Liguo Zhanga, Alexsia Richardsa, M. Inmaculada Barrasaa\ue840, Stephen H. Hughesb\ue840, Richard A. Younga,c, and Rudolf Jaenischa,c,1
\na Whitehead Institute for Biomedical Research, Cambridge, MA 02142; b<\/p>\n
HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer<\/p>\n
Institute, Frederick, MD 21702; and c<\/p>\n
Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02142
\nContributed by Rudolf Jaenisch, April 19, 2021 (sent for review March 29, 2021; reviewed by Anton Berns and Anna Marie Skalka)<\/p>\n
Prolonged detection of severe acute respiratory syndrome coro-
\nnavirus 2 (SARS-CoV-2) RNA and recurrence of PCR-positive tests<\/p>\n
have been widely reported in patients after recovery from COVID-
\n19, but some of these patients do not appear to shed infectious<\/p>\n
virus. We investigated the possibility that SARS-CoV-2 RNAs can be
\nreverse-transcribed and integrated into the DNA of human cells in
\nculture and that transcription of the integrated sequences might<\/p>\n
account for some of the positive PCR tests seen in patients. In sup-
\nport of this hypothesis, we found that DNA copies of SARS-CoV-2<\/p>\n
sequences can be integrated into the genome of infected human
\ncells. We found target site duplications flanking the viral sequences
\nand consensus LINE1 endonuclease recognition sequences at the<\/p>\n
integration sites, consistent with a LINE1 retrotransposon-
\nmediated, target-primed reverse transcription and retroposition<\/p>\n
mechanism. We also found, in some patient-derived tissues, evi-
\ndence suggesting that a large fraction of the viral sequences is<\/p>\n
transcribed from integrated DNA copies of viral sequences, gener-
\nating viral\u2013host chimeric transcripts. The integration and transcrip-
\ntion of viral sequences may thus contribute to the detection of viral<\/p>\n
RNA by PCR in patients after infection and clinical recovery. Because
\nwe have detected only subgenomic sequences derived mainly from
\nthe 3\u2032 end of the viral genome integrated into the DNA of the host
\ncell, infectious virus cannot be produced from the integrated
\nsubgenomic SARS-CoV-2 sequences.
\nSARS-CoV-2 | reverse transcription | LINE1 | genomic integration | chimeric
\nRNAs
\nContinuous or recurrent positive severe acute respiratory
\nsyndrome coronavirus 2 (SARS-CoV-2) PCR tests have been
\nreported in samples taken from patients weeks or months after
\nrecovery from an initial infection (1\u201317). Although bona fide
\nreinfection with SARS-CoV-2 after recovery has recently been
\nreported (18), cohort-based studies with subjects held in strict
\nquarantine after they recovered from COVID-19 suggested that
\nat least some \u201cre-positive\u201d cases were not caused by reinfection<\/p>\n
(19, 20). Furthermore, no replication-competent virus was iso-
\nlated or spread from these PCR-positive patients (1\u20133, 5, 6, 12,<\/p>\n
16), and the cause for the prolonged and recurrent production of
\nviral RNA remains unknown. SARS-CoV-2 is a positive-stranded<\/p>\n
RNA virus. Like other beta-coronaviruses (SARS-CoV-1 and Mid-
\ndle East respiratory syndrome-related coronavirus), SARS-CoV-2<\/p>\n
employs an RNA-dependent RNA polymerase to replicate its
\ngenomic RNA and transcribe subgenomic RNAs (21\u201324). One
\npossible explanation for the continued detection of SARS-CoV-2
\nviral RNA in the absence of virus reproduction is that, in some
\ncases, DNA copies of viral subgenomic RNAs may integrate
\ninto the DNA of the host cell by a reverse transcription
\nmechanism. Transcription of the integrated DNA copies could<\/p>\n
be responsible for positive PCR tests long after the initial in-
\nfection was cleared. Indeed, nonretroviral RNA virus sequences<\/p>\n
have been detected in the genomes of many vertebrate species
\n(25, 26), with several integrations exhibiting signals consistent<\/p>\n
with the integration of DNA copies of viral mRNAs into the
\ngermline via ancient long interspersed nuclear element (LINE)
\nretrotransposons (reviewed in ref. 27). Furthermore, nonretroviral
\nRNA viruses such as vesicular stomatitis virus or lymphocytic
\nchoriomeningitis virus (LCMV) can be reverse transcribed into
\nDNA copies by an endogenous reverse transcriptase (RT), and
\nDNA copies of the viral sequences have been shown to integrate
\ninto the DNA of host cells (28\u201330). In addition, cellular RNAs, for
\nexample the human APP transcripts, have been shown to be<\/p>\n
reverse-transcribed by endogenous RT in neurons with the re-
\nsultant APP fragments integrated into the genome and expressed<\/p>\n
(31). Human LINE1 elements (\u223c17% of the human genome), a<\/p>\n
type of autonomous retrotransposons, which are able to retro-
\ntranspose themselves and other nonautonomous elements such<\/p>\n
as Alu, are a source of cellular endogenous RT (32\u201334). Endog-
\nenous LINE1 elements have been shown to be expressed in aged<\/p>\n
human tissues (35) and LINE1-mediated somatic retro-
\ntransposition is common in cancer patients (36, 37). Moreover,<\/p>\n
expression of endogenous LINE1 and other retrotransposons
\nin host cells is commonly up-regulated upon viral infection,
\nincluding SARS-CoV-2 infection (38\u201340).
\nSignificance
\nAn unresolved issue of SARS-CoV-2 disease is that patients
\noften remain positive for viral RNA as detected by PCR many
\nweeks after the initial infection in the absence of evidence for
\nviral replication. We show here that SARS-CoV-2 RNA can be<\/p>\n
reverse-transcribed and integrated into the genome of the in-
\nfected cell and be expressed as chimeric transcripts fusing viral<\/p>\n
with cellular sequences. Importantly, such chimeric transcripts
\nare detected in patient-derived tissues. Our data suggest that,
\nin some patient tissues, the majority of all viral transcripts are<\/p>\n
derived from integrated sequences. Our data provide an in-
\nsight into the consequence of SARS-CoV-2 infections that may<\/p>\n
help to explain why patients can continue to produce viral RNA
\nafter recovery.
\nAuthor contributions: L.Z., R.A.Y., and R.J. designed research; L.Z. and A.R. performed
\nexperiments; L.Z., A.R., M.I.B., S.H.H., R.A.Y., and R.J. analyzed data; and L.Z. and R.J.
\nwrote the paper with input from all authors.
\nReviewers: A.B., Netherlands Cancer Institute; and A.M.S., Fox Chase Cancer Center.
\nCompeting interest statement: R.J. is an advisor\/co-founder of Fate Therapeutics, Fulcrum
\nTherapeutics, Omega Therapeutics, and Dewpoint Therapeutics. R.A.Y. is a founder and
\nshareholder of Syros Pharmaceuticals, Camp4 Therapeutics, Omega Therapeutics, and
\nDewpoint Therapeutics.
\nThis open access article is distributed under Creative Commons Attribution License 4.0
\n(CC BY).
\n1
\nTo whom correspondence may be addressed. Email: jaenisch@wi.mit.edu.
\nThis article contains supporting information online at https:\/\/www.pnas.org\/lookup\/suppl\/
\ndoi:10.1073\/pnas.2105968118\/-\/DCSupplemental.
\nPublished May 6, 2021.<\/p>\n
PNAS 2021 Vol. 118 No. 21 e2105968118 https:\/\/doi.org\/10.1073\/pnas.2105968118 | 1 of 10<\/p>\n
In this study, we show that SARS-CoV-2 sequences can inte-
\ngrate into the host cell genome by a LINE1-mediated retro-
\nposition mechanism. We provide evidence that the integrated<\/p>\n
viral sequences can be transcribed and that, in some patient
\nsamples, the majority of viral transcripts appear to be derived
\nfrom integrated viral sequences.
\nResults
\nIntegration of SARS-CoV-2 Sequences into the DNA of Host Cells in
\nCulture. We used three different approaches to detect genomic
\nSARS-CoV-2 sequences integrated into the genome of infected<\/p>\n
cells. These approaches were Nanopore long-read sequencing, Illu-
\nmina paired-end whole genomic sequencing, and Tn5 tagmentation-
\nbased DNA integration site enrichment sequencing. All three<\/p>\n
methods provided evidence that SARS-CoV-2 sequences can
\nbe integrated into the genome of the host cell.
\nTo increase the likelihood of detecting rare integration events,
\nwe transfected HEK293T cells with LINE1 expression plasmids
\nprior to infection with SARS-CoV-2 and isolated DNA from the
\ncells 2 d after infection (SI Appendix, Fig. S1A). We detected
\nDNA copies of SARS-CoV-2 nucleocapsid (NC) sequences in
\nthe infected cells by PCR (SI Appendix, Fig. S1B) and cloned the
\ncomplete NC gene (SI Appendix, Fig. S1D) from large-fragment
\ncell genomic DNA that had been gel-purified (SI Appendix, Fig.
\nS1C). The viral DNA sequence (NC) was confirmed by Sanger
\nsequencing (Dataset S1). These results suggest that SARS-CoV-2
\nRNA can be reverse-transcribed, and the resulting DNA could be
\nintegrated into the genome of the host cell.
