Structure and sequence at an RNA template 5′ end influence insertion of transgenes by an R2 retrotransposon protein

Native R2 Rz sequences confer different efficiencies of RNA template function

R2 retrotransposons are grouped into four clades that differ in the number and type of N-terminal zinc finger domains (ZFs) preceding the DNA-binding Myb domain (Kojima et al. 2016). We developed PRINT using A-clade avian R2 proteins containing 3 N-terminal ZFs, with particular success for an R2p from the sparrow Zonotrichia albicollis (ZoAl R2p) (Zhang et al. 2024). In our prior work, the RNA templates paired with ZoAl R2p mRNA for codelivery to cells harbored a compact, relatively minimal length Rz 5′ module adapted from a B-clade R2 retrotransposon in the flour beetle Tribolium castaneum (initially annotated as R2-A, here TriCasA) (Zhang et al. 2024). To evaluate other R2 5′ Rzs as PRINT RNA template 5′ modules, we used sequences from R2 retrotransposons of different species with different predicted lengths of retained rRNA (Fig. 1C). If necessary, we adjusted the rRNA sequence retained within a cleaved Rz to match human rRNA consensus, and in some cases, the peripheral Rz P4 was stabilized with a capping stem–loop as done in structural studies (Ferré-D'Amaré et al. 1998). In vitro transcription (IVT) by T7 RNAP was used to synthesize two RNAs (Fig. 1D): one with a leader sequence and the Rz only, for testing self-cleavage, and a longer PRINT RNA template extended by the transgene expression cassette, a 3′ module for ZoAl R2p binding, and a 3′ tail with 4 nt of downstream rRNA (R4) followed by a 22-nt polyadenosine tract (A22) to promote TPRT.

We compared confirmed or predicted R2 Rzs that cleave at positions 0, −13, or −28 in upstream rRNA (Fig. 1C). TriCasA Rz previously used for PRINT RNA templates self-cleaves at position −28 (Eickbush et al. 2013). The Rz from the active D-clade DroSi R2 self-cleaves at the retrotransposon boundary with rDNA, indicated as position 0, leaving no upstream rRNA (Eickbush et al. 2013). The Rz from the A-clade R2 of the flour beetle T. castaneum (initially annotated R2-B, here TriCasB) self-cleaves at position −13 (Eickbush et al. 2013). We also tested the potential 5′ Rz from the A-clade ZoAl R2 that is species-matched to ZoAl R2p, predicted to cleave at position −28. RNAs synthesized to test self-cleavage extended no more than 50 nt beyond the final P2 strand of the predicted Rz (see Supplemental Fig. S1 for Rz schematics of putative secondary structure base-pairings). If Rzs are active for self-cleavage, they release a transcript 5′ leader (Fig. 1D) that is the hairpin (hp) stem–loop binding site for Pseudomonas phage PP7 coat protein (Lim and Peabody 2002). To evaluate self-cleavage, we resolved IVT RNA lengths by denaturing PAGE and SYBR Gold staining. All of the native-derived Rzs self-cleaved to precisely release the transcript leader with 66%–95% efficiency (Fig. 1E) under the buffer conditions of IVT RNA synthesis.

We next tested each Rz as the 5′ module of an RNA template for transgene insertion to rDNA. The RNA template transgene cassette encoded the chicken β-actin-hybrid intron (CBh) promoter, GFP ORF, and simian virus 40 (SV40) PA signal, enabling comparison of transgene insertion by flow cytometry read-out. Human telomerase reverse transcriptase (hTERT)-immortalized retinal pigment epithelial (RPE-1) cells were used for cotransfection of RNA template and ZoAl R2p mRNA (Fig. 1F). Relative transgene insertion efficiency was scored as the percentage of cells expressing GFP (GFP%) 1-day post-transfection (Fig. 1G shows representative flow cytometry data). We also compared the median GFP intensity of GFP-positive cells, which correlates with the transgene copy number inserted into the multicopy rDNA locus (Zhang et al. 2024). As negative controls, we transfected either ZoAl R2p mRNA or GFP RNA template alone. A small background of cells gated as low-intensity GFP-positive following transfection with RNA template alone (Fig. 1G), so these values were subtracted from the GFP% measured for the parallel transfection of RNA template and R2p mRNA. RNA templates with each tested 5′ module Rz supported GFP transgene insertion, resulting in GFP expression in 2%–13% of cells quantified using assays performed in triplicate (Fig. 1H).

