TY - JOUR
AU1 - Vila-Sanjurjo,, Antón
AU2 - Juarez,, Diana
AU3 - Loyola,, Steev
AU4 - Torres,, Michael
AU5 - Leguia,, Mariana
AB - Abstract Minority Gene Expression Profiling (MGEP) refers to a scenario where the expression profiles of specific genes of interest are concentrated in a small cellular pool that is embedded within a larger, non-expressive pool. An example of this is the analysis of disease-related genes within sub-populations of blood or biopsied tissues. These systems are characterized by low signal-to-noise ratios that make it difficult, if not impossible, to uncover the desired signatures of pathogenesis in the absence of lengthy, and often problematic, technical manipulations. We have adapted ribosome profiling (RP) workflows from the Illumina to the Ion Proton platform and used them to analyze signatures of pathogenesis in an MGEP model system consisting of human cells eliciting <3% productive dengue infection. We find that RP is powerful enough to identify relevant responses of differentially expressed genes, even in the presence of significant noise. We discuss how to deal with sources of unwanted variation, and propose ways to further improve this powerful approach to the study of pathogenic signatures within MGEP systems. Minority gene expression profiling, ribosome profiling, next-generation sequencing, signatures of pathogenesis, dengue Ribosome profiling (RP) is a powerful tool that enables direct measurement of protein expression at the whole cell level [1, 2]. RP is based on the deep sequencing of ribosome footprints (RFs), which are discrete, protected segments of mRNA ~25–35 bp long. Because RFs are generated during nuclease digestion of polysomal extracts, they reflect ribosomal positions along all actively translated mRNAs. In aggregate form, RP data can be used to address topics ranging from mechanisms of translational control, to qualitative and quantitative modeling of differentially expressed (DE) genes (for reviews see [2, 3]). We are interested in viral and bacterial infectious diseases. For many of these there are no vaccines or therapies licensed for human use. For others, therapies may be available, but efficacy can be sub-optimal, particularly in the context of drug resistance and vaccine escape mutants. The pipeline from discovery to licensing of new drugs is slow, highlighting a need for access to the genetic signatures of pathogenesis, as these can help identify predictors of disease severity that guide health interventions, while also supporting gene discovery studies that further advance therapeutic development. Thus, the value of novel tools to measure DE genes in the context of human infectious disease at the systems biology level is clear and growing. The development of RNA-seq revolutionized the world of transcriptomics and led to its current use as a gold standard for DE gene measurements [4]. RP can be regarded as a specialized form of RNA-seq, and thus, it can also be used for DE gene analysis. In fact, the first RP report included some DE gene measurements despite these not being central to the study [5]. RP offers significant advantages over RNA-seq as a DE gene analysis tool. On the one hand, it enables a comprehensive view of how the transcriptome is translated into protein, whereas RNA-seq simply measures steady-state levels of mRNA that may or may not correlate with actual protein expression. Given that it is proteins, and not mRNAs, that carry out most cellular processes, truly accurate measurements of differentially expressed genes require consideration of factors influencing both transcription and translation. This might not be very significant in organisms like yeast, where the rates of protein synthesis are largely determined by mRNA levels [6], but in organisms where the correlation between transcription and translation does not always hold [7], using mRNA levels as proxy for protein expression can be problematic. At best, the information obtained is incomplete; at worst, it can be largely incorrect. On the other hand, RP provides a view of how ribosomes are distributed throughout the transcriptome. As a result, RP addresses important aspects of translation dynamics and regulation not addressable by RNA-seq alone, like ribosome distribution, codon bias, and ribosome pausing [1]. Finally, measurements of protein synthesis via RF occupancy appear to quantitatively reflect absolute protein synthesis [8]. Despite RP’s appeal to address pathogenesis during human disease, a number of questions regarding its application to situations encountered outside the controlled environment of a research lab remain unanswered. With one exception [9], RP has only been tested under uniform experimental conditions, where homogeneous cellular pools were used to measure DE genes. Clinical samples deviate from this ideal situation of uniformity because the DE gene profile(s) of interest are often expressed by a small cellular pool embedded within a larger, non-expressive pool. An example of this is the analysis of disease-related genes within sub-populations of blood or biopsied tissues. We have named this condition minority gene expression profiling (MGEP). The reductionist approach might tempt us to circumvent MGEP scenarios by separation and isolation of the sub-populations of interest, but this is not advisable for two main reasons. First, gene expression is highly sensitive to its environment. When cells are stressed, or if lengthy protocols are employed to separate and isolate specific sub-populations, the observed changes in gene expression may reflect the stress and manipulation the cells were subjected to, rather than the actual signatures of pathogenesis. Second, ribosome-mRNA complexes are well-known for their lability [10], necessitating both careful consideration of the cellular environment and expert technical handling. In these cases, MGEP may be the only way to obtain the true genetic signatures of pathogenesis. To assess the robustness of RP as a tool to detect the genetic signatures of pathogenesis in the context of MGEP, we pushed the limits of the technology by confronting it with a simulated and closely controlled MGEP scenario. Specifically, we studied human cells eliciting very low levels (<3%) of productive dengue virus (DENV) infection, and assessed if RP could detect DE genes in the context of a large uninfected cellular pool (>97%). We validated the response by comparing it to a non-MGEP scenario where 92% of the cells exhibited productive DENV-infection. Additional comparisons to paired RNA-seq data further validated our findings. Here, we show that we are able to detect meaningful DE genes in response to DENV infection in MGEP scenarios that simulate low signal-to-noise conditions encountered not only during infection, but during other conditions like symbiosis, parasitism, mutalism, etc. In the process, we have also adapted Ingolia’s original RP strategy [1] to run on the Ion Proton sequencing platform. This adaptation enables quicker turn-around times during sequencing and provides an alternative to the Illumina platform for which the original RP workflows were developed. METHODS Cell Culture and Viral Infections K-562 (ATCC CCL-243) and A-549 cells (ATCC CCL-185) were maintained in RPMI-1640 (+15% FBS) and EMEM (+10% FBS), respectively, at 37°C and 5% CO2. Viral infections were done by adding either live or heat-inactivated (HI) DENV-2 (1 hr at 55°C), at an MOI = 5, to the media of 5 × 106 cells. Non-infected cells were used as control. Following addition of the virus, both infected and non-infected cells were returned to regular growth conditions until harvest. Immunolocalizations and FACS Analysis 24 hrs post-infection, cells were washed twice with 1X PBS and fixed with BD Cytofix/Cytoperm™ (BD, 554714) according to the manufacturer’s instructions. For labeling cells were incubated (30 min at RT) in 1:20 dilutions of mouse anti-DENV-IgG (ATCC, HB-114) in 1X Perm/Wash buffer with gentle mixing. Cells were then washed 3X and incubated (30 min at RT) in 1:100 dilutions of