TERRA ONTseq: a long-read-based sequencing pipeline to study the human telomeric transcriptome

TERRA ONTseq detects transcripts deriving from a multitude of human chromosome ends

The ONT sequencing technology allows the sequencing of full-length DNA molecules of up to 100 kb, thus bypassing the need to fragment molecules for library preparation and being ideally suited for the sequencing of repetitive nucleic acids (Kwapisz and Morillon 2020; Warburton and Sebra 2023). We affinity-purified TERRA transcripts from nuclear RNA prepared from two telomerase-positive (HeLa and HEK293T) and two ALT (U2OS and SAOS2) cell lines using a biotinylated DNA oligonucleotide comprising five CCCTAA repeats, which are complementary to the UUAGGG tracts within TERRA molecules (Supplemental Fig. S1A; Supplemental Table S1). Biotinylated oligonucleotides were 3′ end blocked with dideoxycytosine to avoid interference with the subsequent step of reverse transcription. As shown by northern blotting, purified RNA fractions were largely devoid of the U6 small nuclear RNA, while containing telomeric UUAGGG repeats (Supplemental Fig. S1B). As quantified by RT-qPCR, the purification efficiencies of TERRA transcripts differed depending on the chromosome end and cell line (Supplemental Fig. S1C), possibly because of different conservations of the (UUAGGG)n sequence. With the exception of SAOS2, TERRA purification consistently provided sufficient material for ONT sequencing across all tested cell lines (Supplemental Fig. S1C).

TERRA-enriched RNA was reverse transcribed with an oligonucleotide complementary to the UUAGGG repeats (Supplemental Fig. S1A; Supplemental Table S1). Because the C-rich oligonucleotide can hybridize randomly along the (UUAGGG)n stretch, including in close proximity of the subtelomeric part of the transcript, this step assures that the majority of reads contain subtelomeric sequences and can be mapped. However, our protocol cannot provide information on the length of the UUAGGG tract. The obtained cDNA was Nanopore sequenced and quality score analysis identified reads up to 11,375, 12,310, and 23,869 bp long for U2OS, HEK293T, and HeLa, respectively (mean read length: 1064, 866, and 964 bp, respectively; median read length: 887, 684, and 779 bp, respectively). Preliminarily alignments of the reads to the build T2T CHM13v2.0/hs1 of the human genome (Nurk et al. 2022) did not detect any splicing signature (data not shown), which is consistent with previous reports (Porro et al. 2014). Hence, we refined the alignment strategy by generating a subtelomeric reference sequence from the T2T genome comprising 25 kb sequences terminating at the fourth telomeric TTAGGG repeats on each chromosome end. Because all cell lines used in this study are from female donors, the recently completed sequence of Chromosome Y was not included in the reference (Rhie et al. 2023).

For each cell line, we calculated the total number of both unique and multimapped reads (Supplemental Table S2), and when all reads from the three cell lines were pooled, at least 10 reads were assigned to 40 out of 46 chromosome ends (19 short arms and 21 long arms) (Fig. 1A,B; Supplemental Figs. S2–S4; Supplemental Table S2). In the majority of the cases, read density peaked in the vicinity of the telomeric repeats and declined when moving toward the inside of the subtelomere, with reads terminating ∼500–750 bp from the telomeric repeats (Fig. 1A,C; Supplemental Figs. S2–S4). A striking different behavior was observed for the 18q and 22p (short arm) subtelomeres, for which reads started 750–1000 bp away from the first telomeric repeat and covered a region of ∼500 bp (Supplemental Figs. S2–S4). This discrepancy may reflect differently structured 18q and 22p subtelomeres in the T2T genome and the cell lines analyzed in this study. Altogether these data reveal that (1) TERRA is transcribed from a large number of telomeres, as previously proposed (Azzalin et al. 2007; Nergadze et al. 2009; Arora et al. 2014; Episkopou et al. 2014; Porro et al. 2014; Feretzaki et al. 2019); and (2) most of TERRA molecules originate from regions very close to the first telomeric repeats and not from distant subtelomeric loci; this confirms that the heterogeneity of TERRA transcripts largely stems from the UUAGGG tracts (Azzalin et al. 2007; Farnung et al. 2012).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