A comparison of GFP% yielded by different R2 HDV-like Rz 5′ module RNA templates suggests several conclusions. First, the highest GFP% (∼13%) was observed with RNA templates harboring the Rzs from TriCasA used in our previous work, and ZoAl, both cleaving at rRNA position −28 (Fig. 1H; Supplemental Fig. S1). The lowest GFP% (∼2%) was observed for the RNA template with TriCasB Rz cleaving at rRNA position −13. Despite the lack of upstream rDNA homology in the cDNA produced from an RNA template with DroSi Rz 5′ module (Fig. 1C), an intermediate GFP% (∼8%–9%) was detected (Fig. 1H; Supplemental Fig. S1). Any RNA template molecules with uncleaved DroSi Rz, like RNA template with self-cleaved DroSi Rz, would lack upstream rDNA homology. We conclude that there is not an obvious relationship between an RNA template's transgene insertion efficiency and its length of 5′ rRNA.

Second, the detection of the junctions formed between rDNA and transgene, which were amplified by PCR from gDNA, gave insight about template 5′ module influence on the mechanism of 5′ junction formation. The presence of PCR products confirms insertion at the 28S rDNA locus after cotransfection of both ZoAl R2p mRNA and GFP RNA template but not either alone (Fig. 1I). No size heterogeneity was detected at the transgene 3′ junction (Fig. 1I, top), as expected from consistent TPRT initiation at the expected rDNA insertion site, congruent with previous PRINT results (Zhang et al. 2024). To detect transgene 5′ junctions, a shared primer was positioned in upstream rDNA, and different reverse primers were used for each different 5′ module. For PRINT using RNA templates with 5′ rRNA, the size of 5′ junction PCR products suggests that annealing occurred between the cDNA 3′ end and upstream rDNA (Fig. 1J, Anneal), as confirmed by sequencing (see below). Accordingly, PCR products were predominantly of homogeneous length using RNA templates with a ZoAl or TriCasA Rz 5′ module (Fig. 1I, bottom, lanes 3 and 4). In contrast, with DroSi Rz as the 5′ module, 5′ junction PCR product sizes were heterogeneous (Fig. 1I, bottom, lane 1) and their size range suggests that cDNA was end-joined to rDNA (Fig. 1J, Join). Minor transgene 5′ junction PCR products from the use of RNA templates with ZoAl or TriCasA Rz as the 5′ module were also consistent with end-joining rather than annealing (Fig. 1I, bottom, products larger than indicated by the filled circle). The 5′ junction PCR products from the use of TriCasB Rz as the RNA template 5′ module had roughly equal distribution between the length indicative of 5′ junction formation by cDNA 3′ end annealing and heterogeneous lengths consistent with end-joining (Fig. 1I, bottom, lane 2). Overall, the cross-species 5′ Rz sampling above reinforces the conclusion that efficient transgene insertion by an A-clade R2p does not require a species-matched Rz as the RNA template 5′ module. Our observations also indicate that RNA template 5′ rRNA has a direct impact on the mechanism underlying 5′ junction formation but is not strictly related to the efficiency of productive transgene insertion.

HDV-like Rz structure contributes to RNA template 5′ module influence on transgene insertion efficiency

Because species-diverse R2 Rz sequences can function as a PRINT RNA template 5′ module, we next tested whether 5′ modules comprised of nonnative, designed HDV-like Rzs would support efficient transgene insertion. We designed Rzs with a P1 stem corresponding to the −28 self-cleavage position (designated Rz-28) and different lengths of rRNA sequence from −28 to 0 in the J1/2 region (Fig. 2A). Each Rz-28 version was active for self-cleavage (Fig. 2B). RNA templates with a 5′ Rz-28 module containing any length of rRNA from 22 to 32 nt, indicated as Rz-28(22) to Rz-28(32), supported comparable transgene insertion efficiency, evident in similar GFP% (Fig. 2C). Rz-28(18) and Rz-28(14) gave sequentially decreasing GFP% and lower GFP median intensity in GFP-positive cells, both indicative of fewer transgene insertions (Fig. 2C). All Rz-28 5′ module RNA templates yielded the expected precise transgene 3′ junctions and predominately precise transgene 5′ junctions (Fig. 2D).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Designed HDV-like Rzs confer efficient RNA template 5′ module function. (A) rRNA lengths of designed Rz-28 HDV-like Rzs, aligned to upstream rDNA sequence; −28 indicates the cleavage position in the rRNA sequence, and values in parentheses indicate the number of nucleotides of rRNA sequence included. (B) Analysis of Rz-28 cleavage from L8 leader. Reaction products were separated by 12% Urea-PAGE. (C) Quantification of insertion efficiency for RNA templates with the indicated Rz-28 5′ modules as determined by flow cytometry. Values indicate mean ± SD, n = 3. (D) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in C.