\nTo demonstrate directly that the SARS-CoV-2 sequences were<\/p>\n
integrated into the host cell genome, DNA isolated from in-
\nfected LINE1-overexpressing HEK293T cells was used for<\/p>\n
Nanopore long-read sequencing (Fig. 1A). Fig. 1 B\u2013D shows an
\nexample of a full-length viral NC subgenomic RNA sequence
\n(1,662 bp) integrated into the cell chromosome X and flanked on
\nboth sides by host DNA sequences. Importantly, the flanking<\/p>\n
sequences included a 20-bp direct repeat. This target site du-
\nplication is a signature of LINE1-mediated retro-integration (41,<\/p>\n
42). Another viral integrant comprising a partial NC subgenomic
\nRNA sequence that was flanked by a duplicated host cell DNA
\ntarget sequence is shown in SI Appendix, Fig. S2 A\u2013C. In both
\ncases, the flanking sequences contained a consensus recognition
\nsequence of the LINE1 endonuclease (43). These results indicate
\nthat SARS-CoV-2 sequences can be integrated into the genomes<\/p>\n
of cultured human cells by a LINE1-mediated retroposition mech-
\nanism. Table 1 summarizes all of the linked SARS-CoV-2\u2013host se-
\nquences that were recovered. DNA copies of portions of the viral<\/p>\n
genome were found in almost all human chromosomes. In addi-
\ntion to the two examples given in Fig. 1 and SI Appendix, Fig. S2,<\/p>\n
we also recovered cellular sequences for 61 integrants for which<\/p>\n
only one of the two host\u2013viral junctions was retrieved (SI Ap-
\npendix, Fig. S2 D\u2013F and Table 1; Nanopore reads containing the<\/p>\n
chimeric sequences summarized in Dataset S2). Importantly,
\nabout 67% of the flanking human sequences included either a
\nconsensus or a variant LINE1 endonuclease recognition sequence
\n(such as TTTT\/A) (SI Appendix, Fig. S2 D\u2013F and Table 1). These<\/p>\n
LINE1 recognition sequences were either at the chimeric junc-
\ntions that were directly linked to the 3\u2032 end (poly-A tail) of viral<\/p>\n
sequences, or within a distance of 8\u201327 bp from the junctions that
\nwere linked to the 5\u2032 end of viral sequences, which is within the
\npotential target site duplication. Both results are consistent with a<\/p>\n
model in which LINE1-mediated retroposition provides a mech-
\nanism to integrate DNA copies of SARS-CoV-2 subgenomic<\/p>\n
fragments into host genomic DNA. About 71% of the viral se-
\nquences were flanked by intron or intergenic cellular sequences<\/p>\n
and 29% by exons (Fig. 1F and Table 1). Thus, the association of
\nthe viral sequences with exons is much higher than would be<\/p>\n
expected for random integration into the genome [human ge-
\nnome: 1.1% exons, 24% introns, and 75% intergenic DNA (44)],<\/p>\n
suggestive of preferential integration into exon-associated target
\nsites. While previous studies showed no preference for LINE1
\nretroposition into exons (45, 46), our finding suggests that LINE1-
\nmediated retroposition of some other RNAs may be different. We
\nnoted that viral\u2013cellular boundaries were frequently close to the 5\u2032
\nor 3\u2032 untranslated regions (UTRs) of the cellular genes, suggesting
\nthat there is a preference for integration close to promoters or
\npoly(A) sites in our experimental system.<\/p>\n
To confirm the integration of SARS-CoV-2 sequences into ge-
\nnomic DNA by another method, we subjected DNA isolated from<\/p>\n
LINE1-transfected and SARS-CoV-2\u2013infected HEK293T cells to
\nIllumina paired-end whole-genome sequencing, using a Tn5-based
\nlibrary construction method (Illumina Nextera) to avoid ligation
\nartifacts. Viral DNA reads were concentrated at the 3\u2032 end of
\nthe SARS-CoV-2 genome (SI Appendix, Fig. S3). We recovered
\n17 viral integrants (sum of two replicates), by mapping human\u2013
\nviral chimeric DNA sequences (Fig. 1E and Table 2, chimeric
\nsequences summarized in Dataset S3); 7 (41%) of the junctions<\/p>\n
contained either a consensus or a variant LINE1 recognition se-
\nquence in the cellular sequences near the junction (Fig. 1E and<\/p>\n
Table 2), consistent with a LINE1-mediated retroposition mech-
\nanism. Similar to the results obtained from Nanopore sequencing,<\/p>\n
about 76% of the viral sequences were flanked by intron or
\nintergenic cellular sequences and 24% by exons (Fig. 1F and
\nTable 2).
\nAbout 32% of SARS-CoV-2 sequences (6\/21 integration
\nevents in Nanopore, 4\/10 in Illumina data) were integrated at
\nLINEs, short interspersed nuclear elements, or long terminal
\nrepeat elements without evidence for a LINE1 recognition site,
\nsuggesting that there may be an alternative reverse transcription\/
\nintegration mechanism, possibly similar to that reported for cells
\nacutely infected with LCMV, which resulted in integrated
\nLCMV sequences fused to intracisternal A-type particle (IAP)
\nsequences (29).<\/p>\n
To assess whether genomic integration of SARS-CoV-2 se-
\nquences could also occur in infected cells that did not over-
\nexpress RT, we isolated DNA from virus-infected HEK293T and<\/p>\n
Calu3 cells that were not transfected with an RT expression
\nplasmid (Fig. 2A). Tn5 tagmentation-mediated DNA integration
\nsite enrichment sequencing (47, 48) (Fig. 2B and SI Appendix,
\nFig. S4A) detected a total of seven SARS-CoV-2 sequences<\/p>\n
fused to cellular sequences in these cells (sum of three inde-
\npendent infections of two cell lines), all of which showed LINE1<\/p>\n
recognition sequences close to the human\u2013SARS-CoV-2 se-
\nquence junctions (Fig. 2 C\u2013F and SI Appendix, Fig. S4 B\u2013D,<\/p>\n
chimeric sequences summarized in Dataset S4).
\nExpression of Viral\u2013Cellular Chimeric Transcripts in Infected Cultured
\nCells and Patient-Derived Tissues. To investigate the possibility that
\nSARS-CoV-2 sequences integrated into the genome can be
\nexpressed, we analyzed published RNA-seq data from
\nSARS-CoV-2\u2013infected cells for evidence of chimeric transcripts
\n(49). Examination of these datasets (50\u201355) (SI Appendix, Fig.<\/p>\n
S5) revealed a number of human\u2013viral chimeric reads (SI Ap-
\npendix, Fig. S6 A and B). These occurred in multiple sample<\/p>\n
types, including cultured cells and organoids from lung\/heart\/
\nbrain\/stomach tissues (SI Appendix, Fig. S6B). The abundance of
\nthe chimeric reads positively correlated with viral RNA level
\nacross the sample types (SI Appendix, Fig. S6B). Chimeric reads
\ngenerally accounted for 0.004\u20130.14% of the total SARS-CoV-2
\nreads in the samples. A majority of the chimeric junctions<\/p>\n
mapped to the sequence of the SARS-CoV-2 NC gene (SI Ap-
\npendix, Fig. S6 C and D). This is consistent with the finding that<\/p>\n
NC RNA is the most abundant SARS-CoV-2 subgenomic RNA
\n(56), making it the most likely target for reverse transcription
\nand integration. However, recent data showed that up to 1% of<\/p>\n
RNA-seq reads from SARS-CoV-2\u2013infected cells can be arti-
\nfactually chimeric as a result of RT switching between RNA<\/p>\n
2 of 10 | PNAS Zhang et al.
\nhttps:\/\/doi.org\/10.1073\/pnas.2105968118 Reverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
\u201cHuman-CoV2-human\u201d chimeric read (Nanopore)
\nSARS-CoV-2
\n(1662 bp)
\nHuman
\n(43 bp)<\/p>\n
\u201cHuman-CoV2-human\u201d chimeric read (Nanopore)
\nalignment on Human ChrX
\nchrX: 137,084,400 137,084,600 137,084,800
\n21.1 q23 q25 q28<\/p>\n
TAAGATAATCCAACTTCATTTTTCTTCAATTGCTATTGCTTCTTGTCATTCTCTAAGAAGCTATTAAATC
\nACATGGGGATAGCACTACTAAAATTAATTTTGCATTGAGCTCTTCCATATAGGTGGCTCTCTAACATTGT
\n……
\nTTATCAGACATTTTAGTTTGTTCGTTTAGAGAACAGATCTACAAGAGATCGAAAGTTGGTTGGTTTGTTA
\nCCTGGGAAGGTATAAACCTTTAATCGCTATTGCTTCTAAAAGGAAAAAATGAAAACAATTGCAGA…<\/p>\n
bp
\nHuman
\n(450 bp)<\/p>\n
Nanopore read alignment<\/p>\n
—>CCAACT TCA T T T T TCT TCAA T TGCT A T TGCT TCT AAAAGGAAAAA TG<\/p>\n
\u201cHuman-CoV2-human\u201d chimeric read (Nanopore)
\nalignment on the SARS-CoV-2 genome
\nScale
\nNC_045512v2:
\n10 kb
\n10,000 15,000 20,000 25,000<\/p>\n
Nanopore read alignment
\nNCBI Genes from NC_045512.2<\/p>\n
ORF1a
\nORF1ab<\/p>\n
S
\nORF3a
\nE
\nM
\nORF6
\nORF7a
\nORF7b
\nORF8
\nN<\/p>\n
ORF10
\nI II III
\n—>
\n10 bases
\n55 60 65 70 75 80 85
\nUUGUAGAUCUGUUCUCUAAACGAACUUUAAAAUCUGU<\/p>\n
—>
\n10 bases
\n28,250 28,255 28,260 28,265 28,270 28,275 28,280
\nGAUUUCAUCUAAACGAACAAACUAAAAUGUCUGAUAA
\nNCBI Genes from NC_045512.2<\/p>\n
ORF8 D 119 F 120 I 121 * 122 N M 1 S 2 D 3 N 4
\n—>
\n10 bases
\n29,860 29,865 29,870 29,875 29,880 29,885
\nCUUCUUAGGAGAAUGACAAAAAAAAAAAAAAAAAAAA<\/p>\n
TRS-L<\/p>\n
TRS-B<\/p>\n
I<\/p>\n
II<\/p>\n
III
\nDay:
\nHEK293T
\n+ LINE1 expression
\n1
\n\u00b1 SARS-CoV-2
\ninfection<\/p>\n
3
\nCell DNA extraction
\n(+RNase)<\/p>\n
A<\/p>\n
B<\/p>\n
C<\/p>\n
D<\/p>\n
E<\/p>\n
Target site duplication and LINE1 endonuclease recognition sequence (TTCT|A)<\/p>\n
F<\/p>\n
Intergenic
\nn=18
\n28.6%
\nIntron
\nn=27
\n42.9%
\nExon\/UTR
\nn=18
\n28.6%
\nHuman-CoV2 chimeric junction
\ndistribution in the human genome
\n\u201cHuman-CoV2\u201d chimeric read
\n(Illumina 2 x 150bp paired-end):<\/p>\n
Nanopore<\/p>\n
Whole genome sequencing
\n\u2022 Nanopore
\n\u2022 Illumina paired-end<\/p>\n
Scale
\nchr12:
\n500 bases<\/p>\n
hg38
\n85,420,000 85,420,500 85,421,000<\/p>\n
Scale
\nNC_045512v2:
\n500 bases
\n28,900 29,100 29,300 29,500 29,700 29,900
\nNCBI Genes from NC_045512.2<\/p>\n
N<\/p>\n
ORF10
\nHuman SARS-CoV-2
\nHuman Chr12<\/p>\n
SARS-CoV-2
\nScale
\nchr12:
\n—><\/p>\n
10 bases
\n85,420,490 85,420,500 85,420,510 85,420,520
\nT TCAGGGT TCAAAACCCGCT TCCCA T A TTTTTTTCCTTTCTTTAA
\nLINE1 endonuclease recognition sequence (TTTT|C)<\/p>\n
Exon\/UTR
\nn=4
\n23.5%
\nIntergenic
\nn=4
\n23.5%
\nIntron
\nn=9
\n53%
\nIllumina<\/p>\n
hg38
\n0 2<\/p>\n
Fig. 1. SARS-CoV-2 RNA can be reverse transcribed and integrated into the host cell genome. (A) Experimental workflow. (B) Chimeric sequence from a
\nNanopore sequencing read showing integration of a full-length SARS-CoV-2 NC subgenomic RNA sequence (magenta) and human genomic sequences (blue)
\nflanking both sides of the integrated viral sequence. Features indicative of LINE1-mediated \u201ctarget-primed reverse transcription\u201d include the target site
\nduplication (yellow highlight) and the LINE1 endonuclease recognition sequence (underlined). Sequences that could be mapped to both genomes are shown
\nin purple with mismatches to the human genomic sequences in italics. The arrows indicate sequence orientation with regard to the human and SARS-CoV-2
\ngenomes as shown in C and D. (C) Alignment of the Nanopore read in B with the human genome (chromosome X) showing the integration site. The human
\nsequences at the junction region show the target site, which was duplicated when the SARS-CoV-2 cDNA was integrated (yellow highlight) and the LINE1
\nendonuclease recognition sequence (underlined). (D) Alignment of the Nanopore read in B with the SARS-CoV-2 genome showing the integrated viral DNA is
\na copy of the full-length NC subgenomic RNA. The light blue highlighted regions are enlarged to show TRS-L (I) and TRS-B (II) sequences (underlined, these are
\nthe sequences where the viral polymerase jumps to generate the subgenomic RNA) and the end of the viral sequence at the poly(A) tail (III). These viral
\nsequence features (I\u2013III) show that a DNA copy of the full-length NC subgenomic RNA was retro-integrated. (E) A human\u2013viral chimeric read pair from
\nIllumina paired-end whole-genome sequencing. The read pair is shown with alignment to the human (blue) and SARS-CoV-2 (magenta) genomes. The arrows
\nindicate the read orientations relative to the human and SARS-CoV-2 genomes. The highlighted (light blue) region of the human read mapping is enlarged to
\nshow the LINE1 recognition sequence (underlined). (F) Distributions of human\u2013CoV2 chimeric junctions from Nanopore (Left) and Illumina (Right) sequencing
\nwith regard to features of the human genome.