TERRA expression in different human cell lines. (A) TERRA ONTseq read coverage for 7q and 8p subtelomeres in U2OS, HeLa, and HEK293T. Mapping was performed against the T2T CHM13v2.0/hs1 subtelomeric reference genome. Alignments span 1250 bp regions starting from the first telomeric repeat. The darker and lighter shade lines represent unique and multimapped reads, respectively. The vertical gray lines indicate the position of the most telomere-proximal 29 bp–like repeats. (B) Quantification of the unique and multimapped reads (darker and lighter shades, respectively) within 1250 bp regions as in A for each human subtelomere. Reads are normalized to the corresponding library size. (C) Metaprofiles of uniquely mapped reads for U2OS, HeLa, and HEK293T cells. Note that the read density profiles between 750 and 1250 bp largely derive from Chromosomes 8q, 18q, and 22p only. (D) RT-qPCR quantifications of TERRA transcripts from 5p, 7q/12q, and 16p subtelomeres in the indicated cell lines. Values are normalized to the number of molecules in U2OS cells. Bars and error bars are means and SDs from three independent experiments. P values (Student's t-test) are indicated.

Read mapping across different chromosome ends was not uniform. For example, the subtelomeres 8p, 7q, and 8q were highly expressed in all cell lines (Fig. 1A,B; Supplemental Figs. S2–S4; Supplemental Table S2). Conversely, some chromosome ends were underrepresented in our alignments, including the ones of the short arms of Chromosomes 1, 2, 3, 4, 6, 15, and 21 and the long arms of Chromosomes 2, 11, 14, and 19 (Fig. 1B; Supplemental Figs. S2–S4; Supplemental Table S2). We conclude that TERRA transcripts from different ends are synthesized and/or degraded at different rates, as previously suggested based on the analysis of a few chromosome ends (Feretzaki et al. 2019; Savoca et al. 2023). Technical limitations, including differences in the efficiencies of purification (Supplemental Fig. S1C) and/or reverse transcription for different TERRA species, may also account for the observed variabilities. Finally, we generated TERRA ONTseq metaprofiles and found that U2OS, HeLa, and HEK293T showed similar profile shapes; however, U2OS cells had the highest overall read count, followed by HEK293T and HeLa (Fig. 1C). These data were validated using RT-qPCRs detecting a few selected TERRA species (Fig. 1D). Because ALT cells are characterized by elevated TERRA levels (Lovejoy et al. 2012; Arora et al. 2014; Episkopou et al. 2014), TERRA ONTseq appears to be able to quantitatively measure TERRA transcripts not only across different chromosome ends but also different samples.

TERRA ONTseq identifies TERRA transcription start site regions

To position TERRA promoters and transcription start sites (TSSs), we first reanalyzed the distribution of tandem repeats related to the canonical 29 bp repeats in our T2T-derived subtelomeric reference and identified them in almost all subtelomeres, with the exception of 6p, 8p, 13p, 14p, 17p, 21p, 22p, Xp, 8q, and 18q (Fig. 2; Supplemental Table S4). More specifically, we found 12 subtelomeres containing the canonical 29 bp sequence (Cons1) (Fig. 2; Supplemental Table S4); 13 subtelomeres containing the canonical 29 bp sequence with a single mismatched nucleotide (Cons2); eight subtelomeres containing a 29 bp sequence with two mismatches compared to Cons1 and 1 to Cons 2 (Cons2 + 1); and eight subtelomeres containing a 29 bp sequence with a variable number of mismatches with respect to Cons1 (Cons1 + n). In addition, the 9q subtelomere contains a 29 bp repeat with two mismatches with respect to Cons2 (Cons2 + 2) (Fig. 2; Supplemental Table S4); the 12q subtelomere contains a 30 bp sequence with 1 nt insertion and two mismatches relative to Cons1 (Cons1 + gap); and finally, the 1q and 17q subtelomeres contain a 34 bp repeat sequence with two insertions of 2 and 3 nt and one mismatch with respect to Cons2 (Cons3). We also predicted telomere-adjacent CpG islands using the T2T reference sequence and identified at least one of them within 2000 bp from the first telomeric repeat on all subtelomeres with the exception of the ones of 13p, 17p, 21p, 22p, Xp, and 8q arms. The sizes of the identified CpG islands are comprised between 259 and 1553 bp (in 12q and 4p, respectively) and often contain multiple repetitions of one or more 29 bp consensus sequences (Fig. 2).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Twenty-nine#bp-like repeats, CpG-rich islands, and TERRA TSSs across all human chromosome ends. Schematic representation of the positions of the different variants of 29 bp–like repeats (triangles) and subtelomeric CpG islands (gray boxes) in the T2T CHM13v2.0/hs1 subtelomeric reference genome. TERRA TSSs identified by TERRA ONTseq are also indicated (pink arrowheads). For each chromosome arm, the most telomere-proximal 2000 bp are shown. Note that Chromosome 5p has two TERRA TSSs. The one on the left was identified in HeLa cells while the one on the right was identified in U2OS and HEK293T cells.