To further investigate the role of Rz-28 in determining the efficiency of RNA template 5′ module function, we mutated various features of the Rz and its 5′-flanking leader sequence. To probe the influence of Rz structure, we disrupted or replaced features of the Rz fold (Fig. 3A). Leader sequences could impact self-cleavage, for example by the formation of alternate stable stems that impede folding of the cleavage-competent structure (Chadalavada et al. 2000; Ruminski et al. 2011). Therefore, we compared the 5′ leader PP7 hp with a leader sequence that is the 8 nt of rRNA upstream in the target site from the −28 cleavage position (L8), which were both effectively released by Rz self-cleavage in vitro (Fig. 3B). We also tested versions of Rz-28(28) that retain potential to fold as a double-nested pseudoknot but do not undergo self-cleavage: no leader (NL) Rzs with or without substitution of the catalytic cytosine (C98A) (Fig. 3A, pink-shaded C). Additionally, we made a leader-less 5′ module with Rz fold disrupted by deleting the 3′ strand of P2 (ΔP2) (Fig. 3A, pink shading). As a final comparison, we used a 5′ module consisting of a simple stem–loop with 5′-terminal 26 nt of upstream rRNA starting from the −28 position (Fig. 3A, SL-28).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

HDV-like Rz structure contributes to RNA template 5′ module function. (A) Predicted secondary structures of RNA template 5′ modules Rz-28(28) (left) and SL-28 (right). rRNA sequence (blue) and Rz regions mutated across the 5′ modules in this figure (pink) are indicated. (B, left) Analysis of Rz-28(28) cleavage from a PP7 hp or L8 leader sequence. Reaction products were separated by 12% Urea-PAGE. (Right) Depiction of the compared Rz leader sequences described in the text. (C) Quantification of insertion efficiency of TriCasA Rz, SL-28, and Rz-28(28) 5′ modules as determined by flow cytometry. Different Rz-28(28) leader sequences and mutated variants are indicated. Values indicate mean ± SD, n = 3. (D) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in C.

PRINT GFP% with RNA templates using a 5′ module SL-28 versus an Rz, either Rz-28(28) or TriCasA Rz, indicated a clear benefit of the Rz structure (Fig. 3C, left three bars). The ΔP2 and C98A mutations were both severely inhibitory (Fig. 3C, right), whereas the Rz-28(28) NL 5′ modules gave an intermediate GFP%. All versions of the Rz-28(28) 5′ module yielded precise transgene 3′ junctions and predominately precise transgene 5′ junctions (Fig. 3D). The few insertions generated using an RNA template with the SL-28 5′ module also had predominately precise 5′ junctions expected from annealing of the cDNA 3′ end to the upstream target site (Fig. 3D, bottom). RNA templates without an active Rz would have a triphosphate 5′ end rather than the 5′ OH produced by self-cleavage, which could influence transgene insertion efficiency, for example, by inducing a RIG-I-mediated innate immune response (Pichlmair et al. 2006). We compared the SL-28 RNA template with or without phosphatase treatment and found that the presence of triphosphate did not explain the reduced efficiency of transgene insertion (Supplemental Fig. S2). Our findings best support a model in which the 5′ module Rz structure itself significantly increases the efficiency of RNA template use for transgene insertion.