\nZhang et al. PNAS | 3 of 10
\nReverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
https:\/\/doi.org\/10.1073\/pnas.2105968118
\nMEDICAL SCIENCES<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
templates, which can occur during the cDNA synthesis step in
\nthe preparation of a RNA-seq library (57). Thus, because there is
\na mixture of host mRNAs and positive-strand viral mRNAs in
\ninfected cells, the identification of genuine chimeric viral\u2013
\ncellular RNA transcripts is compromised by the generation of
\nartifactual chimeras in the assays.
\nWe reasoned that the orientation of an integrated DNA copy
\nof SARS-CoV-2 RNA should be random with respect to the
\norientation of the targeted host gene, predicting that about half
\nthe viral DNAs that were integrated into an expressed host gene
\nshould be in an orientation opposite to the direction of the host
\ncell gene\u2019s transcription (Fig. 3A). As predicted, \u223c50% of viral<\/p>\n
integrants in human genes were in the opposite orientation rel-
\native to the host gene in our Nanopore dataset (integration at<\/p>\n
human genes with LINE1 recognition sequences, Fig. 3B). Thus,
\nfor chimeric transcripts derived from integrated viral sequences,
\nwe would expect that \u223c50% of the chimeric transcripts should
\ncontain negative-strand viral sequences linked to positive-strand
\nhost RNA sequences. We therefore determined the fraction of<\/p>\n
the viral and human\u2013viral chimeric transcripts in infected cul-
\ntured cells\/organoids and in patient-derived tissues containing<\/p>\n
negative-strand viral RNA sequences.
\nThe replication of SARS-CoV2 RNA requires the synthesis of<\/p>\n
negative-strand viral RNA, which serves as template for repli-
\ncation of viral genomic RNA and transcription of viral sub-
\ngenomic positive-strand RNA (21). To assess the prevalence of<\/p>\n
negative-strand viral RNA in acutely infected cells, we<\/p>\n
determined the ratio of total positive to negative-strand RNAs.
\nBetween 0 and 0.1% of total viral reads were derived from
\nnegative-strand RNA in acutely infected Calu3 cells or lung
\norganoids [our data and published data (50, 58)] (Fig. 3C and SI
\nAppendix, Table S1), similar to what has been reported in clinical
\nsamples taken early after infection (59). These results argue that
\nthe level of negative-strand viral RNA is at least 1,000-fold lower
\nthan that of positive-strand viral RNA in acutely infected cells,
\ndue at least in part to a massive production of positive-strand
\nsubgenomic RNA during viral replication. This greatly reduces
\nthe likelihood that random template switching during the reverse
\ntranscription step in the RNA-seq library construction would
\ngenerate a large fraction of the artifactual chimeric reads that
\nwould contain viral negative-strand RNA fused to cellular
\npositive-strand RNA sequences. We determined that between<\/p>\n
0 and 1% of human\u2013viral chimeric reads contained negative-
\nstrand viral sequences in the acutely infected cells\/organoids<\/p>\n
(Fig. 3D and SI Appendix, Table S1), consistent with a small
\nfraction of viral reads being derived from integrated SARS-CoV-2
\nsequences.
\nIn contrast to the results obtained with acutely infected Calu3
\ncells or lung organoids, up to 51% of all viral reads, and up to
\n42.5% of human\u2013viral chimeric reads, were derived from the<\/p>\n
negative-strand SARS-CoV-2 RNA in some patient-derived tis-
\nsues [published data (60, 61), patient clinical background avail-
\nable in the original publications] (Fig. 3 E\u2013G and SI Appendix,<\/p>\n
Tables S2 and S3). Single-cell analysis of patient lung
\nTable 1. Summary of the human-CoV2 chimeric sequences obtained by Nanopore DNA sequencing of infected LINE1-overexpressing
\nHEK293T cells
\nNumber of sequences with
\nhuman-CoV2 junction<\/p>\n
With LINE1 recognition sequence at\/near
\njunction (e.g., TTTT\/A)<\/p>\n
Junction at human
\nintergenic<\/p>\n
Junction at
\nhuman intron<\/p>\n
Junction at human
\nexon\/UTR
\nchr1 10 6 0 6 4
\nchr2 2 2 0 2 0
\nchr3 3 3 0 3 0
\nchr4 2 2 0 1 1
\nchr5 1 1 0 1 0
\nchr6 4 2 3 0 1
\nchr7 2 2 1 1 0
\nchr8 0 0 0 0 0
\nchr9 4 2 0 2 2
\nchr10 5 1 2 1 2
\nchr11 3 2 1 1 1
\nchr12 6 4 2 2 2
\nchr13 3 3 3 0 0
\nchr14 2 2 1 1 0
\nchr15 0 0 0 0 0
\nchr16 2 1 1 1 0
\nchr17 2 0 1 0 1
\nchr18 2 1 0 2 0
\nchr19 1 1 0 0 1
\nchr20 0 0 0 0 0
\nchr21 2 1 1 1 0
\nchr22 1 1 0 1 0
\nchrX 6 5 2 1 3
\nTotal 63 42 18 27 18
\nFraction 66.7% 28.6% 42.9% 28.6%<\/p>\n
Table 2. Summary of the human-CoV2 chimeric sequences obtained by Illumina paired-end
\nwhole-genome DNA sequencing of infected LINE1-overexpressing HEK293T cells
\nRegion features (human) Intergenic Intron Exon\/UTR
\nRegion number 4 9 4
\nWith L1 recognition sequence at\/near junction 2 3 2<\/p>\n
4 of 10 | PNAS Zhang et al.