FITC-conjugated goat-anti-mouse-IgG (Sigma, F9006) in 1X Perm/Wash buffer in the dark with gentle mixing. Cells were washed again 3X and resuspended in 1X PBS (+0.5% BSA). For immunolocalizations cells were counter stained (15 min at RT) with 1:1000 dilutions of Hoechst (Thermo-Fisher H1399). Epifluorescent images were recorded with an Axioplan microscope (Zeiss) and analyzed with Metamorph (Molecular Devices). FACS analysis was performed using an LSRFortessa Cell Analyzer (BD Biosciences) with 50K events per condition. Cell lysates for RP and RNA-seq Analysis Prior to harvest, cycloheximide (Sigma-Aldrich C7698) (100 µg/ml final concentration) was added to the culture media for 2 min. Cultures were then quickly chilled to 4°C using an ethanol/dry ice bath. Cells were pelleted by centrifugation at 1000g (10 min at 4°C), and then washed twice with ice cold PBS. Cell pellets were resuspended in 500µl of ice cold lysis buffer (polysome buffer (20mM Tris pH7.4, 150mM NaCl, 5mM MgCl2, 1mM DTT, 100 µg/ml cycloheximide) plus 25U/ml DNase I (Ambion AM2222) and 1% Triton-X-100), kept on ice for 10 min, and triturated by passing them through a 26G needle 10 times. Lysates were clarified by centrifugation at 20 000g (10 min at 4°C). Cleared lysates were divided in two: 80% was immediately used in RP workflows; the remaining 20% was stored at -80°C and later prepared into RNA-seq libraries. Adaptation of RP workflows to the Ion Proton platform RP libraries were prepared as originally described [1] with minor modifications. The main modification was a re-design of the primers and adapters used for library prep so that sequencing could be done on the Ion Proton platform (Supplement 1). Additional minor modifications, like inclusion of an extra rRNA-depletion oligo, were implemented to improve processing (Supplement 2). Preparation of RNA-seq Libraries Total RNA was isolated using RNeasy Mini Kits (Qiagen 74104). Poly-A RNA was isolated using Oligo(dT)25 Dynabeads (Thermo, 61002) from 5 µg of total RNA (assuming recoveries of 1%) previously spiked-in with ERCC Control Mixes (Thermo, 4456740). RNA-seq libraries were prepared from recovered mRNA using Total RNA-seq Kits v2 (Thermo, 4479789) following the manufacturer’s instructions. Sequencing on the Ion Proton Platform Prior to sequencing all RP and RNA-seq libraries were quality controlled (QC) by Bioanalyzer using Agilent HS DNA chips (Agilent, 5067-4626). RP and RNA-seq libraries were diluted to 13pM and 8pM, respectively, prior to amplification by emulsion PCR using the Ion OneTouch™ enrichment system (Thermo-Fisher 4474779) and Ion PI Template OT2 200 Kit v3 (Thermo-Fisher 4488318) according to the manufacturer’s instructions. Sequencing was done on the Ion Proton (Thermo-Fisher 4474779) with Ion PIv3 200 bp kits (Thermo-Fisher 4488315). Despite being bar-coded, libraries were sequenced individually using one Ion PIv2 chip (Thermo-Fisher 4482321) per sample. Bioinformatics processing and DE analysis of RP data RP data was processed using custom pipelines developed in-house (Supplement 3). Briefly, reads were trimmed at the 5’ and 3’ ends with CutAdapt [11]; contaminating cytoplasmic (cyt) rRNAs were removed with Bowtie [1, 12]; mapping was done with TopHat [13] and Bowtie2 [14]; DE analysis was done with RUVseq [15] and edgeR [15–17]; codon bias and ribosome pausing analysis was done with riboSeqR [18]; functional enrichment analysis was done with STRING v11 [19]. Bioinformatics Processing and DE Analysis of RNA-seq Data RNA-seq data was processed as described for RP data, except that prior to TopHat mapping, merged raw reads were processed with FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/) to maintain read-lengths between 20 and 200 nt and to remove low-quality ends (quality cut-off = 10). RESULTS Validation of the MGEP Model System Our MGEP model system consisted of K-562 cells infected with DENV-2 at an MOI = 5. K-562s [20] are human myelogenous leukemia cells that have been extensively used for DENV studies because they can differentiate into monocytes, one of the primary targets of DENV during natural infection. K-562s provide an ideal MGEP model system because they naturally elicit extremely low DENV infection rates. In agreement with previous observations reporting rates around 1% [21], we measured infection rates of <3% by immunolocalization and FACS analysis (Figure 1). As a control we used A-549 cells [22], which are human lung carcinoma cells that have also been extensively used in DENV studies due to their high infection susceptibility [23–27]. Not surprisingly, A-549s exhibited infection rates >92% under identical experimental conditions, indicating that K-562s can be used as a controlled MGEP model system for the evaluation of RP as a tool for measurements of DE genes. Figure 1. Open in new tabDownload slide K-562s are an ideal MGEP model system because they exhibit very low DENV infection rates (<3%), as compared to control A-549s which exhibit very high infections rates (>92%) under identical experimental conditions. Shown are FACS analyses with 50K events/condition, and immunolocalization images with cell nuclei shown in blue (by Hoecsht staining) and internalized DENV viral particles in green (by FITC-conjugated anti-DENV antibodies). DIC images are provided as reference. Figure 1. Open in new tabDownload slide K-562s are an ideal MGEP model system because they exhibit very low DENV infection rates (<3%), as compared to control A-549s which exhibit very high infections rates (>92%) under identical experimental conditions. Shown are FACS analyses with 50K events/condition, and immunolocalization images with cell nuclei shown in blue (by Hoecsht staining) and internalized DENV viral particles in green (by FITC-conjugated anti-DENV antibodies). DIC images are provided as reference. Quality Assessment of Ion Proton Sequencing Runs The assessment of RP as a tool for the analysis of pathogenic signatures in the context of an MGEP model system depended on both the success of our adaptation of Ingolia’s original RP strategy [1] to run on the Ion Proton platform, and on the proper treatment of sources of unwanted variation normally expected of RP datasets. An additional potential source of difficulty was that we intended to run the minimum number of replicates (n = 3) required for statistical inference analysis [28] in order to challenge the technology to its limit. Thus, we performed triplicate replicates of three independent experimental conditions. The first condition consisted of K-562 cells infected with DENV at an MOI = 5 (DENV); the second of cells infected with a heat-inactivated form of DENV (DENV-HI and equivalent to control #1); and the last of uninfected cells (CTRL and equivalent to control #2). These three conditions are routinely used in DENV infection analyses. To reduce the sources of unwanted variation, every replicate contained one DENV, one DENV-HI, and one CTRL sample. These were processed in parallel, starting with cellular infection on different days and ending with library preparation. Libraries were sequenced individually on the Ion Proton, using Ion PIv2 chips (Thermo-Fisher 4482321) containing 165M wells each. Assuming a 60–70% bead deposition, as per the manufacturer’s specifications, we expected runs to generate 99-115M raw reads each. Instead, we routinely achieved much higher deposition rates (89% average with a range of 86–92%), which in turn resulted in runs of 128-133M raw reads (Table 1). Following sequencing completion, the sequencer automatically performs basic QC processes (elimination of polyclonal reads, low quality reads, etc.) that serve as preliminary surrogates of quality (Table 1). These provided estimates of total “usable” reads per run, ranging between 96-111M, that would later enter downstream bioinformatics pipelines. Table 1. Surrogates of quality for RP sequencing runs of K-562 cells. Similar tables available for A-549 cells, and for RNA-seq runs from both K-562s and A-549s (Supplementary Table 1). N° . NAME . CHIP LOADING . RAW READS . POLYCLONAL READS . LOW QUALITY READS . READ-LENGTH (bp) . TOTAL READS . 