Rz cleavage position in rRNA influences 5′ module function for transgene insertion

We next queried the influence of different rRNA self-cleavage positions on the RNA template 5′ module function. To better control for sequence differences between native R2 Rzs, we used the Rz-28 scaffold to generate Rzs of minimally different sequences that self-cleave at different rRNA positions (Fig. 4A; Supplemental Fig. S3). We first compared Rz-13 and Rz-28 versions, paralleling native R2 Rzs (Fig. 1C), with rRNA lengths from the self-cleavage position to the first-strand nick (13 or 28 nt) or 2 nt shorter (11 or 26 nt), matching the 2 nt offset of first- and second-strand nicking characterized for ZoAl R2p in vitro (Luan et al. 1993; Lee et al. 2024). In parallel, we tested an RNA template 5′ module composed of the native HDV genomic ribozyme (gRz), which lacks upstream rRNA sequence (Fig. 4A; Supplemental Fig. S3). Each Rz sequence supported efficient self-cleavage in vitro (Fig. 4B). PRINT RNA templates with each Rz as a 5′ module supported transgene insertions, but Rz-13 5′ modules provided much less efficient RNA template function (∼1%–4% GFP-positive cells) than Rz-28 or gRz 5′ modules (∼11%–14% GFP-positive cells) (Fig. 4C). The rRNA position of self-cleavage was a much stronger determinant of GFP% than the 2 nt difference in rRNA length. RNA templates with any of the Rz-13 and Rz-28 5′ modules yielded the expected precise transgene 3′ junctions and predominately precise transgene 5′ junctions. The RNA template with a gRz 5′ module yielded heterogeneous transgene 5′ junctions (Fig. 4D), paralleling the 5′ junction products generated using an RNA template with a DroSi Rz 5′ module (Fig. 1I).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Rz cleavage position in rRNA influences 5′ module function for transgene insertion. (A) HDV and designed HDV-like Rz 5′ modules with variable rRNA lengths and cleavage positions within rRNA. 5′ module sequences are aligned to upstream rDNA. (B) Analysis of cleavage from a PP7 hp leader by HDV and designed HDV-like Rzs with variable cleavage positions. (Left) 5′ module Rzs relevant to C and D. (Right) 5′ module Rzs relevant to E and F. Reaction products were separated by 12% Urea-PAGE. (C) Quantification of insertion efficiency of HDV gRz and designed Rzs with −13 and −28 cleavage positions as determined by flow cytometry. Values indicate mean ± SD, n = 3. (D) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in C. (E) Quantification of insertion efficiency of designed Rzs with −21, −24, −28, and −36 cleavage positions as determined by flow cytometry. Values indicate mean ± SD, n = 3. (F) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in E. (G) Depiction of predicted P1 stem base-pairing for WT HDV gRz and GC-depleted mutants (Mut1 and Mut2) with substituted base pairs boxed in red. (H) Quantification of insertion efficiency of WT and GC-depleted mutant HDV gRzs as determined by flow cytometry. Values indicate mean ± SD, n = 3. Normalized fraction of cells expressing GFP relative to WT is indicated in light green.

We expanded the panel of rRNA cleavage positions based on the HDV Rz structural requirement for a 5′ purine, which is most often a guanosine (Cerrone-Szakal et al. 2008). Designed Rz versions with −21, −24, or −36 rRNA cleavage positions (Fig. 4A) self-cleaved in vitro (Fig. 4B, right). However, PRINT RNA templates with these 5′ module Rzs showed markedly reduced transgene insertion efficiency compared to RNA templates with an Rz-28 5′ module (Fig. 4E). As expected, transgene 3′ and 5′ junctions were predominately precise (Fig. 4F). Since Rz-28 versions with a wide range of rRNA lengths support similar PRINT efficiencies (Fig. 2C), these results indicate that directly or indirectly, the Rz rRNA cleavage position greatly impacts RNA template 5′ module function. Among rRNA self-cleavage positions, Rz-28 was optimal for PRINT transgene insertions with precise 5′ junctions, whether assayed with native or designed Rzs. Additionally, we conclude that precise transgene 5′ junction formation is supported by target-site pairing with a cDNA 3′ end with as little as 11 nt of rRNA complementarity, a remarkably limited length requirement.