We then used TERRA ONTseq reads to identify TSS-containing regions by defining a threshold of at least three subtelomeric reads with 5′ ends mapping to the same 10 bp window. This strategy successfully identified 10 bp regions containing TERRA TSSs in 35 subtelomeres (Fig. 2; Supplemental Table S3). Notably, TERRA TSS regions are predominantly positioned within the CpG islands, just downstream from the 29 bp repeats (Fig. 2). This validates the ability of our pipeline to define TERRA TSSs and confirms that TERRA promoters are comprised of a variety of tandemly repeated 29 bp–like sequences. Importantly, some of the identified 29 bp–like repeats are only a few nucleotides away from the first telomeric repeats, as it is the case for the 1p, 3p, 10p, and 14q subtelomeres (Fig. 2), for which virtually no unique TERRA reads were detected (Fig. 1B; Supplemental Figs. S2–S4; Supplemental Table S2). It is plausible that those chromosome ends are also transcribed into TERRA molecules, but those transcripts escaped our analysis because the subtelomeric tract is too short to be mapped.

TERRA ONTseq detects quantitative differences in TERRA molecules

As mentioned, TERRA ONTseq appears to be able to quantitate differences among different samples. To corroborate this assumption using isogenic cell systems, we took advantage of a previously described U2OS-derived cell line (SID4 cells), in which TERRA expression from subtelomeres containing a 20 bp sequence within the 29 bp repeat consensus can be suppressed using activator-like effectors (T-TALEs) fused to the transcription repressor domain SID4X and cloned under the control of a doxycycline (dox) inducible promoter (Silva et al. 2021). Control cells expressing unfused T-TALEs (NLS3 cells) (Silva et al. 2021) were also used. SID4 and NLS3 cells were treated with dox for 24 h or left untreated and subjected to TERRA ONTseq (Supplemental Fig. S5A). T-TALE expression and suppression of TERRA transcription from selected chromosome ends containing 20 bp target sequences were validated by western blot and RT-qPCR, respectively (Supplemental Fig. S5B,C).

SID4 cells generated lower TERRA read counts (normalized to library size) than NLS3 cells already in the absence of dox treatment, possibly because of leaky expression of the T-TALE constructs (Supplemental Fig. S5A,B). The dox treatment decreased the number of reads mapped to subtelomeres containing the T-TALE target in the SID4 samples (average 3.4-fold decrease), whereas a milder effect was observed at subtelomeres devoid of the 20 bp T-TALE target (average 1.4-fold decrease) (Supplemental Fig. S5A; Supplemental Table S2). In NLS3 cells, dox treatment did not cause noticeable changes in TERRA reads at any chromosome ends (Supplemental Fig. S5A; Supplemental Table S2). Overall, these data confirm that TERRA ONTseq detects quantitative differences in TERRA levels between independent samples. Moreover, suppression of TERRA transcription from a fraction of chromosome ends does not trigger a compensatory mechanism increasing transcription from the other ends, at least in U2OS cells.