1 DENV 1R 89% 132 981 052 12% 9% 89 104 106 548 2 DENV 2R 89% 132 518 076 13% 5% 92 103 443 237 3 DENV 3R 86% 128 185 695 11% 8% 90 102 439 759 4 HI 1R 92% 137 409 027 8% 10% 89 110 904 101 5 HI 2R 90% 133 014 606 10% 6% 87 107 646 397 6 HI 3R 89% 131 783 910 10% 7% 89 106 623 063 7 CTRL 1R 86% 127 785 179 12% 12% 89 96 262 662 8 CTRL 2R 87% 129 533 779 12% 8% 89 101 843 535 9 CTRL 3R 89% 132 823 144 10% 10% 90 105 180 973 N° . NAME . CHIP LOADING . RAW READS . POLYCLONAL READS . LOW QUALITY READS . READ-LENGTH (bp) . TOTAL READS . 1 DENV 1R 89% 132 981 052 12% 9% 89 104 106 548 2 DENV 2R 89% 132 518 076 13% 5% 92 103 443 237 3 DENV 3R 86% 128 185 695 11% 8% 90 102 439 759 4 HI 1R 92% 137 409 027 8% 10% 89 110 904 101 5 HI 2R 90% 133 014 606 10% 6% 87 107 646 397 6 HI 3R 89% 131 783 910 10% 7% 89 106 623 063 7 CTRL 1R 86% 127 785 179 12% 12% 89 96 262 662 8 CTRL 2R 87% 129 533 779 12% 8% 89 101 843 535 9 CTRL 3R 89% 132 823 144 10% 10% 90 105 180 973 Open in new tab Table 1. Surrogates of quality for RP sequencing runs of K-562 cells. Similar tables available for A-549 cells, and for RNA-seq runs from both K-562s and A-549s (Supplementary Table 1). N° . NAME . CHIP LOADING . RAW READS . POLYCLONAL READS . LOW QUALITY READS . READ-LENGTH (bp) . TOTAL READS . 1 DENV 1R 89% 132 981 052 12% 9% 89 104 106 548 2 DENV 2R 89% 132 518 076 13% 5% 92 103 443 237 3 DENV 3R 86% 128 185 695 11% 8% 90 102 439 759 4 HI 1R 92% 137 409 027 8% 10% 89 110 904 101 5 HI 2R 90% 133 014 606 10% 6% 87 107 646 397 6 HI 3R 89% 131 783 910 10% 7% 89 106 623 063 7 CTRL 1R 86% 127 785 179 12% 12% 89 96 262 662 8 CTRL 2R 87% 129 533 779 12% 8% 89 101 843 535 9 CTRL 3R 89% 132 823 144 10% 10% 90 105 180 973 N° . NAME . CHIP LOADING . RAW READS . POLYCLONAL READS . LOW QUALITY READS . READ-LENGTH (bp) . TOTAL READS . 1 DENV 1R 89% 132 981 052 12% 9% 89 104 106 548 2 DENV 2R 89% 132 518 076 13% 5% 92 103 443 237 3 DENV 3R 86% 128 185 695 11% 8% 90 102 439 759 4 HI 1R 92% 137 409 027 8% 10% 89 110 904 101 5 HI 2R 90% 133 014 606 10% 6% 87 107 646 397 6 HI 3R 89% 131 783 910 10% 7% 89 106 623 063 7 CTRL 1R 86% 127 785 179 12% 12% 89 96 262 662 8 CTRL 2R 87% 129 533 779 12% 8% 89 101 843 535 9 CTRL 3R 89% 132 823 144 10% 10% 90 105 180 973 Open in new tab Quality Assessment of Ribosome Profiling Data Having validated the Ion Proton sequencing output in terms of quantity of reads, we used a custom bioinformatics pipeline (Supplement 3) assembled from publicly available software to assess RP data quality. Each RP sample was individually processed with this pipeline (Figure 2). To ensure that only high-quality reads would be used in the analysis we took a conservative approach, requiring that all RFs be properly flanked by both 5’ and 3’ adapter sequences added during library preparation. CutAdapt [11] was used to discard reads with improperly flanked RFs, and to trim 5’ and 3’ adapter sequences away from properly flanked ones. Given this conservative approach, 21–33% of the initial 98-110M reads that entered the pipeline were discarded, leaving 66-83M (67–79%) for further processing. Following trimming of 5’ and 3’ adapter sequences, Bowtie was used to filter out contaminating cytoplasmic (cyt) rRNAs [1, 12]. Our laboratory procedures include physical depletion of cyt-rRNAs known to be a major contaminant in all RP workflows [5, 29]. However, despite performing one round of rRNA depletion with a specific depletion oligo pool, 30–55% (31-56M) of the reads that originally entered the pipeline matched cyt-rRNAs and were eliminated, leaving 22–42% (22-44M) for further processing. Others have reported comparable numbers for similar workflows [29], indicating that cyt-rRNA contamination is a serious nuisance in RP studies. Figure 2. Open in new tabDownload slide Flowchart of the algorithm used for read processing. Individual steps and software are indicated in grey. Minimum and maximum yield is indicated as follows: discarded reads, black, italicized font; accepted reads, black, bold font. Inset shows the minimum and maximum final yield per replicate (note that each replicate consists of three independent samples, as explained in the text). Transcriptome mapped reads were used to assess read quality, while genome mapped reads were used for DE analysis. Figure 2. Open in new tabDownload slide Flowchart of the algorithm used for read processing. Individual steps and software are indicated in grey. Minimum and maximum yield is indicated as follows: discarded reads, black, italicized font; accepted reads, black, bold font. Inset shows the minimum and maximum final yield per replicate (note that each replicate consists of three independent samples, as explained in the text). Transcriptome mapped reads were used to assess read quality, while genome mapped reads were used for DE analysis. Next we used Tophat [13] to map processed reads (22-44M) to the human genome (GRCh37), successfully aligning 8–33% (8-35M) of the original input (Figure 2). As recommended by an Ion Proton-specific alignment protocol developed by SevenBridges (https://www.sevenbridges.com/ion-proton-rna-seq-alignment/), we also used Bowtie2 [14] to map unaligned reads (9-15M), adding an additional 2–6% (2-6M) of the original input to the pool of aligned reads. This combined mapping protocol resulted in a total of 10-41M reads matching human genomic sequences. In a separate alignment, we used Bowtie [1, 12] to map processed reads to the DENV-2 transcriptome, resulting in a total of 0.02-0.12M DENV2-mapped reads in DENV-infected cells only. As expected, the relative proportion of human and DENV reads in each replicate varied between treatment groups, with live DENV treatment being the only group with significant numbers of RFs aligning to the DENV genome. In these samples, coverage of the ~10.7 kb DENV genome ranged from 1.7 to 11 reads/DENV nucleotide (not shown). In contrast, in control samples (heat-inactivated DENV-2 or no virus control), coverage of the DENV genome was close to zero (not shown). It is worth noting that the total number of DENV reads (0.02-0.12M) in the DENV treatment group is drastically smaller than the total number of human reads (10-41M) in the same group. This is not surprising given that cellular gene expression would be expected to far outnumber viral gene expression, particularly in an extreme MGEP scenario where only 3% of the cellular pool is productively infected. Taken together, the alignment results are a preliminary positive indicator that the RP data obtained is consistent with what would be expected of an MGEP model system. We extended dataset QC through an assessment of “bona fide” RFs using RiboSeqR [18]. This package was specifically developed to analyze RP data, and because it requires transcriptome alignments, we mapped processed reads to the human GRCh38 transcriptome, successfully aligning 5–25% (5-27M) of the original input (Figure 2). This subset of RFs was then used to generate read-length distributions for various conditions, including GRCh38 transcriptome, DENV-2, and contaminating cyt-rRNAs aligned reads (Figure 3). During nuclease digestion, ribosomes provide protection from digestion, resulting in discrete RFs of an average length centered around 29–30 nts [5, 29]. As previously established [30], RF length is an important metric for distinguishing bona fide RFs from other sources of signal. A unimodal distribution with a peak at 29–30 nt and the expected size-range distribution (~25–35 bp) is only apparent from human GRCh38 and DENV-2 transcriptome mapped reads (Figure 3A and B). In contrast, the signal from contaminating cyt-rRNAs (Figure 3C) elicits a dramatically different distribution that lacks the 29–30 nt peak, and instead, is enriched in fragments smaller than 20 nt [30]. Collectively, these patterns constitute the expected distribution signatures of RP datasets, indicating that following subtraction of contaminating cyt-rRNAs, our dataset is composed of bona fide RFs. Figure 3. Open in new