Differences in insertion efficiency using RNA templates with 5′ module Rzs that have different rRNA positions of self-cleavage could result from multiple factors. We noted a correlation that Rzs with a GC-rich P1 stem, such as gRz, TriCasA Rz, and versions of Rz-28, were the most effective 5′ modules for PRINT RNA templates. Comparison across HDV-like Rz P1 stems in 5′ modules of RNA templates that supported different PRINT efficiencies extended this correlation (Supplemental Fig. S4A,B). Although self-cleavage position varies across different species’ R2 HDV-like Rzs, Rz-28 is the most common (Eickbush et al. 2013), indicating that its GC-only P1 stem (Supplemental Fig. S3A) may be favorable. To investigate the influence of the GC content of the P1 stem in an Rz 5′ module independent of the rRNA sequence, we generated HDV gRz mutants with compensatory P1 base-pairing changes that decreased GC content (Fig. 4G) but still supported efficient self-cleavage (Supplemental Fig. S4C). The 5′ module gRzs with reduced P1 GC content compromised PRINT RNA template function, with the most GC-depleted variant generating ∼50% fewer GFP-positive cells that also had lower median GFP intensity (Fig. 4H). As expected, transgene insertions using RNA templates with gRZ P1 stem mutants had precise 3′ junctions and heterogeneous 5′ junctions (Supplemental Fig. S4D). These results suggest a model in which the thermodynamic stability of an HDV or HDV-like Rz fold is one determinant of an RNA template 5′ module contribution to PRINT efficiency. Because the P1 stem sequence is dependent on the position of rRNA self-cleavage, R2 Rzs that differ in the position of self-cleavage would have different levels of efficiency as the 5′ module of a PRINT RNA template.

Uracil base modification improves transgene insertion efficiency in RNA templates with suboptimal 5′ modules

We previously reported an increase in transgene insertion efficiency using RNA templates synthesized with 100% substitution of U by pseudouridine (Ψ) or N1-methyl-pseudouridine (m1Ψ), with TriCasA Rz as the RNA template 5′ module (Zhang et al. 2024). It is understood that Ψ increases the thermodynamic stability of RNA duplexes by favoring improved base-stacking and base-pairing, which could contribute to a Ψ-mediated increase in biological stability of mRNAs (Davis 1995; Karikó et al. 2008; Hudson et al. 2013; Mauger et al. 2019). U substitution with m1Ψ is also stabilizing but appears to be more effective in suppressing innate immune responses and confers less mispairing than Ψ (Mauger et al. 2019; Morais et al. 2021). In contrast, other U analogs may negatively influence RNA stability; for example, 5-methoxy-uridine (mOU) decreases the thermodynamic stability of RNA folding (Mauger et al. 2019). We hypothesized that the influence of using U analogs in an RNA template for PRINT might vary with 5′ module RNA structure.

As an initial evaluation, we compared Rz-28(28) to the simple stem–loop SL-28. Curiously, upon U replacement with 5-methyl-uridine (mU) or moU, no decrease in Rz-28(28) self-cleavage was evident (Fig. 5A, first three lanes). In contrast, upon U replacement with Ψ or m1Ψ, limited to no self-cleavage activity was detected (Fig. 5A, final two lanes). PRINT RNA templates harboring the 5′ module Rz-28(28) showed improved transgene insertion efficiency with U replacement with Ψ and to a lesser extent m1Ψ, similar insertion efficiency using mU, and severely decreased insertion efficiency using moU (Fig. 5B), paralleling results reported previously for PRINT RNA templates with TriCasA Rz as a 5′ module (Zhang et al. 2024). PRINT using an RNA template with an SL-28 5′ module showed a dramatic improvement in transgene insertion efficiency upon U replacement with Ψ, or slightly less improvement with m1Ψ (Fig. 5B). RNA templates with the Rz-28(28) versus SL-28 5′ module made using Ψ or m1Ψ supported a surprisingly similar GFP% and transgene 3′ junction PCR signal, although the Rz-28(28) RNA template generated higher median GFP intensity indicative of increased average transgene copy number (Fig. 5B,C, top). Transgene 5′ junction formation for both 5′ module RNA templates was predominantly precise (Fig. 5C, bottom). Some end-joining junction formation was detected for the Ψ- or m1Ψ-modified Rz-28(28) RNA templates (Fig. 5C, bottom), likely because these RNA templates retained non-rRNA leader sequence due to lack of Rz self-cleavage.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