While analyzing the T-TALE data, we noticed that the 6q and 7q subtelomeres did not respond to dox treatment, despite being annotated in the T2T genome as containing numerous copies of the 20 bp target sequence (Supplemental Fig. S6A; Supplemental Table S4). This raised the possibility that the 6q and 7q subtelomeres contain 29 bp repeat-like sequences in the CHM13 cell line that was used for T2T sequencing (Nurk et al. 2022) but not in other cell lines. We thus focused on the highly expressed 7q subtelomere, which is annotated in the T2T genome assembly as having 29 copies of the Cons1 + n 29 bp repeat variant (Supplemental Fig. S6A; Supplemental Table S4). The same locus is devoid of 29 bp repeats in the GRCh38/hg38 human genome assembly (Supplemental Fig. S6A,B; Supplemental Table S4). Additionally, a previous study on the architecture of 7q subtelomere did not detect 29 bp repeats in 10 DNA samples from unrelated individuals (Baird et al. 2000). We PCR-amplified HeLa and U2OS genomic DNA with oligonucleotides flanking the 29 bp repeats in the T2T 7q subtelomere and obtained only amplicons of sizes consistent with the lack of the 29 bp repeats tract (1537 and 697 bp for fragments with and without insertion, respectively; data not shown). We also cloned the amplified fragments into a plasmid and Sanger sequenced independent inserts for each cell line, and did not detect 29 bp repeat-like sequences in any of the inserts (Supplemental Fig. S6A,B). Hence, the 29 bp repeats present in the T2T genome at the 7q subtelomere might correspond to an insertion polymorphism that is not shared by all individuals or by HeLa and U2OS cells. This explains why our T-TALEs did not affect the expression of 7q TERRA and suggests a similar explanation also for the 6q subtelomere. Given these findings, we used the GRCh38/hg38 sequence for the 7q alignments performed later during this study.

Analysis of 7q TERRA expression

The highly expressed 7q TERRA accounts for 26%, 5%, and 10% of total unique TERRA reads in U2OS, HeLa, and HEK293T cells, respectively (Fig. 3A). To better characterize this transcript, we first performed northern blotting using a probe corresponding to a 441 bp fragment from the 7q subtelomere (7q/12q probe) (Supplemental Fig. S6C). Because of the high identity between 7q and 12q subtelomeres (Supplemental Fig. S6C), our probe is expected to detect TERRA species from both loci; nonetheless, based on our TERRA ONTseq data, 74% of the signal generated by this probe should derive only from 7q TERRA (Fig. 1B; Supplemental Table S2). We isolated total and nuclear RNA from U2OS, HeLa, and HEK293T cells and subjected it to northern blotting using 7q/12q and total (UUAGGG) TERRA probes. The signals generated by the two probes were similar and appeared as smears ranging from <0.5 to >9 kb (Fig. 3B,C).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Expression of 7q TERRA. (A) Distribution of TERRA reads from 7q, 12q, 29 bp-like-containing subtelomeres and from repeat-devoid subtelomeres in the indicated cell lines. Only unique reads were considered for the analysis. (B) Northern blot hybridizations of total and nuclear RNA from U2OS, HeLa, and HEK293T cells using probes for the 7q/12q TERRA or UUAGGG repeats (total TERRA). The positions of 28S and 18S ribosomal RNAs are indicated on the right. GAPDH and U6 were used as loading controls. RNaseA treatments were used to assure that signals derived from RNA and not genomic DNA contaminations. (C) Total and nuclear RNA from U2OS cells was incubated with RNaseH in the presence of oligonucleotides complementary to a sequence from 7q/12q TERRA or UUAGGG repeats (Tel). Oligonucleotides complementary to a GAPDH sequence were also included as indicated and served as a control for the RNaseH treatment. Products were analyzed by northern blotting as in B. (D) U2OS cells were treated with actinomycin D for the indicated times. Total RNA was extracted and analyzed by northern blotting as in C. The long-lived GAPDH mRNA was used as a loading control. Half-lives and R² values were calculated using a scatter plot analysis after the 7q/12q or the TERRA signals were normalized to the GAPDH signals. Data points are from two independent experiments and the values for time 0 are set to 1.