\nhttps:\/\/doi.org\/10.1073\/pnas.2105968118 Reverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
A<\/p>\n
B<\/p>\n
C<\/p>\n
F
\nCell line Viral primer Human LINE1 recognition Human genomic
\ntarget chromosome sequence feature
\nHEK293T Near 3\u2019 end of Chr15 TTTT|G Intergenic
\nviral genome
\nChr1 TTTT|G Intergenic
\nCalu3 (rep1) Near 5\u2019 end of Chr18 CTTT|A Intergenic
\nNC gene
\nChr2 TTTA|A UTR
\nChr12 TTTA|A Intron
\nCalu3 (rep2) Near 5\u2019 end of Chr12 TTTT|C Intron
\nNC gene
\nChr4 TTTT|A Intron
\nHuman-CoV2 chimeric sequence summary:
\nRead 1 Read 2
\n\u201cHuman-CoV2\u201d chimeric read (HEK293T)
\nnamuH 2-VoC-SRAS<\/p>\n
CTCAAACAGCCCTGCTTCAACTAGGGGAGAAAACACAGTGTTTGAATACCA
\nTGTGATGGTATCCATCCTGTTCCAGGTGGAGGATGCAGAGGATGGCCCCC
\nTGCACCTTCCAGGATAAGAAGATCCTGATGCAGTCAACTTACCACCAGGA
\n5\u2019–<\/p>\n
–3\u2019<\/p>\n
Read 1 (151 nt):<\/p>\n
5\u2019–<\/p>\n
–3\u2019
\nAGCGGAGTACGATCGAGTGTACAGTGAACAATGCTAGGGAGAGCTGCCTA
\nTATGGAAGAGCCCTAATGTGTAAAATTAATTTTAGTAGTGCTATCCCCATGT
\nGATTTTAATAGCTTCTTAGGAGAATGACAAAAAAAAAAAAAAAAGAATG
\nRead 2 (151 nt): primer<\/p>\n
\u201cHuman-CoV2\u201d chimeric read (HEK293T)
\nalignment on human Chr15
\nScale
\nchr15:
\n500 bases<\/p>\n
hg38
\n68,544,500 68,545,000<\/p>\n
Scale
\nchr15:
\n—>
\n10 bases
\n68,544,775 68,544,785 68,544,795
\nCATACACATTCTTTTTTTTTTTTTTTTTTTGAGATG
\n<— GTATGT GTAAG AAAAAAAAAAAAAAAAAAACTCTAC<\/p>\n
\u201cHuman-CoV2\u201d chimeric read (HEK293T)
\nalignment on the SARS-CoV-2 genome<\/p>\n
Scale
\nNC_045512v2:
\n500 bases
\n28,900 29,100 29,300 29,500 29,700 29,900
\nNCBI Genes from NC_045512.2<\/p>\n
N<\/p>\n
ORF10
\nprimer<\/p>\n
D<\/p>\n
E
\nLINE1 endonuclease recognition sequence (TTTT|G)
\nT T T TTT T T T TTT T T T TTTT
\nA A AAA A A A AAA A A A AAAA A<\/p>\n
SARS-CoV-2 Human<\/p>\n
Random adapter tag
\n-mentation by Tn5<\/p>\n
Primers<\/p>\n
PCR enrichment and
\nIllumina paired-end sequencing<\/p>\n
SARS-CoV-2 genome<\/p>\n
CGCGGAGTACGATCGAGTG
\nDay:
\nHEK293T or Calu3
\n+ SARS-CoV-2 infection
\n1
\nCell DNA extraction
\n(+RNase)<\/p>\n
PCR enrichment and Illumina
\npaired-end sequencing<\/p>\n
0 2<\/p>\n
Fig. 2. Evidence for integration of SARS-CoV-2 cDNA in cultured cells that do not overexpress a reverse transcriptase. (A) Experimental workflow. (B) Ex-
\nperimental design for the Tn5 tagmentation-mediated enrichment sequencing method used to map integration sites in the host cell genome. (C) A<\/p>\n
human\u2013viral chimeric read pair supporting viral integration. The reads are aligned with the human (blue) and SARS-CoV-2 (magenta) genomic sequences. The
\narrows indicate the read orientations relative to the human and SARS-CoV-2 genomes as shown in D and E. Sequence of the viral primer used for enrichment
\nis shown with green highlight in the read (corresponding to the green arrow illustrated in B). Sequences that could be mapped to both genomes are shown in
\npurple. (D) Alignment of the read pair in C with the human genome (chromosome 15, blue arrow). The highlighted (light blue) region of the human sequence<\/p>\n
is enlarged to show the LINE1 recognition sequence (underlined) with a 19-base poly-dT sequence (purple highlight) that could be annealed by the viral poly-
\nA tail for \u201ctarget-primed reverse transcription.\u201d Additional 5-bp human sequence (GAATG, blue) was captured in read 2 (C), supporting a bona fide inte-
\ngration site. (E) Alignment of the read pair in C with the SARS-CoV-2 genome (magenta). The viral primer sequence is shown with green highlight. (F)<\/p>\n
Summary of seven human\u2013viral chimeric sequences identified by the enrichment sequencing method in the two cell lines showing the integrated human
\nchromosomes, LINE1 recognition sequences close to the chimeric junction, and human genomic features at the read junction.
\nZhang et al. PNAS | 5 of 10
\nReverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
https:\/\/doi.org\/10.1073\/pnas.2105968118
\nMEDICAL SCIENCES<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
bronchoalveolar lavage fluid (BALF) cells from patients with
\nsevere COVID [published data (61)] showed that up to 40% of
\nall viral reads were derived from the negative-strand SARS-CoV-2
\nRNA (SI Appendix, Fig. S7). Fractions of negative-strand RNA in
\ntissues from some patients were orders of magnitude higher than
\nthose in acutely infected cells or organoids (Fig. 3 C\u2013G). In fixed
\n(formalin-fixed, paraffin-embedded [FFPE]) autopsy samples, in 4
\nout of 14 patients (Fig. 3E and SI Appendix, Table S2), and in
\nBALF samples, in 4 out of 6 patients (Fig. 3G and SI Appendix,<\/p>\n
Table S3), at least \u223c20% of the viral reads were derived from
\nnegative-strand viral RNA. In contrast to acutely infected cells
\n(Fig. 3 C and D and SI Appendix, Table S1), there was little or no
\nevidence for virus reproduction in these autopsy samples (60). As
\nsummarized in SI Appendix, Table S2, there were negative-strand
\nviral sequences in a large fraction of the human\u2013viral chimeric
\nreads (up to \u223c40%) in samples from one patient. Different
\nsamples derived from the same patient revealed a similarly high
\nfraction of negative viral strand\u2013human RNA reads. Several<\/p>\n
C<\/p>\n
F<\/p>\n
B<\/p>\n
Patient case<\/p>\n
AAAAAA
\nAAAAAA<\/p>\n
RNA-seq:
\n~100% CoV2 reads from positve strand
\nPositive strand CoV2 (sub)genomic RNA Random integration and transcription
\nHuman gene CoV2 integrant
\n~50% same orientation: CoV2 reads from positive strand<\/p>\n
A<\/p>\n
D
\nFraction of CoV2 chimeric
\nreads from negaitive strand<\/p>\n
Fraction of CoV2 chimeric
\nreads from negative strand
\nCalu3
\nLung organoid<\/p>\n
Patient case<\/p>\n
Fraction of CoV2 reads
\nfrom negative strand
\n1 2 3 5 6 7 8 9 10 11 A B C D
\n0.0001
\n0.001
\n0.01
\n0.1
\n1<\/p>\n
0
\n0.5<\/p>\n
1 8 9 11 C D
\n0.0001
\n0.001
\n0.01
\n0.1
\n1<\/p>\n
0
\n0.5
\n0.0001
\n0.001
\n0.01
\n0.1
\n1<\/p>\n
Fraction of CoV2 reads<\/p>\n
0<\/p>\n
from negative strand
\nCalu3
\nLung organoid
\n0.0001
\n0.001
\n0.01
\n0.1
\n1<\/p>\n
0<\/p>\n
G
\nSame Opposite
\n0.0
\n0.2
\n0.4
\n0.6
\n0.8
\nRelative orientation:
\nHuman gene vs
\nCoV2 integrant
\nFraction
\n15
\n(54%)
\n13
\n(46%)<\/p>\n
~50% opposite orientation: CoV2 reads from negative strand<\/p>\n
EFraction of CoV2 reads
\nfrom negative strand
\nC143
\nC145
\nC146
\nC148
\nC149
\nC152<\/p>\n
0.0001
\n0.001
\n0.01
\n0.1
\n1<\/p>\n
0
\n0.5<\/p>\n
Patient case<\/p>\n
Fig. 3. Negative-strand viral RNA-seq reads suggest that integrated SARS-CoV-2 sequences are expressed. (A) Schema predicting fractions of positive- or
\nnegative-strand SARS-CoV-2 RNA-seq reads that are derived from viral (sub)genomic RNAs or from transcripts of integrated viral sequences. The arrows
\n(Right) showing the orientation of an integrated SARS-CoV-2 (magenta) positive strand relative to the orientation of the host cellular gene (blue). (B)
\nFractions of SARS-CoV-2 sequences integrated into human genes with same (n = 15) or opposite (n = 13) orientation of the viral positive strand relative to the
\npositive strand of the human gene. A total of 28 integration events at human genes with LINE1 endonuclease recognition sequences were identified from our
\nNanopore DNA sequencing of infected LINE1-overexpressing HEK293T cells (Fig. 1A). (C) Fraction of total viral reads that are derived from negative-strand
\nviral RNA in acutely infected cells or organoids (see SI Appendix, Table S1 for details). (D) Fraction of human\u2013viral chimeric reads that contain viral sequences
\nderived from negative-strand viral RNA in acutely infected cells or organoids (see SI Appendix, Table S1 for details). (E) Fraction of total viral reads that are
\nderived from negative-strand viral RNA in published patient RNA-seq data (autopsy FFPE samples, GSE150316, samples with no viral reads or of low library
\nstrandedness quality not included; see SI Appendix, Table S2 for details; reanalysis results consistent with the original publication). (F) Fraction of human\u2013viral
\nchimeric reads that contain viral sequences derived from negative-strand viral RNA in published patient RNA-seq data (autopsy FFPE samples, GSE150316; see
\nSI Appendix, Table S2 for details). (G) Fraction of total viral reads that are derived from negative-strand viral RNA in published patient RNA-seq data (BALF
\nsamples, GSE145926; see SI Appendix, Table S3 for details). The red dashed lines in E\u2013G indicate the level at which 50% of all viral reads (E and G) or viral
\nsequences in human\u2013viral chimeric reads (F) were from negative-strand viral RNAs, a level expected if all the viral sequences were derived from integrated
\nsequences.<\/p>\n
6 of 10 | PNAS Zhang et al.
\nhttps:\/\/doi.org\/10.1073\/pnas.2105968118 Reverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
other patient samples revealed lower fraction of negative viral
\nstrand RNA\u2013human RNA chimeras, which were, however, still
\nsignificantly higher than what was found in acutely infected cells
\n(Fig. 3 D and F and SI Appendix, Table S1 and S2). Because the
\nability to identify viral\u2013human chimeric reads using short-read
\nRNA-seq is limited, our analysis failed to show significant<\/p>\n
numbers of chimeric reads in patient BALF samples (SI Appen-
\ndix, Table S3). In summary, our data suggest that in some patient-
\nderived tissues, where the total number of SARS-CoV-2 sequence-
\npositive cells may be small, a large fraction of the viral transcripts<\/p>\n
could have been transcribed from SARS-CoV-2 sequences inte-
\ngrated into the host genome.<\/p>\n
Discussion
\nWe present here evidence that SARS-CoV-2 sequences can be
\nreverse-transcribed and integrated into the DNA of infected
\nhuman cells in culture. For two of the integrants, we recovered
\n\u201chuman\u2013viral\u2013human\u201d chimeric reads encompassing a direct
\ntarget site repeat (20 or 13 bp), and a consensus recognition site
\nof the LINE1 endonuclease was present on both ends of the host
\nDNA that flanked the viral sequences. These and other data are<\/p>\n
consistent with a target primed reverse transcription and retro-
\nposition integration mechanism (41, 42) and suggest that en-
\ndogenous LINE1 RT can be involved in the reverse transcription<\/p>\n
and integration of SARS-CoV-2 sequences in the genomes of
\ninfected cells.