Uracil base modification improves transgene insertion efficiency in RNA templates with suboptimal 5′ modules. (A) Analysis of Rz-28(28) cleavage from a PP7 hp leader when either U or the substituted U analog is used for Rz synthesis. Reaction products were separated by 12% Urea-PAGE. Here and subsequently, altered mobility of the cleaved PP7 hp with different U analogs is due to folding and not differences in molecular weight (see Materials and Methods). (B) Quantification of insertion efficiency of SL-28 (green) and Rz-28(28) (gray) RNA templates when U or indicated U analogs are used for RNA template synthesis, as determined by flow cytometry. Values indicate mean ± SD, n = 3. (C) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in B. (D) Analysis of HDV gRz cleavage from a PP7 hp leader when either U or the substituted U analog is used for Rz synthesis. Reaction products were separated by 12% Urea-PAGE. (E) Quantification of insertion efficiency of various 5′ module RNA templates synthesized with either U (green) or fully substituted with pseudouridine (gray), as determined by flow cytometry. Average fold enhancement, defined as the percentage of cells expressing GFP yielded by a 5′ module template synthesized with pseudouridine over the same template synthesized with U, is indicated below each 5′ module. Values indicate mean ± SD, n = 3. (F) PCR amplification of 3′ (top) and 5′ (bottom) rDNA-transgene junctions from gDNA in a representative replicate of the experiment in E. (G) Average transgene copy number per cell as measured within the 3′ module (blue) or 5′ promoter region (dark gray) relative to RPP30 by ddPCR. ddPCR was performed using gDNA extracted from cells PRINTed with U or pseudouridine-substituted RNA templates. Values indicate mean ± SD, n = 3. (H) Percentage of productive insertions (PI%, the copy number of promoter relative to 3′ module) as determined from ddPCR analysis in G. Fold enhancement, defined as the percentage of productive insertions yielded by a 5′ module template synthesized with pseudouridine over the same template synthesized with U, is indicated below each 5′ module. Values indicate mean ± SD, n = 3.

To extend this investigation, we created a 5′ module consisting of only 5′ rRNA (R28), with no complementary strand to form a stem–loop. As a set, we compared PRINT efficiency using RNA templates made with U or Ψ and with 5′ module R28, SL-28, gRz, Rz-13(13), Rz-28(28), or TriCasA Rz. Self-cleavage of gRz showed less compromise with any U modification (Fig. 5D), whereas the other Rzs mirrored the Rz-28(28) profile of self-cleavage inhibition by Ψ or m1Ψ (data not shown). RNA templates with the SL-28 5′ module yielded slightly higher PRINT efficiency than RNA templates with the R28 5′ module, both showing ∼40-fold increase in GFP% using Ψ versus U (Fig. 5E). The Rz-13 5′ module also supported improved transgene insertion efficiency using Ψ, with an eightfold increase in GFP%. The other 5′ module Rzs (gRz, Rz-28, and TriCasA Rz), showed a similar approximately threefold improvement using Ψ (Fig. 5E), despite the distinction that only gRZ retained efficient self-cleavage with U replacement by Ψ (Fig. 5A,D). As expected, transgene 3′ junction signal intensity paralleled GFP% (Fig. 5F, top).

Independent of synthesis with U or Ψ, most RNA templates with 5′ modules harboring rRNA generated precise transgene 5′ junctions (Fig. 5F, bottom), in contrast with the gRz 5′ module (lanes 7–8). However, of interest, with the 5′ module Rz-13, Ψ substitution entirely changed the preferred mode of transgene 5′ junction formation from annealing to end-joining, evident in longer and more heterogeneous 5′ junction PCR products (Fig. 5F, lanes 9 and 10). We propose that this altered 5′ junction formation reflects a challenge to annealing of the cDNA 3′ end with upstream rDNA when the RNA template 5′ leader sequence is retained due to Ψ substitution, which is sufficient to eliminate precise junction formation when the RNA template 5′ module contains only 13 nt of rRNA. However, retention of the RNA template 5′ leader sequence did not inhibit precise transgene 5′ junction formation to the same degree if a longer length of rRNA was present, as in RNA templates with Rz-28(28) or TriCasA Rz 5′ module. Low levels of 5′ junction end-joining PCR products were detected for RNA templates with these modules using Ψ for RNA synthesis (Fig. 5F, lanes 2 and 12).