tabDownload slide Analysis of K-562 read-length distributions. Boxplots show the length distribution of reads aligning to the GFCh38 transcriptome (A), the DENV-2 transcriptome (B), and to a reference consisting of contaminating cyt rRNAs (C). Boxplots were generated with riboSeqR and custom scripts and show the median (internal horizontal bars), the range (whiskers), the values of the 1st and 3rd quartiles (external horizontal bars), and outliers (dots). Note that only GRCh38 and DENV-2 boxplots elicit the expected read-length distributions peaking around 29–30 nts. Note that for the sake of comparison, distributions are plotted under the same display scale (presented as % of reads in each plot) despite having vastly different numbers of reads represented in each plot (5-27M for GRCh38, .02-.12M for DENV-2, and 31-56M for cyt rRNAs aligned reads; Figure 2). Figure 3. Open in new tabDownload slide Analysis of K-562 read-length distributions. Boxplots show the length distribution of reads aligning to the GFCh38 transcriptome (A), the DENV-2 transcriptome (B), and to a reference consisting of contaminating cyt rRNAs (C). Boxplots were generated with riboSeqR and custom scripts and show the median (internal horizontal bars), the range (whiskers), the values of the 1st and 3rd quartiles (external horizontal bars), and outliers (dots). Note that only GRCh38 and DENV-2 boxplots elicit the expected read-length distributions peaking around 29–30 nts. Note that for the sake of comparison, distributions are plotted under the same display scale (presented as % of reads in each plot) despite having vastly different numbers of reads represented in each plot (5-27M for GRCh38, .02-.12M for DENV-2, and 31-56M for cyt rRNAs aligned reads; Figure 2). Having demonstrated that our datasets are enriched in bona fide RFs we performed codon bias and ribosome pausing analyses to further assess read quality (Figure 4). Codon bias refers to the preference of ribosomes to translate one reading frame (normally coding frame 1) over the other two (reading frames 2 and 3). Codon bias analyses of both GRCh38- (Figure 4A) and DENV-2-aligned RFs (Figure 4B) show almost identical per-frame distributions of RF read-lengths, with a clear preference for coding frame 1 (shown in light grey) over the other two (shown in dark grey and black). Besides the observed preference for coding frame 1, analysis of total ribosome density relative to the coding DNA sequence (CDS) position in the most abundant RFs (those 29 nts in length) also shows maximum ribosome density ~12 nt upstream of the 5’ start codon (Figure 4C). This is consistent with the positioning of initiating ribosomes just upstream of the start codon, protecting 4 codons 5’ of the ribosomal P site [5]. Ribosome density rapidly drops within the first 40–50 bases of CDSs, and thereafter is uniformly maintained at a lower density up to the translation termination region, where it end abruptly ~15 nt, or 5 codons, away from the termination signal. This is consistent with the dissociation of ribosomes from the mRNA once translation terminates [31]. Finally, a higher-resolution analysis of ribosome density shows that ribosome pausing events are consistent among replicates (Figure 4D), with distinct ribosome pausing sites (black arrows) along the DENV-2 mRNA despite different numbers of data points in each replicate. Besides the clear peak observed at position 85, which corresponds to position ~12 upstream of the 5’ start codon, additional smaller peaks of unambiguous frame preference are apparent at positions 66, 84, 94, and 103. Furthermore, despite positive correlation coefficients among replicates along the whole DENV mRNA, these become particularly high (0.99) in these regions. Taken together, codon bias and ribosome pausing analyses indicate that our dataset consists of high quality, bona fide RFs that elicit the expected signatures of traditional RP datasets where a low signal-to-noise ratio is not a concern. Figure 4. Open in new tabDownload slide Analysis of K-562 codon bias and ribosome pausing. Number of RFs with lengths from 26–30 nts in the 0, 1, and 2 reading frames mapping to the GRCh38 (A) and DENV-2 transcriptomes (B), as determined by riboSeqR [18]. Reading frames are coded as indicated in insets, with Frame 0 as the coding frame. C, Representative 29-nt RF density at the 5’ and 3’ ends of a CDS that combines all cellular mRNAs, as determined by riboSeqR. D, Position of 29-nt RFs near the start codon (underlined) of the DENV2 mRNA. The RNA sequence is displayed on the ×axis. The curves represent the abundance of reads along the displayed sequence. Replicates are color coded as indicated in insets. Arrows indicate pause sites discussed in the text. Figure 4. Open in new tabDownload slide Analysis of K-562 codon bias and ribosome pausing. Number of RFs with lengths from 26–30 nts in the 0, 1, and 2 reading frames mapping to the GRCh38 (A) and DENV-2 transcriptomes (B), as determined by riboSeqR [18]. Reading frames are coded as indicated in insets, with Frame 0 as the coding frame. C, Representative 29-nt RF density at the 5’ and 3’ ends of a CDS that combines all cellular mRNAs, as determined by riboSeqR. D, Position of 29-nt RFs near the start codon (underlined) of the DENV2 mRNA. The RNA sequence is displayed on the ×axis. The curves represent the abundance of reads along the displayed sequence. Replicates are color coded as indicated in insets. Arrows indicate pause sites discussed in the text. Analysis of Differential Gene Expression The ultimate litmus test for RP as an appropriate tool to interrogate MGEP scenarios is an ability to detect differentially expressed (DE) genes. To assess for the presence of DE genes using EdgeR [17], we first converted GHRC37-aligned reads into counts. Given that we expected a weak DE gene signal from our MGEP system, we assumed that modeling the almost unavoidable presence of batch effects [32] was key to uncovering the desired signal. An obvious source of batch effects in our dataset was the different processing dates for each replicate. Recall that each replicate consisted of one DENV-2-treated (DENV) sample, one heat inactivated (HI) sample and one control (CTRL) sample. Exploratory muti-dimensional scaling (MDS) plots confirmed the presence of batch effects (Figure 5A), showing that samples clustered together by processing date along two dimensions, principal components 1 and 2 (PC1 and PC2). Replicate 2 (in black) is clearly different from the other two (in grey), particularly along the first dimension PC1, which accounts for 89.38% of the variance observed, and is likely a result of the differences first noticed during read processing of replicates (Figure 2, inset). To address this issue we included the processing date of each replicate as a covariate into EdgeR GLMs. Initial attempts to identify DE genes in the DENV treated dataset (DENV vs. CTRL comparison), using EdgeR with batch effects, resulted in P-value histograms that lacked a spike at zero, which is indicative of near-zero P-values normally associated with truly DE genes (Figure 5B). The normalization package RUVseq is capable of estimating and removing factors of unwanted variation, such as batch effects [15]. Following RUVseq normalization, we obtained a uniform P-value histogram with a clear spike at zero, indicative of a reasonable DE analysis (Figure 5C), and after adjusting for a 5% false discovery rate, we identified a total of 471 DE genes (Figure 5D). Figure 5. Open in new tabDownload slide Analysis of K-562 differentially expressed genes. A, MDS plot shows clustering of samples relative to processing date. Samples from the same replicate (1, 2, or 3) are circled. PC1 and PC2 represent 89.38% and 2.98% of the total variance, respectively. B, Histogram of P-values after DE analysis with EdgeR. C, Histogram of P-values after DE analysis with EdgeR and RUVSeq. Inset shows the same histogram with an enlarged number of breakpoints. D, Smear plot shows the total gene tags identified, with DE tags in grey. Figure 