We then used a northern blot–based approach previously developed to analyze TERRA transcripts from specific subtelomeres (Nergadze et al. 2009). We incubated total and nuclear RNA from U2OS cells with an oligonucleotide corresponding to a unique sequence shared by 7q and 12q subtelomeres and located 85 bp from the telomeric tract in the 7q subtelomere or with a (CCCTAA)5 oligonucleotide (7q_12q or Tel oligonucleotide, respectively) (Supplemental Fig. S6C). In both reactions, we also included a control oligonucleotide annealing to GAPDH mRNA. We treated the RNA with RNaseH, which digests specifically RNA molecules engaged in DNA/RNA hybrids, and analyzed it by northern blotting (Fig. 3C). When compared to the undigested controls, the 7q/12q TERRA signal in RNaseH-treated sample incubated with the telomeric C-rich oligonucleotide appeared as a smear between ∼1000 and 500 bases; this smear should correspond to the 7q/12q TERRA subtelomeric sequences largely devoid of telomeric repeats. RNaseH digestion in the presence of the 7q_12q oligonucleotide produced a slightly lower major band compared to the digestion with the telomeric oligonucleotide, confirming that the 7q/12q probe mainly detects 7q/12q TERRA and that these transcripts contain long telomeric UUAGGG tracts (Fig. 3C). Hybridization with a telomeric probe showed, as expected, a significantly reduced signal from the RNA incubated with the Tel oligonucleotide, whereas the incubation with the 7q_12q oligonucleotide led to a substantial disappearance of longer TERRA species above 9 kb (Fig. 3C). Attesting the efficiency and specificity of the RNaseH treatment, all GAPDH transcripts were shorter in samples treated with RNaseH, whereas the length of the unrelated U6 transcript remained unaffected (Fig. 3C). Altogether, these experiments confirm that 7q TERRA constitutes a substantial fraction of total TERRA and reveals that individual 7q TERRA species contain long and heterogeneous UUAGGG repeat tracts in all tested cell lines.

To determine whether the increased levels of 7q TERRA derive from increased transcription or RNA stability, we measured 7q/12q TERRA stability by treating U2OS cells with actinomycin D to halt transcription and collecting total RNA every hour over 8 h. Northern blot analysis showed that 7q/12q TERRA has a half-life of 4.8 h, whereas total TERRA has a half-life of >8 h (Fig. 3D), which is consistent with previously reported values (Azzalin and Lingner 2008; Porro et al. 2010; Savoca et al. 2023). Hence the abundance of 7q TERRA cannot be ascribed to overall increased RNA stability, but rather it should derive from higher transcription rates.

Analysis of 7q TERRA promoter regulation

7q/12q TERRA levels are higher in U2OS ALT cells than in HeLa and HEK293T telomerase-positive cells; however, those differences are not as pronounced as the ones detected for total UUAGGG repeats (Fig. 3B). We expanded our analysis to a larger panel of cell lines and again observed that 7q/12q TERRA was, on average, 1.5 times higher in ALT cells compared to telomerase-positive and primary lung fibroblasts, whereas the total UUAGGG signal was roughly 5.5 times higher in ALT cells compared to the others (Fig. 4A). We conclude that 7q/12q TERRA expression is not regulated identically to the majority of TERRA species.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Transcriptional regulation of 7q TERRA. (A) Total RNA from ALT, telomerase-positive (Telomerase+), or primary (Pr.) cells was analyzed by northern blotting using probes for the 7q/12q subtelomere or UUAGGG repeats (total TERRA). GAPDH and U6 RNAs were used as loading controls. (B) Total and nuclear RNA from HCT116 and DKO cells was analyzed by northern blotting as in A. U2OS RNA was included in the same blots for comparison with an ALT cell line. (C) Northern blot analysis of eGFP expression in HeLa cells transfected with the p7q promoter reporter. An empty vector plasmid devoid of promoter sequences (pEV) and the p3.5 plasmid containing a 29 bp repeat promoter were used as negative and positive controls, respectively. Twenty-four hours after transfection, cells were selected using puromycin and 48 h later harvested for RNA preparation. The bar graph shows the quantification of eGFP signals normalized through the corresponding U6 signals. Bars and error bars are the means and SDs from three independent experiments. P values (Student's t-test) are indicated. (D) Representative images of eGFP fluorescence in HeLa cells treated as in C. eGFP is shown in green, DAPI-stained DNA in blue. The bar graph shows the total eGFP fluorescence per cell. Bars and error bars are the means and SDs from two independent experiments. P values (Student's t-test) are indicated. (E) Heatmap representation of 7q CpG island methylation for the indicated cell lines. The CpG island coordinates are chr7:159,335,203–159,335,443 in the GRCh38/hg38 reference genome. The position of 7q TERRA TSS gauged from TERRA ONTseq is indicated.