Transgenes generated by PRINT can be 5′-truncated, reducing the number of productive insertions. Previously, we found that 5′ truncation decreased using the same RNA template with Ψ relative to U (Zhang et al. 2024). To investigate whether RNA templates with different 5′ module sequences shared a similar improvement in the percentage of intact-transgene, productive insertions (PI%) with Ψ, we used droplet digital PCR (ddPCR) to determine average transgene 5′-promoter versus 3′-end copy number relative to the reference gene RPP30 (Zhang et al. 2024) for TriCasA Rz, SL-28, and gRz 5′ module templates synthesized with either U or Ψ. For RNA templates synthesized with U, transgene 5′-promoter and 3′-end copy number were both enhanced by the presence of a 5′ Rz (Fig. 5G, compare bar 3 with bars 1 and 5; gray bars are transgene 5′ copy number and superimposed blue bars are transgene 3′ copy number). All RNA templates synthesized with Ψ gave consistently higher copy numbers (Fig. 5G, compare bars within each set of 1–2, 3–4, and 5–6). For each RNA template, we calculated the PI% as the ratio of 5′ copy number to 3′ copy number. PI% was ∼11%, 5%, or 16% using U-containing RNA templates with TriCasA Rz, SL-28, or gRz 5′ modules, respectively (Fig. 5H). Using Ψ-containing RNA templates, PI% was ∼25%, 25%, or 39% for these same 5′ modules, respectively. The fold improvement in PI% from Ψ modification was greatest for the RNA template with the SL-28 5′ module (approximately fivefold increase; Fig. 5H). The highest PI% was achieved using the RNA template with the gRz 5′ module, whether compared in U or Ψ (Fig. 5H). Overall, we conclude that Ψ improves PRINT RNA template function for all 5′ modules, but the degree of improvement is highly dependent on 5′ module sequence and structure.

Sequence analysis of transgene 5′ junctions suggests several mechanisms of their formation

Previously, we used Illumina whole-genome sequencing (WGS) of gDNA from PRINTed cells to quantify the on-target specificity of rDNA insertion, which was >99% for ZoAl R2p (Zhang et al. 2024). Here, we turned to WGS of gDNA harvested similarly from PRINTed cells 1 day after 2-RNA delivery to query similarities and differences in the sequences of transgene junctions formed using RNA templates with or without 5′ module Rz structure and with or without rRNA homology. To maximize the number of full-length transgene insertions, RNA templates were composed of only a varying 5′ module fused to a minimized 3′ module in which GeFo 3′ UTR was truncated to its terminal 98 nt, without an internal GFP expression cassette (Supplemental Fig. S5A). RNA templates were synthesized with Ψ using TriCasA Rz, SL-28, or gRz as the 5′ module. TriCasA Rz and SL-28 share rRNA homology but differ in Rz presence, whereas TriCasA Rz and gRz share Rz structure but differ in the presence of rRNA homology. Analysis of transgene 3′ junctions indicated that cells given ZoAl R2p mRNA and any RNA template gained rDNA-inserted transgenes with very few off-target insertions (Supplemental Fig. S5B), consistent with prior analysis at greater sequencing depth (Zhang et al. 2024). Transgene 5′ junctions were classified from the transgene side as either FL or 5′-truncated (Fig. 6A). Unexpectedly, transgene insertions using an RNA template with the gRz 5′ module were often slightly 5′-truncated, most commonly within 50 nt of the 5′ terminus (Fig. 6B).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 6.

