Visualizing the transcription and replication of influenza A viral RNAs in cells by multiple direct RNA padlock probing and in situ sequencing (mudRapp-seq)

Abstract

Influenza A viruses (IAVs) contain eight negative-sense single-stranded viral RNA (vRNA) molecules, which are transcribed into messenger RNA (mRNA) and replicated via complementary RNA (cRNA). These processes are tightly regulated, but the precise molecular mechanisms governing the switch from transcription to replication remain elusive. Here, we introduce multiple direct RNA-assisted padlock probing in combination with in situ sequencing (mudRapp-seq) to visualize the transcription and replication of all eight IAV vRNA and mRNA molecules at the single-cell level. We demonstrate that direct RNA padlock probing is three times more efficient than conventional probes that target cDNA. Individual probes showed variations in efficiency, partly due to the RNA structure of the target, which was mitigated by employing multiple padlock probes per target. Applying mudRapp-seq to an infection time course, we observed early mRNA expression, followed by vRNA accumulation ∼3 h later. Individual viral segments exhibited differential expression, particularly in the mRNA population. Both bulk and single-cell analyses revealed a correlation between the expression of "M" mRNA and the onset of the transcription-to-replication switch. Our findings demonstrate that mudRapp-seq offers significant potential for elucidating viral replication mechanisms and may be applicable to studying other RNA viruses and cellular RNA processes.

Graphical Abstract

Figure 1.

Direct RNA padlock probing enables highly sensitive and strain-specific detection of IAV genomes. (A) Schematic comparison of classical cDNA padlock probing and direct RNA padlock probing. After hybridization to the targets (cDNA or RNA), the PLPs are ligated, followed by RCA to generate concatemers. The loop region of the PLP contains the detection probe sequence, and after RCA, the RCPs are hybridized with a detection probe and detected as spots. (B) Sensitivity comparison between classical padlock probing and direct RNA padlock probing. Madin–Darby canine kidney (MDCK) cells were infected with 1 multiplicity of infection (MOI) and samples were probed for the PB1 segment (either direct RNA or cDNA) after 6 hpi (scale bar = 50 μm). (C) Graph showing the average RCPs detected per infected cell in case of direct RNA probing and cDNA probing. Y-axis is limited to the range 0–100; one FOV (field of view) with 196 RCPs per infected cell in 5PLP vRNA is not shown. cDNA* is sample padlock probed with an additional primer to prime the RCA reaction. (D) Specificity assessment of the direct RNA padlock probing strategy using two closely related H1N1 IAV strains, i.e. PR8 (in green) and St. Petersburg (StPt) (in magenta). Samples were infected with PR8 only, PR8 and StPt (co-infection), or StPt only, and all the samples were padlock probed for both strains’ probe set. (scale bar = 50 μm). (E) Quantification demonstrating the specificity of the direct RNA padlock probing strategy. RCP spots were detected for both PLP sets only in the case of co-infection, confirming strain-specific detection.

Figure 2.

Multiple-direct RNA padlock probing improves the detection efficiency of the RNA targets. (A) Schematic representation of PLP binding sites on the NA segment of IAV strain PR8 vRNA. (B) Images demonstrating heterogeneity in PLP binding across different regions of the NA segment (scale bar = 50 μm). (C) Binding efficiency of individual PLPs for the NA segment. (D) Quantification of unspecific binding of the individual PLP. Negative control was performed on non-infected MDCK cells. (E) Probing PR8-infected MDCK cells (1 MOI) after 6 hpi with an increasing number of PLPs. (F) Graph showing the increase in target detection efficiency as a function of the number of PLPs per target. The line represents smoothed data obtained through local polynomial regression fitting. (G) Negative control with all 10 NA PLPs on uninfected cells.

Figure 3.

DMS probing reveals the impact of RNA structure on PLP binding efficiency. (A) Schematic illustration of DMS probing principle: single-stranded RNA regions are reactive to DMS, base-paired RNA and the region hybridized with PLP will be totally protected from DMS modification. (B) DMS reactivity profile with and without PLP binding along the NA vRNA segment. Individual PLPs are displayed as blue arrows, colored by binding efficiency as measured by RCP/cell. Illustration generated with genomes. (C) Correlation analysis of DMS reactivity at target sites with and without PLP binding. The scatter plot shows the relationship between DMS reactivity in both conditions, with the difference highlighting structural changes upon probe binding. (D) Heat map plot of the high-binding region (PLP1) and low-binding region (PLP9) on the NA segment. PLP9 binding site has high DMS reactivity at the ligation junction.

Figure 4.

mudRapp-seq workflow for in situ RNA detection and quantification. (A) The loop region of the PLP can be customized with a target-specific nucleotide barcode and by hybridizing sequencing primers to the RCPs and employing Illumina four-color chemistry for in situ sequencing. (B) These barcodes can be detected and quantified via sequencing. After sequencing primer hybridization, incorporation mix is added and the sample is imaged, followed by cleavage of the dye attached to nucleotide to prepare the sample for the next round of incorporation. After multiple sequencing cycles, maximum intensity projection is performed on z-stack images, followed by image alignment and spot registration. The fluorescent spots are then basecalled for each sequencing cycle to determine the barcodes. Finally, the spatial localization of detected targets is registered, enabling further quantifications.

Figure 5.

Simultaneous detection of the vRNA and viral mRNA in infected cells. (A) Barcode design on the PLP for simultaneous detection of the vRNA and mRNA in cells. The first round of incorporation distinguishes between the vRNA and mRNA molecules in infected cells, and in combination with the second round of incorporation specific segments can be identified. (B) Representative image of PR8-infected MDCK cells with 0.3 MOI and 1 MOI after the first round of incorporation where the vRNA and mRNA can be distinguished (scale bar = 50 μm). (C) Graph showing the bulk transcription and replication pattern of PR8 infected with 0.3 and 1 MOI for 0–8 hpi. Points are individual FOVs, and lines are locally estimated scatter plot smoothing (loess) smoothed conditional means. (D) Distribution of the single vRNA and mRNA segments at 0–8 h for 0.3 and 1 MOI infected cells. Points are individual FOVs, and lines are loess smoothed conditional means. y-axis is limited to range 0–20, and some points exceed this range and are therefore not shown; all points are included in calculating the smoothing.

Figure 6.

Single-cell analysis of transcription and replication heterogeneity during early infection. (A) Diagram of variations in transcription and replication at single-cell level at early time points of infection. We fixed the cells every hour till 8 h post-infection to capture the variation in transcription and replication every hour at the single-cell level. For single-cell visualization of the RCPs after in situ sequencing, cell and nucleus segmentation is performed. (B) Cumulative distribution of total vRNA and mRNA at each time point of infection. The number of cells included is on the x-axis and the total number of RCP spots in these cells on the y-axis; cells are sorted descending by number of RCPs. (C) Boxplot illustrating the levels of total vRNA on a log scale in cells categorized as low, medium, or high expressers for individual mRNA segments. Abundance of total vRNA (D) and mRNA (E) in cells at hpi >5 missing exactly 1 vRNA segment. x-axis shows the missing segment, and the y-axis the total sum of vRNA and mRNA in cells. Statistical test was Wilcoxon rank sum test with continuity correction.