5. Open in new tabDownload slide Analysis of K-562 differentially expressed genes. A, MDS plot shows clustering of samples relative to processing date. Samples from the same replicate (1, 2, or 3) are circled. PC1 and PC2 represent 89.38% and 2.98% of the total variance, respectively. B, Histogram of P-values after DE analysis with EdgeR. C, Histogram of P-values after DE analysis with EdgeR and RUVSeq. Inset shows the same histogram with an enlarged number of breakpoints. D, Smear plot shows the total gene tags identified, with DE tags in grey. Validation of DE Gene Analysis in the Context of MGEP as a Model System for Gene Discovery Having established that RP can be used to detect DE genes in MGEP model systems, we investigated the nature of the DE gene set to assess its validity specifically in the context of an MGEP model system of DENV infection. To maximize the number of genes analyzed, particularly given the MGEP nature of the study and its potential as a gene-discovery tool, we did not establish cut-off fold-change values for lowly expressed genes, and instead, analyzed the entire dataset considering its context. Of the 471 total DE genes identified (Supplementary Table 2), 363 were up-regulated and 108 were down-regulated in response to DENV-2 infection (Figure 6A and B). Of these, 130 (27.6%) were protein coding genes, 144 (30.6%) were pseudogenes, 92 (19.5%) were non-coding ncRNAs, and the remaining 105 (22.3%) had no annotation at the time of analysis. Detailed inspection of the 20 most relevant DE tags reveals that, as suspected, there is often a lack of correlation in the directionality of gene expression between transcription and translation (Table 2). In fact, 13 of these genes were not detected in a parallel DE gene analysis based on RNAseq data derived from the same samples (Table 2), confirming that the use of mRNA levels as proxy for protein expression can potentially be problematic. Table 2. Summary of the most interesting DE tags identified and their relationship to DENV . . Symbol . Name . Relation to dengue . Reference . by RP . by RNAseq . by RP . . ENSEMBL ID . . . . . in K-562s . in K-562s . in A-549s . Proteins ENSG00000112715 VEGFA Vascular endothelial growth factor A Predictor of dengue disease severity [33–35] Up Yes (Up) No ENSG00000132694 ARHGEF11 Rho guanine nucleotide exchange factor (GEF) 11 VEGF-interacting protein [36] Up No No ENSG00000038382 TRIO Trio Rho guanine nucleotide exchange factor VEGF-interacting protein, upregulated in patients with dengue fever and dengue hemorrhagic fever [37] Up No No ENSG00000104419 NDRG1 N-Myc downstream-regulated gene 1 Restricts Hepatitis C Virus Propagation [38] Up Yes (Up) Yes (Up) ENSG00000176171 BNIP3 BCL2 Interacting Protein 3 Up-regulated in dengue infected Human Monocyte derived Dendritic cells vs. uninfected cells [39] Up No No ENSG00000108679 LGALS3BP Galecting 3 bindign protein Up-regulated in in vitro dengue virus infections of human umbilical vein endothelial cells (HUVECs) [40] Down No No ENSG00000119922 IFIT2 Interferon-induced protein with tetratricopeptide repeats 2 Induced by DENV infection [41, 42] Up No Yes (Up) ENSG00000106538 RARRES2 Retinoic acid receptor responder 2 Influences respiratory syncitial virus (RSV) replication [43] Up No No ENSG00000137869 CYP19A1 Cytochrome P450s (CYP) Associated to dengue Hemorrhagic Fever according to HuGE Navigator Gene-Phenotype Associations [44] Up No No ENSG00000132821 VSTM2L V-set and transmembrane domain containing 2 like Perturbed expression in the presence of viruses [44] Up No No ENSG00000133138 TBC1D8B TBC1 domain family member 8B Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No No ENSG00000167996 FTH1 FTH1 ferritin heavy chain 1 Up-regulated in high-throughput quantitative proteomic analysis of dengue virus type 2 infected A549 cells [23] Up No No ENSG00000144504 ANKMY1 Ankyrin repeat and MYND domain containing 1 Perturbed expression in the presence of viruses [44] Up No No ENSG00000163046 ANKRD30BL Ankyrin repeat domain 30B-like Perturbed expression in the presence of viruses [44] Up No Yes (Up) ENSG00000186352 ANKRD37 Ankyrin repeat domain 37 Perturbed expression in the presence of viruses [44] Up Yes (Up) No ENSG00000135299 ANKRD6 Ankyrin repeat domain 6 Perturbed expression in the presence of viruses [44] Up No No ENSG00000051108 HERPUD1 Homocysteine-inducible ER protein with ubiquitin-like domain 1 Perturbed expression in the presence of viruses, including up-regulation after zika infection of human cortical neural precursors [44] Up Yes (Up) Yes (Up) ENSG00000125968 ID1 ID1 inhibitor of DNA binding 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Down Yes (Down) Yes (Down) ENSG00000115738 ID2 ID2 inhibitor of DNA binding 2 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) ENSG00000117318 ID3 ID3 inhibitor of DNA binding 3 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) lncRNAs ENSG00000251562 MALAT1 Metastasis associated lung adenocarcinoma transcript 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No Yes (Up) ENSG00000218510 LINC00339 Long intergenic non-protein coding RNA 339 subjected to transcription readthrough during herpes simplex virus 1 infection [46] Down No Yes (Down) miRNAs ENSG00000247095 MIR210HG MIR210 host gene Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up Yes (Up) No ENSG00000207567 MIR142 microRNA 144 Hematopoietic-specific miRNA attenuates genetically engineered DENV2 strains carrying MIR142 target sites [47] Up Yes (Up) No ENSG00000199004 MIR21 microRNA 21 Promotes DENV2 replication in HepG2 cells [48] Down No No ENSG00000207618 miR144 microRNA 144 Up-regulated after Crimean Congo Virus (CCV) infection of human subjects [49] Up No No . . Symbol . Name . Relation to dengue . Reference . by RP . by RNAseq . by RP . . ENSEMBL ID . . . . . in K-562s . in K-562s . in A-549s . Proteins ENSG00000112715 VEGFA Vascular endothelial growth factor A Predictor of dengue disease severity [33–35] Up Yes (Up) No ENSG00000132694 ARHGEF11 Rho guanine nucleotide exchange factor (GEF) 11 VEGF-interacting protein [36] Up No No ENSG00000038382 TRIO Trio Rho guanine nucleotide exchange factor VEGF-interacting protein, upregulated in patients with dengue fever and dengue hemorrhagic fever [37] Up No No ENSG00000104419 NDRG1 N-Myc downstream-regulated gene 1 Restricts Hepatitis C Virus Propagation [38] Up Yes (Up) Yes (Up) ENSG00000176171 BNIP3 BCL2 Interacting Protein 3 Up-regulated in dengue infected Human Monocyte derived Dendritic cells vs. uninfected cells [39] Up No No ENSG00000108679 LGALS3BP Galecting 3 bindign protein Up-regulated in in vitro dengue virus infections of human umbilical vein endothelial cells (HUVECs) [40] Down No No ENSG00000119922 IFIT2 Interferon-induced protein with tetratricopeptide repeats 2 Induced by DENV infection [41, 42] Up No Yes (Up) ENSG00000106538 RARRES2 Retinoic acid receptor responder 2 Influences respiratory syncitial virus (RSV) replication [43] Up No No ENSG00000137869 CYP19A1 Cytochrome P450s (CYP) Associated to dengue Hemorrhagic Fever according to HuGE Navigator Gene-Phenotype Associations [44] Up No No ENSG00000132821 VSTM2L V-set and transmembrane domain containing 2 like Perturbed expression in the presence of viruses [44] Up No No ENSG00000133138 TBC1D8B TBC1 domain family member 8B Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No No ENSG00000167996 FTH1 FTH1 ferritin heavy chain 1 Up-regulated in high-throughput quantitative proteomic analysis of dengue virus type 2 infected A549 cells [23] Up No No ENSG00000144504 ANKMY1 Ankyrin repeat and MYND domain containing 1 Perturbed expression in the presence of viruses [44] Up No No ENSG00000163046 ANKRD30BL Ankyrin repeat domain 30B-like Perturbed expression in the presence of viruses [44] Up No Yes (Up) ENSG00000186352 ANKRD37 Ankyrin repeat domain 37 Perturbed expression in the presence of viruses [44] Up Yes (Up) No ENSG00000135299 ANKRD6 Ankyrin repeat domain 6 Perturbed expression in the presence of viruses [44] Up No No ENSG00000051108 HERPUD1 Homocysteine-inducible