\nApproximately 30% of viral integrants analyzed in cultured<\/p>\n
cells lacked a recognizable nearby LINE1 endonuclease recog-
\nnition site. Thus, it is also possible that integration can occur by<\/p>\n
another mechanism. Indeed, there is evidence that chimeric
\ncDNAs can be produced in cells acutely infected with LCMV by
\ncopy choice with endogenous IAP elements during reverse
\ntranscription. This mechanism is expected to create a chimeric
\ncDNA complementary to both LCMV and IAP. In some cases,<\/p>\n
the resulting chimeric cDNAs were integrated without the gen-
\neration of a target site duplication (29). A recent study has also<\/p>\n
suggested that the interaction between coronavirus sequences<\/p>\n
and endogenous retrotransposon could be a potential viral in-
\ntegration mechanism (40).<\/p>\n
It will be important, in follow-up studies, to demonstrate the
\npresence of SARS-CoV-2 sequences integrated into the host
\ngenome in patient tissues. However, this will be technically
\nchallenging because only a small fraction of cells in any patient
\ntissues are expected to be positive for viral sequences (61).<\/p>\n
Consistent with this notion, it has been estimated that only be-
\ntween 1 in 1,000 and 1 in 100,000 mouse cells infected with<\/p>\n
LCMV either in culture or in the animal carried viral DNA
\ncopies integrated into the genome (30). In addition, only a
\nfraction of patients may carry SARS-CoV-2 sequences integrated
\nin the DNA of some cells. However, with more than 140 million
\nhumans infected with SARS-CoV-2 worldwide (as of April,
\n2021), even a rare event could be of significant clinical relevance.<\/p>\n
It is also challenging to estimate the frequency of retro-
\nintegration events in cell culture assays since infected cells usu-
\nally die and are lost before sample collection. For the same<\/p>\n
reason, no clonal expansion of integrated cells is expected in
\nacute infection experiments. Moreover, the chance of integration
\nat the same genomic locus in different patients\/tissues may be
\nlow, due to a random integration process.
\nThe presence of chimeric virus\u2013host RNAs in cells cannot
\nalone be taken as strong evidence for transcription of integrated
\nviral sequences because template switching can happen during
\nthe reverse transcription step of cDNA library preparation.
\nHowever, we found that only a very small fraction (0\u20131%) of<\/p>\n
chimeric reads from acutely infected cells contained negative-
\nstrand viral RNA sequences, whereas, in the RNA-seq libraries<\/p>\n
prepared from some patients, the fraction of total viral reads,
\nand the fraction of human\u2013viral chimeric reads that were derived<\/p>\n
from negative-strand SARS-CoV-2 RNAs was substantially
\nhigher. For retrotransposon-mediated integration events, the
\norientation of the reverse-transcribed SARS-CoV-2 RNA should
\nbe random with respect to the orientation of a host gene. Thus,
\nfor chimeric RNAs derived from integrated viral sequences,
\nabout half of the chimeric reads will link positive-strand host
\nRNA sequences to negative-strand viral sequences. In some
\npatient samples, negative-strand viral reads accounted for
\n40\u201350% of the total viral RNA sequences and a similar fraction<\/p>\n
of the chimeric reads contained negative-strand viral RNA se-
\nquences, suggesting that the majority if not all of the viral RNAs<\/p>\n
in these samples were derived from integrated viral sequences.
\nIt is important to note that, because we have detected only
\nsubgenomic sequences derived mainly from the 3\u2032 end of the
\nviral genome integrated into the DNA of the host cell, infectious
\nvirus cannot be produced from such integrated subgenomic<\/p>\n
SARS-CoV-2 sequences. The possibility that SARS-CoV-2 se-
\nquences can be integrated into the human genome and expressed<\/p>\n
in the form of chimeric RNAs raises several questions for future
\nstudies. Do integrated SARS-CoV-2 sequences express viral
\nantigens in patients and might these influence the clinical course
\nof the disease? The available clinical evidence suggests that, at
\nmost, only a small fraction of the cells in patient tissues express<\/p>\n
viral proteins at a level that is detectable by immunohisto-
\nchemistry. However, if a cell with an integrated and expressed<\/p>\n
SARS-CoV-2 sequences survives and presents a viral- or neo-
\nantigen after the infection is cleared, this might engender con-
\ntinuous stimulation of immunity without producing infectious<\/p>\n
virus and could trigger a protective response or conditions such
\nas autoimmunity as has been observed in some patients (62, 63).
\nThe presence of LCMV sequences integrated in the genomes of
\nacutely infected cells in mice led the authors to speculate that
\nexpression of such sequences \u201cpotentially represents a naturally
\nproduced form of DNA vaccine\u201d (30). It is not known how many
\nantigen-presenting cells are needed to elicit an antigen response,
\nbut derepressed LINE1 expression, induced by viral infection or
\nby exposure to cytokines (38\u201340), may stimulate SARS-CoV-2
\nintegration into the genome of infected cells in patients. More
\ngenerally, our results suggest that integration of viral DNA in
\nsomatic cells may represent a consequence of a natural infection<\/p>\n
that could play a role in the effects of other common disease-
\ncausing RNA viruses such as dengue, Zika, or influenza virus.<\/p>\n
Our results may also be relevant for current clinical trials of
\nantiviral therapies (64). If integration and expression of viral
\nRNA are fairly common, reliance on extremely sensitive PCR
\ntests to determine the effect of treatments on viral replication
\nand viral load may not always reflect the ability of the treatment
\nto fully suppress viral replication because the PCR assays may
\ndetect viral transcripts that derive from viral DNA sequences
\nthat have been stably integrated into the genome rather than
\ninfectious virus.
\nMaterials and Methods
\nCell Culture and Plasmid Transfection. HEK293T cells were obtained from ATCC
\n(CRL-3216) and cultured in DMEM supplemented with 10% heat-inactivated
\nFBS (HyClone; SH30396.03) and 2 mM L-glutamine (MP Biomedicals;
\nIC10180683) following ATCC\u2019s method. Calu3 cells were obtained from ATCC
\n(HTB-55) and cultured in EMEM (ATCC; 30-2003) supplemented with 10%
\nheat-inactivated FBS (HyClone; SH30396.03) following ATCC\u2019s method.
\nPlasmids for human LINE1 expression, pBS-L1PA1-CH-mneo (CMV-LINE-1),
\nwas a gift from Astrid Roy-Engel, Tulane University Health Sciences Center,
\nNew Orleans, LA (Addgene plasmid #51288 ; http:\/\/addgene.org\/51288;
\nRRID:Addgene_51288) (65); EF06R (5\u2032UTR-LINE-1) was a gift from Eline
\nLuning Prak, University of Pennsylvania, Philadelphia, PA (Addgene plasmid
\n#42940 ; http:\/\/addgene.org\/42940; RRID:Addgene_42940) (66). Transfection
\nwas done with Lipofectamine 3000 (Invitrogen; L3000001) following
\nmanufacturer\u2019s protocol.<\/p>\n
Zhang et al. PNAS | 7 of 10
\nReverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
https:\/\/doi.org\/10.1073\/pnas.2105968118
\nMEDICAL SCIENCES<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
SARS-CoV-2 Infection. SARS-CoV-2 USA-WA1\/2020 (GenBank: MN985325.1)
\nwas obtained from BEI Resources and expanded and tittered on Vero cells.
\nCells were infected in DMEM plus 2% FBS for 48 h using a multiplicity of
\ninfection (MOI) of 0.5 for infection of HEK293T cells and an MOI of 1 or 2 for
\nCalu3 cells. All sample processing and harvest with infectious virus were
\ndone in the BSL3 facility at the Ragon Institute.
\nNucleic Acids Extraction and PCR Assay. Cellular DNA extraction was done using
\na published method (31). For purification of genomic DNA, total cellular DNA
\nwas fractionated on a 0.4% (wt\/vol) agarose\/1\u00d7 TAE gel for 1.5 h with a 3 V\/cm
\nvoltage, with \u03bb DNA-HindIII Digest (NEB; N3012S) as size markers. Large
\nfragments (>23.13 kb) were cut out, frozen in \u221280 \u00b0C, and then crushed with a
\npipette tip. Three volumes (vol\/wt) of high T-E buffer (10 mM Tris\u201310 mM<\/p>\n
EDTA, pH 8.0) were added, and then NaCl was added to give a final concen-
\ntration of 200 mM. The gel solution was heated at 70 \u00b0C for 15 min with<\/p>\n
constant mixing and then extracted with phenol:chloroform:isoamyl alcohol<\/p>\n
(25:24:1, vol\/vol\/vol) (Life Technologies; 15593031) and chloroform:isoamyl al-
\ncohol 24:1 (Sigma; C0549-1PT). DNA was precipitated by the addition of so-
\ndium acetate and isopropyl alcohol. For samples with low DNA concentration,<\/p>\n
glycogen (Life Technologies; 10814010) was added as a carrier to aid
\nprecipitation.
\nRNA extraction was done with RNeasy Plus Micro Kit (Qiagen; 74034)
\nfollowing manufacturer\u2019s protocol.<\/p>\n
To detect DNA copies of SARS-CoV-2 sequences, we chose four NC gene-
\ntargeting PCR primer sets that are used in COVID-19 tests [SI Appendix, Fig.<\/p>\n
S1A, primer source from World Health Organization (67), modified to match
\nthe genome version of NC_045512.2]. See SI Appendix, Table S4 for PCR
\nprimer sequences used in this study. PCR was done using AccuPrime Taq DNA
\nPolymerase, high fidelity (Life Technologies; 12346094). PCR products were
\nrun on 1% or 2% (wt\/vol) agarose gel to show amplifications.