Sequence analysis of transgene 5′ junctions suggests several mechanisms of their formation. (A) Categorization of insertion 5′ junctions yielded by RNA templates with TriCasA, SL-28, and gRz 5′ modules. 5′ junctions are assessed as either full-length, indicating that the full amount of the nonhomology region (in the case of annealing) or total (in the case of other categories) transgene cDNA length is detected, or truncated. Within “full-length” and “truncated,” 5′ junctions are classified according to the categories in C. (B) Quantification of transgene 5′ end-joining positions for insertions yielded by different 5′ module RNA templates. Colored lines above templates indicate 5′ module regions, with white corresponding to the L8 leader, red corresponding to each Rz, and blue corresponding to SL-28. The remaining transgene sequence is an identical 98 nt 3′ module. Black vertical dashed line indicates the end of the region of rDNA homology. Red vertical dashed line indicates the end of 5′ module. (C) Depiction of different mechanisms of 5′ junction formation deduced from WGS analysis of cells with transgene insertions yielded by cotransfection of ZoAl R2p mRNA and RNA templates with different 5′ modules. (D) Quantification of number of nucleotides of inferred microhomology between transgene cDNA and rDNA at 5′ junctions. Nucleotides of microhomology (gray bars) are compared to the expected frequency of microhomology by chance (black lines, see Materials and Methods). (E) rDNA-transgene join position frequency for insertions with no detected microhomology yielded by different 5′ module RNA templates. Join positions are determined relative to the position of first-strand nicking, defined as 0. Black, full-length insertion events. Gray, truncated insertion events.

To evaluate the transgene-flanking 5′ junction sequence, we first classified four categories of junction by general mechanism of formation (Zhang et al. 2024). RNA templates with upstream rDNA homology generated 5′ junctions in the “Anneal” category, resulting from cDNA 3′ end base-pairing with upstream target site and therefore no duplication of the rDNA sequence shared by template and target site (Fig. 6C, upper left). These junctions were the predominant type for both TriCasA and SL-28 5′ modules (Fig. 6A, green bars). The “Join” category has direct fusion of cDNA to rDNA (Fig. 6C, lower left), which was the predominant type of junction for the gRz 5′ module and for 5′-truncated transgenes produced from any RNA template (Fig. 6A, purple bars). Other types of transgene 5′ junctions were observed with low frequency (Fig. 6A,C), including “Snap-back,” in which the template-complementary cDNA 3′ end is extended by priming synthesis on a nearby DNA strand before rDNA junction formation, or “Extra template,” in which the template-complementary cDNA 3′ end is extended by copying another RNA template (Zhang et al. 2024).

Using the Join category of 5′ junctions, we investigated whether the rDNA-transgene fusion occurred with potential microhomology. Although 5′ junctions occurred with potential microhomology more often than expected by chance, many junctions formed without a prediction of even 1 bp bridging the joined ends (Fig. 6D). However, some microhomology use could be obscured by ZoAl R2p-mediated nontemplated addition of one or a few nucleotides to the cDNA or upstream target-site 3′ end before junction formation (Stage and Eickbush 2009; Eickbush et al. 2013). The possibility of nontemplated sequence addition by ZoAl R2p, and/or a cellular polymerase, is supported by the detection of one or more extra nucleotides between rDNA and transgene in a minority of Join events without microhomology (Supplemental Table S1).

Also using the Join category of 5′ junctions, we defined rDNA positions of fusion to transgene sequence as a potential indication of where second-strand nicking occurs. Joins without microhomology were optimal for this analysis, because the 3′-most rDNA position could be defined unambiguously. The most common rDNA position of fusion was directly opposite the bottom-strand nick used to initiate TPRT (Fig. 6E, position 0). This precision could reflect second-strand nicking opposite the first-strand nick and/or at upstream location(s) with subsequent DNA synthesis that fills in the recessed 3′ end. Other rDNA positions of transgene joining were more heterogeneous for transgene insertions using an RNA template with the SL-28 5′ module than the TriCasA or gRz 5′ module, but they were all predominantly within 10 bp of the first-strand nick (Fig. 6E). At transgene 5′ junctions formed with potential microhomology, we determined “maximal rDNA alignment” by assigning any potential microhomology to the rDNA and “maximal transgene alignment” by assigning any potential microhomology to the transgene (Supplemental Fig. S5C–E). For these junctions as well, rDNA positions of joining to the transgene were predominantly within 10 bp of the first-strand nick. Maximal rDNA alignment recapitulated the observations for junctions with no microhomology (compare Fig. 6E; Supplemental Fig. S5E). Overall, we conclude that transgene 5′ junction formation by end-joining mechanism(s) gives similar junction features for each tested RNA template, with the major influence of the 5′ module rDNA sequence being the shift in 5′ junction formation from end-joining to cDNA 3′-end annealing with the upstream target site.