transferred between organisms 
. Technically, this might be difficult to achieve, as we estimate there are at least 533 source organisms that have contributed to the bdelloid genome by HGT, but we will explore this in future work.
The novel biochemistry implicated includes the ability to degrade toxins, and indeed to exploit them and a range of otherwise unmetabolisable organic molecules as food sources, and to use novel biosynthetic pathways to generate protective molecules, for example antioxidants, or nutrients that are rare in the environment. This is expected to increase bdelloid stress tolerance and competitiveness, and could be important for anhydrobiosis. Bdelloids do not produce trehalose or other non-reducing disaccharides 
and have unusual LEA proteins 
, and therefore mechanisms associated with desiccation tolerance in other anhydrobiotes do not apply. Recently, the bdelloid A. vaga
was shown to have high antioxidant capacity; this reduces protein oxidation, which is thought to be a major problem caused by desiccation and the dry state 
. Antioxidants in bdelloids have not been characterised, but it will be of interest to determine how far HGT plays a role; this is currently under investigation.
It is also tempting to speculate that HGT facilitates long-term persistence in the absence of sex: asexuals are unable to bring together novel gene combinations arising within a population since they lack conventional genetic exchange mechanisms; equally, asexuals cannot eliminate detrimental mutations readily 
. Uptake and expression of genes from other organisms is a means of diversifying functional capacity, particularly biochemical capacity, and the potential to replace defective genes with foreign counterparts could protect against loss of function through mutation.