\nNanopore DNA Sequencing and Analysis. A total of 1.6 \u03bcg of DNA extracted
\nfrom HEK293T cells transfected with the pBS-L1PA1-CH-mneo (CMV-LINE-1)<\/p>\n
plasmid and infected with SARS-CoV-2 was used to make a sequencing li-
\nbrary with the SQK-LSK109 kit (Oxford Nanopore Technologies) and se-
\nquenced on one R9 PromethION flowcell (FLO-PRO002) for 3 d and 5 min.<\/p>\n
The sequencing data were base-called using Guppy 4.0.11 (Oxford Nanopore
\nTechnologies) using the high-accuracy model.<\/p>\n
Nanopore reads were mapped using minimap2 (68) (version 2.15) with pa-
\nrameters \u201c-p 0.3 -ax map-ont\u201d and a fasta file containing the human genome<\/p>\n
sequence from ENSEMBL release 93 (ftp:\/\/ftp.ensembl.org\/pub\/release-93\/fasta\/<\/p>\n
homo_sapiens\/dna\/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz) concat-
\nenated to the SARS-CoV-2 sequence, GenBank ID: MN988713.1, \u201cSevere acute<\/p>\n
respiratory syndrome coronavirus 2 isolate SARS-CoV-2\/human\/USA\/IL-CDC-IL1\/
\n2020, complete genome.\u201d From the SAM file, we selected all the sequences that<\/p>\n
mapped to the viral genome and divided them into groups based on the hu-
\nman chromosomes they mapped to. We blasted the selected sequences, using<\/p>\n
blastn, against a BLAST database made with the human and virus sequences
\ndescribed above. We parsed the blast output into a text file containing one row<\/p>\n
per high-scoring segment pair (HSP) with a custom perl script. We further fil-
\ntered that file, for each sequence, by selecting all the viral HSPs and the top<\/p>\n
three human HSPs. We inspected those files visually to identify sequences
\ncontaining human\u2013viral\u2013human or human\u2013viral junctions. For a few sequences,<\/p>\n
longer than 30 kb, we inspected the top 15 human HSPs. Additionally, we vi-
\nsually inspected all the identified reads containing human and viral sequences<\/p>\n
by the University of California, Santa Cruz (UCSC) BLAT (69) tool. Due to errors
\nin Nanopore sequencing and\/or base-calling, artifactual \u201chybrid sequences\u201d
\nexist in a subset of these reads, sometimes with Watson and Crick strands from
\nthe same DNA fragment present in the same read. Therefore, we only focused
\non chimeric sequences showing clear human\u2013viral junctions and analyzed
\nknown LINE1-mediated retroposition features such as target-site duplications
\nand LINE1 endonuclease recognition sequences for evidence of integration.<\/p>\n
Tn5 Tagmentation-Mediated Integration Site Enrichment. We used a tagmentation-
\nbased method to enrich for viral integration sites (47, 48). Briefly, we used Tn5<\/p>\n
transposase (Diagenode; C01070010) to randomly tagment the cellular DNA
\nwith adapters (adapter A, the Illumina Nextera system). Tagmentation was done
\nusing 100 ng of DNA for 10 min at 55 \u00b0C, followed by stripping off the Tn5
\ntransposase from the DNA with SDS. We used a reverse primer targeting the
\nnear-5\u2032 end of SARS-CoV-2 NC gene (CCAAGACGCAGTATTATTGGGTAAA) or a<\/p>\n
forward primer targeting the near-3\u2032 end of SARS-CoV-2 genome (CTTGTGCAG-
\nAATGAATTCTCGTAACT) to linearly amplify (PCR0, 45 cycles) the tagmented DNA<\/p>\n
fragments containing viral sequences. We took the product of PCR0 and am-
\nplified the DNA fragments containing adapter and viral sequences (potential<\/p>\n
integration sites) using 15\u201320 cycles of PCR1, with a barcoded (i5) Nextera primer<\/p>\n
(AATGATACGGCGACCACCGAGATCTACACNNNNNNNNTCGTCGGCAGCGTC,
\nNNNNNNNN indicates the barcode) against the adapter sequence and a viral
\nprimer. The viral primer was designed to either target the near-5\u2032 end of
\nSARS-CoV-2 NC gene (GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGCCGAC
\nGTTGTTTTGATCG, viral sequence underlined) or target the near-3\u2032 end of
\nSRAS-CoV-2 genome (GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCGCGGA
\nGTACGATCGAGTG, viral sequence underlined). The viral primer also contained
\nan adapter sequence for further PCR amplification. We amplified the PCR1
\nproduct by 15\u201320 cycles of PCR2, using a short primer (AATGATACGGCGACCACC
\nGA) against the i5 Nextera primer sequence and a barcoded (i7) Nextera primer
\n(CAAGCAGAAGACGGCATACGAGATNNNNNNNNGTCTCGTGGGCTCGG,
\nNNNNNNNN indicates the barcode) against the adapter sequence introduced by<\/p>\n
the viral primer in PCR1. The final product of the PCR2 amplification was frac-
\ntionated on 1.5% agarose gel (Sage Science; HTC1510) with PippinHT (Sage<\/p>\n
Science; HTP0001) and 500- to 1,000-bp pieces were selected for Illumina paired-
\nend sequencing. All three PCR steps (PCR0\u2013PCR2) were done with KAPA HiFi<\/p>\n
HotStart ReadyMix (KAPA;KK2602).
\nIllumina DNA Sequencing and Analysis. We constructed libraries for HEK293T cell<\/p>\n
whole-genome sequencing using the Tn5-based Illumina DNA Prep kit (Illu-
\nmina; 20018704). The whole-genome sequencing libraries or the libraries from<\/p>\n
Tn5-mediated integration site enrichment after sizing (described above) were<\/p>\n
subjected to Illumina sequencing. qPCR was used to measure the concentra-
\ntions of each library using KAPA qPCR library quant kit according to the<\/p>\n
manufacturer\u2019s protocol. Libraries were then pooled at equimolar concentra-
\ntions, for each lane, based on qPCR concentrations. The pooled libraries were<\/p>\n
denatured using the Illumina protocol. The denatured libraries were loaded
\nonto an SP flowcell on an Illumina NovaSeq 6000 and run for 2 \u00d7 150 cycles
\naccording to the manufacturer\u2019s instructions. Fastq files were generated and
\ndemultiplexed with the bcl2fastq Conversion Software (Illumina).
\nTo identify human\u2013SARS-CoV-2 chimeric DNA reads, raw sequencing reads were
\naligned with STAR (70) (version 2.7.1a) to a human plus SARS-CoV-2 genome made
\nwith a fasta file containing the human genome sequence version hg38 with no
\nalternative chromosomes concatenated to the SARS-CoV-2 sequence from National
\nCenter for Biotechnology Information (NCBI) reference sequence NC_045512.2. The
\nfollowing STAR parameters were used to call chimeric reads: \u2013alignIntronMax 1
\n\\\u2013chimOutType Junctions SeparateSAMold WithinBAM HardClip
\n\\\u2013chimScoreJunctionNonGTAG 0 \\\u2013alignSJstitchMismatchNmax -1\u20131 -1\u20131
\n\\\u2013chimSegmentMin 25 \\\u2013chimJunctionOverhangMin 25 \\\u2013outSAMtype
\nBAM SortedByCoordinate. We extracted viral reads from the generated
\nBAM file by samtools (71) (version 1.11) using command: samtools view -b<\/p>\n
Aligned.sortedByCoord.out.bam NC_045512v2 > NC_Aligned.sortedBy-
\nCoord.out.bam. We extracted human\u2013viral chimeric reads by using the read<\/p>\n
names from the STAR generated Chimeric.out.junction file to get the read
\nalignments from the STAR generated Chimeric.out.sam file by Picard (http:\/\/<\/p>\n
broadinstitute.github.io\/picard), using command: java -jar picard.jar Filter-
\nSamReads I = Chimeric.out.sam O = hv-Chimeric.out.sam READ_LIST_FILE = hv-
\nChimeric.out.junction.ids FILTER = includeReadList. We further confirmed each<\/p>\n
of the chimeric reads and filtered out any unconvincing reads (too short or
\naligned to multiple sites of the human genome) by visual inspection with
\nthe UCSC BLAT (69) tool. We also loaded the STAR generated Aligned.-
\nsortedByCoord.out.bam file or the NC_Aligned.sortedByCoord.out.bam file
\ncontaining extracted viral reads to the UCSC browser SARS-CoV-2 genome
\n(NC_045512.2) to search for additional chimeric reads that were missed by
\nthe STAR chimeric calling method. To generate genome coverage file, we
\nused the bamCoverage from the deepTools suite (72) (version 3.5.0) to convert
\nthe STAR generated Aligned.sortedByCoord.out.bam file to a bigwig file binned
\nat 10 bp, using command: bamCoverage -b Aligned.sortedByCoord.out.bam -o
\nAligned.sortedByCoord.out.bw\u2013binSize 10.<\/p>\n
RNA-Seq and Analysis. To identify human\u2013SARS-CoV-2 chimeric reads, pub-
\nlished RNA-seq data were downloaded from Gene Expression Omnibus<\/p>\n
(GEO) with the accession numbers GSE147507 (50), GSE153277 (51),
\nGSE156754 (52), GSE157852 (53), GSE153684 (54), and GSE154998 (55)
\n(summarized in SI Appendix, Fig. S5C). Raw sequencing reads were aligned with
\nSTAR (70) (version 2.7.1a) to human plus SARS-CoV-2 genome and transcriptome
\nmade with a fasta file containing the human genome sequence version hg38 with
\nno alternative chromosomes concatenated to the SARS-CoV-2 sequence from NCBI<\/p>\n
reference sequence NC_045512.2, and a gtf file containing the human gene an-
\nnotations from ENSEMBL version GRCh38.97 concatenated to the SARS-CoV-2<\/p>\n
gene annotations from NCBI (http:\/\/hgdownload.soe.ucsc.edu\/goldenPath\/wuh-
\nCor1\/bigZips\/genes\/). The following STAR parameters (56) were used to call chi-
\nmeric reads unless otherwise specified (SI Appendix, Fig. S5C):\u2013chimOutType<\/p>\n
Junctions SeparateSAMold WithinBAM HardClip \\\u2013chimScoreJunctionNonGTAG
\n8 of 10 | PNAS Zhang et al.
\nhttps:\/\/doi.org\/10.1073\/pnas.2105968118 Reverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
0 \\\u2013alignSJstitchMismatchNmax -1\u20131 -1\u20131 \\\u2013chimSegmentMin 50
\n\\\u2013chimJunctionOverhangMin 50.
\nFor RNA-seq strandedness analysis, we generated RNA-seq data using RNA
\nfrom SARS-CoV-2\u2013infected Calu3 cells. Stranded libraries were constructed
\nwith the Kapa mRNA HyperPrep kit (Roche; 08098115702). Libraries were
\nqPCR\u2019ed using a KAPA qPCR library quant kit as per manufacturer\u2019s protocol.