Because TERRA 29 bp repeat promoters are silenced by CpG methylation mediated by DNMT1 and DNMT3b (Nergadze et al. 2009; Feretzaki et al. 2019), it is possible that CpG methylation is not substantially affecting 7q/12q TERRA expression. We performed northern blotting to analyze 7q/12q and total TERRA expression in human DKO cells and parental HCT116 cells. As expected, we detected a very faint signal for total TERRA in HCT116 parental cell lines when compared to the very strong signal detected for DKO cells (16%–20% of the DKO signal) (Fig. 4B). On the contrary, the 7q/12q TERRA signal was easily detected in HCT116 cells and its intensity was roughly half of the one in DKO cells (44%–49% of the DKO signal) (Fig. 4B). These data suggest that the 7q/12q promoters might be overall hypomethylated across different cell lines regardless of their telomere elongation mechanism, thus explaining why 7q/12q TERRA transcripts are more homogeneously expressed among cell lines and constitute a large fraction of total TERRA transcripts.

To test the hypothesis above, we first set up to isolate the 7q TERRA promoter. We generated a promoter reporter plasmid (p7q) by cloning a 695 bp fragment from the 7q subtelomere, overlapping a predicted CpG island of 333 bp (Fig. 2; Supplemental Fig. S6C), in front of an eGFP coding sequence. We transfected the plasmid into HeLa cells and analyzed eGFP expression by northern blotting and fluorescence microscopy. As a control, we transfected cells with a plasmid in which the eGFP cDNA is preceded by a 29 bp repeat promoter from the XqYq subtelomere (p3.5 plasmid) (Nergadze et al. 2009). We found that the selected 7q subtelomeric sequence was able to induce the expression of eGFP (Fig. 4C,D), confirming that we isolated a novel TERRA promoter sequence that does not contain 29 bp repeats. The p7q plasmid produced eGFP transcripts ∼5× less abundant, and functional eGFP proteins ∼2.4× less abundant than the ones produced from the p3.5 plasmid (Fig. 4C,D). This indicates that the 29 bp repeat promoters are stronger than the 7q promoter; however, they are likely to be silenced more efficiently in cells.

Having defined the 7q promoter, we performed bisulfite sequencing of that region using genomic DNA from U2OS, HeLa, HEK293T, HLF, HCT116, and DKO cells and primers flanking the identified CpG island (Supplemental Fig. S6C). Unfortunately, we were not able to isolate amplification products for the 7q subtelomere from HEK293T, HCT116, and DKO cells, possibly because of polymorphisms within the regions annealing with the chosen oligonucleotides. Nonetheless, we could analyze CpG methylation in the other three cell lines. 7q methylation patterns from U2OS, HeLa, and HLF did not show any striking differences (Fig. 4D; Supplemental Fig. S6C), indicating that CpG methylation has only a secondary role in regulating 7q TERRA expression. On the other hand, it was previously reported that the 29 bp repeat promoters are more methylated in HLF and HeLa cells (low TERRA levels) than in U2OS cells (high TERRA levels), pointing to methylation as a primary mechanism silencing the expression of 29 bp repeat promoters (Nergadze et al. 2009).

Conclusions

TERRA ONTseq allows users to simultaneously analyze TERRA transcripts from nearly all telomeres and provides robust and reproducible data. Although confirming that 29 bp repeat-like sequences are the main TERRA promoters, other promoter regions devoid of such repeats, such as the one of the 7q subtelomere identified here, are present in humans and substantially contribute to TERRA cellular levels. Similarly, TERRA ONTseq should allow us to identify and validate all TERRA promoter sequences. TERRA ONTseq also showed that the most transcribed chromosome ends are consistent between cell lines, likely pointing to the existence of conserved regulatory elements maintained across different cell types. Given the qualitative and quantitative nature of TERRA ONTseq, we are convinced that it will become an extremely useful tool for a rapid and precise analysis of the human TERRA transcriptome in different laboratory and clinical settings. Further methodological improvements will broaden the reach of this tool—for example, by developing protocols to also sequence the entire UUAGGG tract within TERRA or by adapting the TERRA ONTseq pipeline to organisms other than humans.