\nLibraries were then pooled at equimolar concentrations, for each lane, based<\/p>\n
on qPCR concentrations. The pooled libraries were denatured using the Illu-
\nmina protocol. The denatured libraries were loaded onto an HiSeq 2500<\/p>\n
(Illumina) and sequenced for 120 cycles from one end of the fragments.
\nBasecalls were performed using Illumina offline basecaller (OLB) and then
\ndemultiplexed. We downloaded published RNA-seq data (stranded libraries)
\nfrom GEO with the accession numbers GSE147507 (50) (Calu3, SI Appendix,
\nTable S1), GSE148697 (58) (lung organoids, SI Appendix, Table S1), and
\nGSE150316 (60) (patient FFPE tissues, SI Appendix, Table S2). Raw RNA-seq
\nreads were aligned as described above, using parameters\u2013chimSegmentMin
\n30 \\\u2013chimJunctionOverhangMin 30 to call chimeric reads. We extracted total
\nviral reads and human\u2013viral chimeric reads as described above. We convert
\nthe viral read BAM files into Bed files using the bamToBed utility in BEDTools
\n(73). We then counted the total and stranded read numbers in the converted
\nBED files.
\nPublished single-cell RNA-seq data were downloaded from GEO with the
\naccession number GSE145926 (61) (patient BALF samples, SI Appendix, Table S3).<\/p>\n
For bulk analysis, duplicate reads with the same read1 (UMI) and read2 se-
\nquences in raw fastq files were removed by dedup_hash (https:\/\/github.com\/<\/p>\n
mvdbeek\/dedup_hash). Then the pool of read2 were aligned as described above,
\nusing parameters \u2013chimSegmentMin 30 \\\u2013chimJunctionOverhangMin 30 to call<\/p>\n
chimeric reads. Read strandedness was analyzed as described above. For single-
\ncell analysis, we generated a custom genome by Cell Ranger (10\u00d7 Genomics Cell<\/p>\n
Ranger 3.0.2) (74) mkref, using a fasta file containing the human genome<\/p>\n
sequence from ENSEMBL release 93 (ftp:\/\/ftp.ensembl.org\/pub\/release-93\/fasta\/<\/p>\n
homo_sapiens\/dna\/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz) concate-
\nnated to the SARS-CoV-2 sequence, GenBank ID: MN988713.1, and a gtf file<\/p>\n
containing human and viral annotations. Read mapping, assigning reads to cell<\/p>\n
barcodes and removing PCR duplicates were done with Cell Ranger (10\u00d7 Ge-
\nnomics Cell Ranger 4.0.0) (74) count, using the custom genome described above.<\/p>\n
We processed the counts using Seurat (version 3.2.2) (75). We removed cells that
\nhad less than 200 genes detected or more than 20% of transcript counts deriving
\nfrom the mitochondria. For each cell, we counted the number of reads mapping
\nto either the positive or negative viral strand.
\nData Availability. All data supporting the findings of this study are available
\nwithin the article and supporting information. All sequencing data generated
\nin this study have been deposited to the Sequence Read Archive, https:\/\/www.<\/p>\n
ncbi.nlm.nih.gov\/sra (accession no. PRJNA721333). All published data ana-
\nlyzed in this study are cited in this article with accession methods provided in<\/p>\n
Materials and Methods.
\nACKNOWLEDGMENTS. We thank members in the laboratories of R.J. and
\nR.A.Y. and other colleagues from Whitehead Institute and Massachusetts
\nInstitute of Technology (MIT) for helpful discussions and resources. We thank
\nThomas Volkert and staff from the Whitehead genomics core, and Stuart Levine
\nfrom the MIT\/Koch Institute BioMicro center for sequencing support. We thank<\/p>\n
Lorenzo Bombardelli for sharing protocol and advice for Tn5 tagmentation-
\nmediated integration enrichment sequencing. We thank Jerold Chun, Inder<\/p>\n
Verma, Joseph Ecker, and Daniel W. Bellott for discussion and suggestions. This
\nwork was supported by grants from the NIH to R.J. (1U19AI131135-01;
\n5R01MH104610-21) and by a generous gift from Dewpoint Therapeutics and
\nfrom Jim Stone. S.H.H. was supported by the Intramural Research Program of the
\nCenter for Cancer Research of the National Cancer Institute. Finally, we thank
\nNathans Island for inspiration.<\/p>\n
1. Korean Disease Control and Prevention Agency, Findings from investigation and analysis of
\nre-positive cases. https:\/\/www.kdca.go.kr\/board\/board.es?mid=a30402000000&bid=0030.
\nAccessed 12 June 2020.
\n2. J. Bullard et al., Predicting infectious severe acute respiratory syndrome coronavirus 2
\nfrom diagnostic samples. Clin. Infect. Dis. 71, 2663\u20132666 (2020).
\n3. X. He et al., Temporal dynamics in viral shedding and transmissibility of COVID-19.
\nNat. Med. 26, 672\u2013675 (2020).
\n4. N. Li, X. Wang, T. Lv, Prolonged SARS-CoV-2 RNA shedding: Not a rare phenomenon.
\nJ. Med. Virol. 92, 2286\u20132287 (2020).<\/p>\n
5. M. J. Mina, R. Parker, D. B. Larremore, Rethinking COVID-19 test sensitivity\u2014a strat-
\negy for containment. N. Engl. J. Med. 383, e120 (2020).<\/p>\n
6. N. Sethuraman, S. S. Jeremiah, A. Ryo, Interpreting diagnostic tests for SARS-CoV-2.
\nJAMA 323, 2249\u20132251 (2020).
\n7. J.-R. Yang et al., Persistent viral RNA positivity during the recovery period of a patient
\nwith SARS-CoV-2 infection. J. Med. Virol. 92, 1681\u20131683 (2020).
\n8. J. An et al., Clinical characteristics of recovered COVID-19 patients with re-detectable
\npositive RNA test. Ann. Transl. Med. 8, 1084 (2020).
\n9. D. Chen et al., Recurrence of positive SARS-CoV-2 RNA in COVID-19: A case report. Int.
\nJ. Infect. Dis. 93, 297\u2013299 (2020).
\n10. L. Lan et al., Positive RT-PCR test results in patients recovered from COVID-19. JAMA
\n323, 1502\u20131503 (2020).
\n11. D. Loconsole et al., Recurrence of COVID-19 after recovery: A case report from Italy.
\nInfection 48, 965\u2013967 (2020).<\/p>\n
12. J. Lu et al., Clinical, immunological and virological characterization of COVID-19 pa-
\ntients that test re-positive for SARS-CoV-2 by RT-PCR. EBioMedicine 59, 102960 (2020).<\/p>\n
13. S. Luo, Y. Guo, X. Zhang, H. Xu, A follow-up study of recovered patients with COVID-
\n19 in Wuhan, China. Int. J. Infect. Dis. 99, 408\u2013409 (2020).<\/p>\n
14. G. Ye et al., Clinical characteristics of severe acute respiratory syndrome coronavirus 2
\nreactivation. J. Infect. 80, e14\u2013e17 (2020).
\n15. R. W\u00f6lfel et al., Virological assessment of hospitalized patients with COVID-2019.
\nNature 581, 465\u2013469 (2020).
\n16. M. Cevik et al., SARS-CoV-2, SARS-CoV, and MERS-CoV viral load dynamics, duration
\nof viral shedding, and infectiousness: A systematic review and meta-analysis. Lancet
\nMicrobe 2, e13\u2013e22 (2021).
\n17. A. L. Rasmussen, S. V. Popescu, SARS-CoV-2 transmission without symptoms. Science
\n371, 1206\u20131207 (2021).
\n18. K. K. To et al., COVID-19 re-infection by a phylogenetically distinct SARS-coronavirus-2
\nstrain confirmed by whole genome sequencing. Clin. Infect. Dis., 10.1093\/cid\/ciaa1275
\n(2020).
\n19. J. Huang et al., Recurrence of SARS-CoV-2 PCR positivity in COVID-19 patients: A
\nsingle center experience and potential implications. medRxiv [Preprint] (2020). https:\/\/
\ndoi.org\/10.1101\/2020.05.06.20089573 (Accessed 6 June 2020).
\n20. B. Yuan et al., Recurrence of positive SARS-CoV-2 viral RNA in recovered COVID-19
\npatients during medical isolation observation. Sci. Rep. 10, 11887 (2020).<\/p>\n
21. P. V\u2019Kovski, A. Kratzel, S. Steiner, H. Stalder, V. Thiel, Coronavirus biology and rep-
\nlication: Implications for SARS-CoV-2. Nat. Rev. Microbiol. 19, 155\u2013170 (2021).<\/p>\n
22. L. Alanagreh, F. Alzoughool, M. Atoum, The human coronavirus disease COVID-19: Its<\/p>\n
origin, characteristics, and insights into potential drugs and its mechanisms. Patho-
\ngens 9, 331 (2020).<\/p>\n
23. E. de Wit, N. van Doremalen, D. Falzarano, V. J. Munster, SARS and MERS: Recent
\ninsights into emerging coronaviruses. Nat. Rev. Microbiol. 14, 523\u2013534 (2016).<\/p>\n
24. A. R. Fehr, S. Perlman, Coronaviruses: An overview of their replication and patho-
\ngenesis. Methods Mol. Biol. 1282, 1\u201323 (2015).<\/p>\n
25. V. A. Belyi, A. J. Levine, A. M. Skalka, Unexpected inheritance: Multiple integrations
\nof ancient bornavirus and ebolavirus\/marburgvirus sequences in vertebrate genomes.
\nPLoS Pathog. 6, e1001030 (2010).<\/p>\n
26. M. Horie et al., Endogenous non-retroviral RNA virus elements in mammalian ge-
\nnomes. Nature 463, 84\u201387 (2010).<\/p>\n
27. M. Horie, K. Tomonaga, Non-retroviral fossils in vertebrate genomes. Viruses 3,
\n1836\u20131848 (2011).<\/p>\n
28. A. Shimizu et al., Characterisation of cytoplasmic DNA complementary to non-
\nretroviral RNA viruses in human cells. Sci. Rep. 4, 5074 (2014).<\/p>\n
29. M. B. Geuking et al., Recombination of retrotransposon and exogenous RNA virus
\nresults in nonretroviral cDNA integration. Science 323, 393\u2013396 (2009).
\n30. P. Klenerman, H. Hengartner, R. M. Zinkernagel, A non-retroviral RNA virus persists in
\nDNA form. Nature 390, 298\u2013301 (1997).
\n31. M. H. Lee et al., Somatic APP gene recombination in Alzheimer\u2019s disease and normal
\nneurons. Nature 563, 639\u2013645 (2018).
\n32. C. R. Huang, K. H. Burns, J. D. Boeke, Active transposition in genomes. Annu. Rev.
\nGenet. 46, 651\u2013675 (2012).
\n33. H. H. Kazazian Jr, J. V. Moran, Mobile DNA in health and disease. N. Engl. J. Med. 377,
\n361\u2013370 (2017).
\n34. J. M. Coffin, H. Fan, The discovery of reverse transcriptase. Annu. Rev. Virol. 3, 29\u201351
\n(2016).<\/p>\n
35. M. De Cecco et al., L1 drives IFN in senescent cells and promotes age-associated in-
\nflammation. Nature 566, 73\u201378 (2019).<\/p>\n
36. B. Rodriguez-Martin et al.; PCAWG Structural Variation Working Group; PCAWG
\nConsortium, Pan-cancer analysis of whole genomes identifies driver rearrangements
\npromoted by LINE-1 retrotransposition. Nat. Genet. 52, 306\u2013319 (2020).
\n37. E. C. Scott et al., A hot L1 retrotransposon evades somatic repression and initiates
\nhuman colorectal cancer. Genome Res. 26, 745\u2013755 (2016).<\/p>\n
38. R. B. Jones et al., LINE-1 retrotransposable element DNA accumulates in HIV-1-in-
\nfected cells. J. Virol. 87, 13307\u201313320 (2013).<\/p>\n
39. M. G. Macchietto, R. A. Langlois, S. S. Shen, Virus-induced transposable element ex-
\npression up-regulation in human and mouse host cells. Life Sci. Alliance 3,<\/p>\n
e201900536 (2020).
\n40. Y. Yin, X. Z. Liu, X. He, L. Q. Zhou, Exogenous coronavirus interacts with endogenous
\nretrotransposon in human cells. Front. Cell. Infect. Microbiol. 11, 609160 (2021).
\n41. H. Kaessmann, N. Vinckenbosch, M. Long, RNA-based gene duplication: Mechanistic
\nand evolutionary insights. Nat. Rev. Genet. 10, 19\u201331 (2009).<\/p>\n
42. S. Lanciano, G. Cristofari, Measuring and interpreting transposable element expres-
\nsion. Nat. Rev. Genet. 21, 721\u2013736 (2020).<\/p>\n
43. T. A. Morrish et al., DNA repair mediated by endonuclease-independent LINE-1 ret-
\nrotransposition. Nat. Genet. 31, 159\u2013165 (2002).<\/p>\n
44. J. C. Venter et al., The sequence of the human genome. Science 291, 1304\u20131351
\n(2001).<\/p>\n
Zhang et al. PNAS | 9 of 10
\nReverse-transcribed SARS-CoV-2 RNA can integrate into the genome of cultured human
\ncells and can be expressed in patient-derived tissues<\/p>\n
https:\/\/doi.org\/10.1073\/pnas.2105968118
\nMEDICAL SCIENCES<\/p>\n
Downloaded at University of Patras on May 19, 2021<\/p>\n
45. T. Sultana et al., The landscape of L1 retrotransposons in the human genome is
\nshaped by pre-insertion sequence biases and post-insertion selection. Mol. Cell 74,
\n555\u2013570.e7 (2019).<\/p>\n
46. D. A. Flasch et al., Genome-wide de novo L1 retrotransposition connects endonu-
\nclease activity with replication. Cell 177, 837\u2013851.e28 (2019).<\/p>\n
47. D. L. Stern, Tagmentation-based mapping (TagMap) of mobile DNA genomic inser-
\ntion sites. bioRxiv [Preprint] (2017). https:\/\/doi.org\/10.1101\/037762 (Accessed 16 Feb-
\nruary 2021).<\/p>\n
48. S. Picelli et al., Tn5 transposase and tagmentation procedures for massively scaled
\nsequencing projects. Genome Res. 24, 2033\u20132040 (2014).
\n49. L. Zhang et al., SARS-CoV-2 RNA reverse-transcribed and integrated into the human<\/p>\n
genome. bioRxiv [Preprint] (2020). https:\/\/doi.org\/10.1101\/2020.12.12.422516 (Ac-
\ncessed 16 March 2021).<\/p>\n
50. D. Blanco-Melo et al., Imbalanced host response to SARS-CoV-2 drives development of
\nCOVID-19. Cell 181, 1036\u20131045.e9 (2020).
\n51. J. Huang et al., SARS-CoV-2 infection of pluripotent stem cell-derived human lung
\nalveolar type 2 cells elicits a rapid epithelial-intrinsic inflammatory response. Cell Stem
\nCell 27, 962\u2013973.e7 (2020).
\n52. J. A. Perez-Bermejo et al., SARS-CoV-2 infection of human iPSC-derived cardiac cells
\nreflects cytopathic features in hearts of patients with COVID-19. Sci. Transl. Med.,
\n10.1126\/scitranslmed.abf7872 (2021).
\n53. F. Jacob et al., Human pluripotent stem cell-derived neural cells and brain organoids
\nreveal SARS-CoV-2 neurotropism predominates in choroid plexus epithelium. Cell
\nStem Cell 27, 937\u2013950.e9 (2020).
\n54. G. G. Giobbe et al., SARS-CoV-2 infection and replication in human fetal and pediatric
\ngastric organoids. bioRxiv [Preprint] (2020). https:\/\/doi.org\/10.1101\/2020.06.24.167049
\n(Accessed 28 October 2020).
\n55. S. E. Gill et al., Transcriptional profiling of leukocytes in critically ill COVID19 patients:
\nImplications for interferon response and coagulation. Intensive Care Med. Exp. 8, 75
\n(2020).
\n56. D. Kim et al., The architecture of SARS-CoV-2 transcriptome. Cell 181, 914\u2013921.e10
\n(2020).
\n57. B. Yan et al., Host-virus chimeric events in SARS-CoV2 infected cells are infrequent
\nand artifactual. bioRxiv [Preprint] (2021). https:\/\/doi.org\/10.1101\/2021.02.17.431704
\n(Accessed 20 February 2021).
\n58. Y. Han et al., Identification of candidate COVID-19 therapeutics using hPSC-derived
\nlung organoids. bioRxiv [Preprint] (2020). https:\/\/doi.org\/10.1101\/2020.05.05.079095
\n(Accessed 10 March 2021).<\/p>\n
59. S. Alexandersen, A. Chamings, T. R. Bhatta, SARS-CoV-2 genomic and subgenomic
\nRNAs in diagnostic samples are not an indicator of active replication. Nat. Commun.
\n11, 6059 (2020).
\n60. N. Desai et al., Temporal and spatial heterogeneity of host response to SARS-CoV-2
\npulmonary infection. Nat. Commun. 11, 6319 (2020).
\n61. M. Liao et al., Single-cell landscape of bronchoalveolar immune cells in patients with
\nCOVID-19. Nat. Med. 26, 842\u2013844 (2020).
\n62. M. C. Dalakas, Guillain-Barr\u00e9 syndrome: The first documented COVID-19-triggered
\nautoimmune neurologic disease: More to come with myositis in the offing. Neurol.
\nNeuroimmunol. Neuroinflamm. 7, e781 (2020).<\/p>\n
63. S. Pfeuffer et al., Autoimmunity complicating SARS-CoV-2 infection in selective IgA-
\ndeficiency. Neurol. Neuroimmunol. Neuroinflamm. 7, e881 (2020).<\/p>\n
64. A. Baum et al., REGN-COV2 antibodies prevent and treat SARS-CoV-2 infection in
\nrhesus macaques and hamsters. Science 370, 1110\u20131115 (2020).
\n65. B. J. Wagstaff, M. Barnerssoi, A. M. Roy-Engel, Evolutionary conservation of the
\nfunctional modularity of primate and murine LINE-1 elements. PLoS One 6, e19672
\n(2011).
\n66. E. A. Farkash, G. D. Kao, S. R. Horman, E. T. Prak, Gamma radiation increases
\nendonuclease-dependent L1 retrotransposition in a cultured cell assay. Nucleic Acids
\nRes. 34, 1196\u20131204 (2006).
\n67. WHO, World Health Organization (WHO) resource of in-house\u2013developed molecular
\nassays. https:\/\/www.who.int\/docs\/default-source\/coronaviruse\/whoinhouseassays.pdf?
\nsfvrsn=de3a76aa_2. Accessed 6 June 2020.
\n68. H. Li, Minimap2: Pairwise alignment for nucleotide sequences. Bioinformatics 34,
\n3094\u20133100 (2018).
\n69. W. J. Kent, BLAT\u2013the BLAST-like alignment tool. Genome Res. 12, 656\u2013664 (2002).
\n70. A. Dobin et al., STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 29, 15\u201321
\n(2013).
\n71. H. Li et al.; 1000 Genome Project Data Processing Subgroup, The sequence alignment\/
\nmap format and SAMtools. Bioinformatics 25, 2078\u20132079 (2009).
\n72. F. Ram\u00edrez et al., deepTools2: A next generation web server for deep-sequencing data
\nanalysis. Nucleic Acids Res. 44, W160\u2013W165 (2016).
\n73. A. R. Quinlan, I. M. Hall, BEDTools: A flexible suite of utilities for comparing genomic
\nfeatures. Bioinformatics 26, 841\u2013842 (2010).
\n74. G. X. Zheng et al., Massively parallel digital transcriptional profiling of single cells.
\nNat. Commun. 8, 14049 (2017).
\n75. T. Stuart et al., Comprehensive integration of single-cell data Cell 177,1888\u20131902.e21 (2019).<\/p>\n","protected":false},"excerpt":{"rendered":"
Liguo Zhanga, Alexsia Richardsa, M. Inmaculada Barrasaa\ue840, Stephen H. Hughesb\ue840, Richard A. Younga,c, and Rudolf Jaenischa,c,1 a Whitehead Institute for Biomedical Research, Cambridge, MA 02142; b HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, MD 21702; and c Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02142 Contributed by … <\/p>\n