ER protein with ubiquitin-like domain 1 Perturbed expression in the presence of viruses, including up-regulation after zika infection of human cortical neural precursors [44] Up Yes (Up) Yes (Up) ENSG00000125968 ID1 ID1 inhibitor of DNA binding 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Down Yes (Down) Yes (Down) ENSG00000115738 ID2 ID2 inhibitor of DNA binding 2 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) ENSG00000117318 ID3 ID3 inhibitor of DNA binding 3 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) lncRNAs ENSG00000251562 MALAT1 Metastasis associated lung adenocarcinoma transcript 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No Yes (Up) ENSG00000218510 LINC00339 Long intergenic non-protein coding RNA 339 subjected to transcription readthrough during herpes simplex virus 1 infection [46] Down No Yes (Down) miRNAs ENSG00000247095 MIR210HG MIR210 host gene Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up Yes (Up) No ENSG00000207567 MIR142 microRNA 144 Hematopoietic-specific miRNA attenuates genetically engineered DENV2 strains carrying MIR142 target sites [47] Up Yes (Up) No ENSG00000199004 MIR21 microRNA 21 Promotes DENV2 replication in HepG2 cells [48] Down No No ENSG00000207618 miR144 microRNA 144 Up-regulated after Crimean Congo Virus (CCV) infection of human subjects [49] Up No No Open in new tab Table 2. Summary of the most interesting DE tags identified and their relationship to DENV . . Symbol . Name . Relation to dengue . Reference . by RP . by RNAseq . by RP . . ENSEMBL ID . . . . . in K-562s . in K-562s . in A-549s . Proteins ENSG00000112715 VEGFA Vascular endothelial growth factor A Predictor of dengue disease severity [33–35] Up Yes (Up) No ENSG00000132694 ARHGEF11 Rho guanine nucleotide exchange factor (GEF) 11 VEGF-interacting protein [36] Up No No ENSG00000038382 TRIO Trio Rho guanine nucleotide exchange factor VEGF-interacting protein, upregulated in patients with dengue fever and dengue hemorrhagic fever [37] Up No No ENSG00000104419 NDRG1 N-Myc downstream-regulated gene 1 Restricts Hepatitis C Virus Propagation [38] Up Yes (Up) Yes (Up) ENSG00000176171 BNIP3 BCL2 Interacting Protein 3 Up-regulated in dengue infected Human Monocyte derived Dendritic cells vs. uninfected cells [39] Up No No ENSG00000108679 LGALS3BP Galecting 3 bindign protein Up-regulated in in vitro dengue virus infections of human umbilical vein endothelial cells (HUVECs) [40] Down No No ENSG00000119922 IFIT2 Interferon-induced protein with tetratricopeptide repeats 2 Induced by DENV infection [41, 42] Up No Yes (Up) ENSG00000106538 RARRES2 Retinoic acid receptor responder 2 Influences respiratory syncitial virus (RSV) replication [43] Up No No ENSG00000137869 CYP19A1 Cytochrome P450s (CYP) Associated to dengue Hemorrhagic Fever according to HuGE Navigator Gene-Phenotype Associations [44] Up No No ENSG00000132821 VSTM2L V-set and transmembrane domain containing 2 like Perturbed expression in the presence of viruses [44] Up No No ENSG00000133138 TBC1D8B TBC1 domain family member 8B Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No No ENSG00000167996 FTH1 FTH1 ferritin heavy chain 1 Up-regulated in high-throughput quantitative proteomic analysis of dengue virus type 2 infected A549 cells [23] Up No No ENSG00000144504 ANKMY1 Ankyrin repeat and MYND domain containing 1 Perturbed expression in the presence of viruses [44] Up No No ENSG00000163046 ANKRD30BL Ankyrin repeat domain 30B-like Perturbed expression in the presence of viruses [44] Up No Yes (Up) ENSG00000186352 ANKRD37 Ankyrin repeat domain 37 Perturbed expression in the presence of viruses [44] Up Yes (Up) No ENSG00000135299 ANKRD6 Ankyrin repeat domain 6 Perturbed expression in the presence of viruses [44] Up No No ENSG00000051108 HERPUD1 Homocysteine-inducible ER protein with ubiquitin-like domain 1 Perturbed expression in the presence of viruses, including up-regulation after zika infection of human cortical neural precursors [44] Up Yes (Up) Yes (Up) ENSG00000125968 ID1 ID1 inhibitor of DNA binding 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Down Yes (Down) Yes (Down) ENSG00000115738 ID2 ID2 inhibitor of DNA binding 2 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) ENSG00000117318 ID3 ID3 inhibitor of DNA binding 3 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) lncRNAs ENSG00000251562 MALAT1 Metastasis associated lung adenocarcinoma transcript 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No Yes (Up) ENSG00000218510 LINC00339 Long intergenic non-protein coding RNA 339 subjected to transcription readthrough during herpes simplex virus 1 infection [46] Down No Yes (Down) miRNAs ENSG00000247095 MIR210HG MIR210 host gene Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up Yes (Up) No ENSG00000207567 MIR142 microRNA 144 Hematopoietic-specific miRNA attenuates genetically engineered DENV2 strains carrying MIR142 target sites [47] Up Yes (Up) No ENSG00000199004 MIR21 microRNA 21 Promotes DENV2 replication in HepG2 cells [48] Down No No ENSG00000207618 miR144 microRNA 144 Up-regulated after Crimean Congo Virus (CCV) infection of human subjects [49] Up No No . . Symbol . Name . Relation to dengue . Reference . by RP . by RNAseq . by RP . . ENSEMBL ID . . . . . in K-562s . in K-562s . in A-549s . Proteins ENSG00000112715 VEGFA Vascular endothelial growth factor A Predictor of dengue disease severity [33–35] Up Yes (Up) No ENSG00000132694 ARHGEF11 Rho guanine nucleotide exchange factor (GEF) 11 VEGF-interacting protein [36] Up No No ENSG00000038382 TRIO Trio Rho guanine nucleotide exchange factor VEGF-interacting protein, upregulated in patients with dengue fever and dengue hemorrhagic fever [37] Up No No ENSG00000104419 NDRG1 N-Myc downstream-regulated gene 1 Restricts Hepatitis C Virus Propagation [38] Up Yes (Up) Yes (Up) ENSG00000176171 BNIP3 BCL2 Interacting Protein 3 Up-regulated in dengue infected Human Monocyte derived Dendritic cells vs. uninfected cells [39] Up No No ENSG00000108679 LGALS3BP Galecting 3 bindign protein Up-regulated in in vitro dengue virus infections of human umbilical vein endothelial cells (HUVECs) [40] Down No No ENSG00000119922 IFIT2 Interferon-induced protein with tetratricopeptide repeats 2 Induced by DENV infection [41, 42] Up No Yes (Up) ENSG00000106538 RARRES2 Retinoic acid receptor responder 2 Influences respiratory syncitial virus (RSV) replication [43] Up No No ENSG00000137869 CYP19A1 Cytochrome P450s (CYP) Associated to dengue Hemorrhagic Fever according to HuGE Navigator Gene-Phenotype Associations [44] Up No No ENSG00000132821 VSTM2L V-set and transmembrane domain containing 2 like Perturbed expression in the presence of viruses [44] Up No No ENSG00000133138 TBC1D8B TBC1 domain family member 8B Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No No ENSG00000167996 FTH1 FTH1 ferritin heavy chain 1 Up-regulated in high-throughput quantitative proteomic analysis of dengue virus type 2 infected A549 cells [23] Up No No ENSG00000144504 ANKMY1 Ankyrin repeat and MYND domain containing 1 Perturbed expression in the presence of viruses [44] Up No No ENSG00000163046 ANKRD30BL Ankyrin repeat domain 30B-like Perturbed expression in the presence of viruses [44] Up No Yes (Up) ENSG00000186352 ANKRD37 Ankyrin repeat domain 37 Perturbed expression in the presence of viruses [44] Up Yes (Up) No ENSG00000135299 ANKRD6 Ankyrin repeat domain 6 Perturbed expression in the presence of viruses [44] Up No No ENSG00000051108 HERPUD1 Homocysteine-inducible ER protein with ubiquitin-like domain 1 Perturbed expression in the presence of viruses, including up-regulation after zika infection of human cortical neural precursors [44] Up Yes (Up) Yes (Up) ENSG00000125968 ID1 ID1 inhibitor of DNA binding 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Down Yes (Down) Yes (Down) ENSG00000115738 ID2 ID2 inhibitor of DNA binding 2 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) ENSG00000117318 ID3 ID3 inhibitor of DNA binding 3 Perturbed expression in the presence of viruses [44] Down Yes (Down) Yes (Down) lncRNAs ENSG00000251562 MALAT1 Metastasis associated lung adenocarcinoma transcript 1 Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up No Yes (Up) ENSG00000218510 LINC00339 Long intergenic non-protein coding RNA 339 subjected to transcription readthrough during herpes simplex virus 1 infection [46] Down No Yes (Down) miRNAs ENSG00000247095 MIR210HG MIR210 host gene Up-regulated in patients with dengue shock syndrome, relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45] Up Yes (Up) No ENSG00000207567 MIR142 microRNA 144 Hematopoietic-specific miRNA attenuates genetically engineered DENV2 strains carrying MIR142 target sites [47] Up Yes (Up) No ENSG00000199004 MIR21 microRNA 21 Promotes DENV2 replication in HepG2 cells [48] Down No No ENSG00000207618 miR144 microRNA 144 Up-regulated after Crimean Congo Virus (CCV) infection of human subjects [49] Up No No Open in new tab Figure 6. Open in new tabDownload slide DE genes identified by RP. A, Distribution of total DE features identified in K-562s. B, Volcano plot of individual DE genes identified relative to their logFCs and false discovery rates (FDRs). Features below the DE threshold (FDR > 0.05) are in black. Features above the threshold are either in white or grey according to legend. C, Volcano plot of DE protein-coding genes identified in K-562s. Highly interesting genes described in the text are indicated by name and shown as large black dots (and listed in Table 2). White dots represent genes with abs(logFC) > 1 and FDR < 0.05; light grey dots, abs(logFC) > 1 and FDR > 0.05; dark grey dots, abs(logFC) < 1 and FDR < 0.05; and black dots, abs(logFC) < 1 and FDR > 0.05. D, As in (C) but displaying data from A-549s. Figure 6. Open in new tabDownload slide DE genes identified by RP. A, Distribution of total DE features identified in K-562s. B, Volcano plot of individual DE genes identified relative to their logFCs and false discovery rates (FDRs). Features below the DE threshold (FDR > 0.05) are in black. Features above the threshold are either in white or grey according to legend. C, Volcano plot of DE protein-coding genes identified in K-562s. Highly interesting genes described in the text are indicated by name and shown as large black dots (and listed in Table 2). White dots represent genes with abs(logFC) > 1 and FDR < 0.05; light grey dots, abs(logFC) > 1 and FDR > 0.05; dark grey dots, abs(logFC) < 1 and FDR < 0.05; and black dots, abs(logFC) < 1 and FDR > 0.05. D, As in (C) but displaying data from A-549s. Recently we have carried out similar RP- and RNAseq-based analyses using the A-549 cell line [22], which is frequently used as a model for DENV infection due to its high susceptibility to infection [25]. In our hands, A-549s elicit infection rates higher than 92% under identical infection conditions to the ones used for K-562s (Figure 1). Although the details of that analysis will be published elsewhere, a brief comparison of DE gene responses in the two systems is relevant, specifically because the A-549 response (high infection, non-MGEP system) can serve as a positive control with which to validate the MGEP response. We hypothesized that if the responses observed were a true MGEP signal, then a sub-set of the genes identified as DE in K-562s would likely overlap with those identified in A-549s. Indeed, 28 of the 363 up-regulated genes (7.71%) and 17 of the 108 down-regulated genes (15.74%) in K-562s are similarly up- and down-regulated in A-549s (Figures 6C and 6D, and complete list in Supplementary Table 3), indicating the measured DE gene response is likely the signature of DENV infection in this MGEP model system. Finally, to complete validation of this MGEP model system we carried out functional enrichment analysis to elucidate whether MGEP analysis by means of RP could lead to the identification of DE gene sets known to be implicated in DENV infection. We used STRING v11 [19] to analyze the subset of DE gene tags coding for proteins only (130 total). When the minimum required interaction score was set to “medium confidence (0.400),” STRING detected 129 nodes (proteins) and 60 edges or functional associations (29 more than expected if the nodes were selected randomly), with an average node degree of 0.93, an average local clustering coefficient of 0.294, and a PPI enrichment P-value of 2.64e-06. A number of the DE gene tags identified in our study appeared as players in functional interactions within iron homeostasis, hypoxia and apoptosis gene sets that have been previously defined (Figure 7A-C) [50–53]. This is reassuring given that these processes are known to be active during DENV infection [54–60]. To further test the resilience of the MGEP model system we increased the required minimum interaction score to “highest confidence (0.900).” At this level, 11 edges were detected (6 more than expected), with an average node degree of 0.171, an average local clustering coefficient of 0.0724, and a PPI enrichment P-value of .012. Interestingly, this increase in stringency did not change the total number of functional associations present in the calculated network, only the individual false discovery rate, which were lower at higher confidence levels, as expected (not shown). This is significant, particularly in the context of MGEP systems, because it suggests that lowering the confidence levels may be a way to identify potentially interesting genes that are not initially seen in MGEP systems because of their inherent low signal-to-noise ratio. If so, these identifications could later be validated using alternative experimental approaches. An example of this is illustrated by CBX4 (Chromobox Homolog 4). STRING identified functional enrichment around CBX4 (Figure 7D), a protein that was not detected using EdgeR. However, others have shown that CBX4 is down-regulated in patients with dengue shock syndrome relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45], demonstrating that this approach to data mining is potentially very useful. In addition, STRING produces a wealth of additional information related to the identified functional enrichment sets (Table 3), including relevant literature digests (Supplementary Table 4), that provide further useful leads in the search for signatures of pathogenesis that might not be detected in a basic MGEP analysis. Table 3. STRING v11 output of INTERPRO and PFAM protein domains and features of the enrichment analysis of 130 DE proteins tags, with the required minimum interaction score set to “medium confidence (0.400).” . INTERPRO Protein Domains and Features . . . . . . . . Domain . Description . Observed gene count . Background gene count . FDR . Matching proteins in network (IDs) . Matching proteins in network (labels) . INTERPRO Protein Domains IPR028139 Humanin family 5 11 1.60E-05 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 IPR026052 DNA-binding protein inhibitor 3 4 .0016 ENSP00000365280 ID1 ENSP00000234091 ID2 ENSP00000363689 ID3 PFAM Protein Domains PF15040 Humanin family 6 12 2.13E-07 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 ENSP00000439666 MTRNR2L8 . INTERPRO Protein Domains and Features . . . . . . . . Domain . Description . Observed gene count . Background gene count . FDR . Matching proteins in network (IDs) . Matching proteins in network (labels) . INTERPRO Protein Domains IPR028139 Humanin family 5 11 1.60E-05 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 IPR026052 DNA-binding protein inhibitor 3 4 .0016 ENSP00000365280 ID1 ENSP00000234091 ID2 ENSP00000363689 ID3 PFAM Protein Domains PF15040 Humanin family 6 12 2.13E-07 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 ENSP00000439666 MTRNR2L8 Open in new tab Table 3. STRING v11 output of INTERPRO and PFAM protein domains and features of the enrichment analysis of 130 DE proteins tags, with the required minimum interaction score set to “medium confidence (0.400).” . INTERPRO Protein Domains and Features . . . . . . . . Domain . Description . Observed gene count . Background gene count . FDR . Matching proteins in network (IDs) . Matching proteins in network (labels) . INTERPRO Protein Domains IPR028139 Humanin family 5 11 1.60E-05 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 IPR026052 DNA-binding protein inhibitor 3 4 .0016 ENSP00000365280 ID1 ENSP00000234091 ID2 ENSP00000363689 ID3 PFAM Protein Domains PF15040 Humanin family 6 12 2.13E-07 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 ENSP00000439666 MTRNR2L8 . INTERPRO Protein Domains and Features . . . . . . . . Domain . Description . Observed gene count . Background gene count . FDR . Matching proteins in network (IDs) . Matching proteins in network (labels) . INTERPRO Protein Domains IPR028139 Humanin family 5 11 1.60E-05 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 IPR026052 DNA-binding protein inhibitor 3 4 .0016 ENSP00000365280 ID1 ENSP00000234091 ID2 ENSP00000363689 ID3 PFAM Protein Domains PF15040 Humanin family 6 12 2.13E-07 ENSP00000439228 MTRNR2L1 ENSP00000442159 MTRNR2L10 ENSP00000443339 MTRNR2L3 ENSP00000382856 MTRNR2L4 ENSP00000439985 MTRNR2L7 ENSP00000439666 MTRNR2L8 Open in new tab Figure 7. Open in new tabDownload slide Functional enrichment analysis with STRING v11. A, Iron homeostasis gene set [50]. B, Hypoxia gene set 1 in light grey [51] and gene set 2 in dark grey [52]. C, Apoptosis gene set [61]. D, CBX4 gene set [53]. Figure 7. Open in new tabDownload slide Functional enrichment analysis with STRING v11. A, Iron homeostasis gene set [50]. B, Hypoxia gene set 1 in light grey [51] and gene set 2 in dark grey [52]. C, Apoptosis gene set [61]. D, CBX4 gene set [53]. DISCUSSION This proof-of-concept study assesses the ability of RP to uncover potential phenotypic signatures in MGEP systems. Our main question was whether RP is sufficiently powerful to detect DE genes in small cellular pools embedded within larger, non-expressive pools. Our MGEP model system consisted of human cells eliciting <3% productive DENV infection. We show that RP can extract relevant signals in this context, confirming the approach has enormous potential to uncover pathogenic signatures associated with human disease. We envision this will be particularly useful to study infectious processes, but other biologically relevant associations, like symbiosis, parasitism and mutualism, also stand to benefit. Our study included an adaptation of the original RP protocol [5] from the Illumina to the Ion platform, thus taking advantage of shorter sequencing turnaround times, and also expanding the availability of platforms with which to conduct PR studies. We show the adaptation successfully identified ribosome pausing sites in a replicate-consistent manner (Figure 4), even when ribosome density was extremely low, as in the case of reads mapped to DENV (1.7 to 11 reads/DENV nucleotide, not shown). This constituted the first indication that the system was capable of reproducibility at the replicate level, despite representing only 3% of the signal of interest. Nevertheless, sources of unwanted variation were a major issue that needed to be properly addressed, as expected. We demonstrated that highly sensitive normalization tools like RUVseq [15, 62] are crucial to uncover desired signals (Figure 5). We identified 471 DE genes (Supplementary Table 2), including 130 (27.6%) protein coding genes, 144 (30.6%) pseudogenes, and 92 (19.5%) ncRNAs. Many of the protein coding genes appeared in functional enrichment analyses (Figure 7), and have documented functions during DENV pathogenesis or related viral processes (Table 2). For example, VEGFA (vascular endothelial growth factor A), which along with its receptor mediates vascular permeability, can serve as a predictor of dengue disease severity [33–35]. VEGFs are known to interact with Rho guanine nucleotide exchange factors (Rho-GEFs) that integrate diverse cellular signals into coordinated functional responses [36]. We identified two such Rho-GEFs, ARHGEF11 (Rho guanine nucleotide exchange factor (GEF) 11) and TRIO (trio Rho guanine nucleotide exchange factor). Both are VEGF-interacting proteins [36] and TRIO is up-regulated in patients with dengue fever and dengue hemorrhagic fever [37]. Another interesting protein was IFIT2, which is induced during DENV infection and known to have anti-viral functions [41, 42]. Interestingly, the gene with the highest logFC was RARRES2 (Figure 6C and Table 2), which is a host-cell factor that influences respiratory syncitial virus (RSV) replication [43]. Although a specific role for RARRES2 during DENV infection is unknown, its role during RSV infection serves as precedent for possible role(s) during other viral infections, including DENV. A related gene, RARRES3, although not identified in K-562 cells, was identified in A-549 control cells, and is up-regulated in patients with dengue shock syndrome relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45, 63]. Other proteins in our screen (LGALS3BP, CYP19A1, TBC1D8B, FTH1 and ID1) have also been implicated in dengue infection, while an additional subset (NDRG1, VSTM2L, ANKMY1, ANKRD30BL, ANKRD37, ANKRD6, ID2 and ID3) has roles in other viral processes (Table 2). Comparison of DE gene responses in the MGEP (<3% infection) and non-MGEP (>92% infection) model systems revealed an even larger subset of interesting DE genes. Specifically, 28 of 363 (7.71%) and 17 of 108 (15.74%) DE genes were similarly up- or down-regulated, respectively, in both systems (Figure 6). IFIT2, for example, had an unremarkable expression increase in K-562s in comparison to A-549s. However, this increase was measurable, and most importantly, associated to the most statistically significant P-value in the set. Another gene with a significant P-value up-regulated in both cases was HERPUD1. HERPUD1 is induced during endoplasmic reticulum stress [64], a well-known result of RNA viral infection [65]. Although the role of HERPUD1 during DENV infection has not been established, other viruses directly up-regulate HERPUD1 [66]. This would suggest that if/when HERPUD1 is up-regulated during DENV infection, it may play a role in ER-related changes. We also found a set of three proteins, ID-1, ID-2 and ID-3, that were down-regulated in both systems. Members of the ID (inhibitor of DNA binding/differentiation) family of proteins inhibit DNA binding by heterodimerization with basic helix–loop–helix proteins [67]. While a role for these proteins during DENV infection has not been described, the fact that they are identified as a group that is down-regulated in two strikingly different cell lines merits further attention. Our study also identified a number of tags associated to pseudogenes and ncRNAs (Supplementary Table 2). RFs originating from pseudogenes and ncRNAs now appear to be pervasive in RP studies [30, 68–72]. We initially considered that their presence might represent an artifact resulting from the use of a sucrose cushion for monosome isolation (see Materials and Methods). However, their ubiquitous nature and the fact they were picked up as significantly DE in our analysis warrants further discussion. Two ncRNA subspecies, specifically long non-coding (lnc)RNAs and micro (mi)RNAs, are emerging as important regulators of pathogenesis, the immune system, viral replication and gene expression (Table 2) [73–75]. RP data has been used to show that although a majority of lncRNAs remain non-coding, around 5% of them elicit behavior (based on measurements of ribosome release scores) similar to protein coding genes [76], suggesting that some RNA species previously termed “non-coding” may not be as non-coding as anticipated. This may explain why lncRNAs residing outside of annotated protein-coding genes are pervasively translated with RF distributions indistinguishable from those of bona fide RFs [30]. Their mechanism(s) of expression remain unknown, but their regulatory roles in viral replication, the innate immune response, and pathogenesis are starting to emerge [77–79]. Of the 8 lncRNAs identified here, 6 were novel products without known relationship to DENV or other viruses. One identified lncRNA, MALAT1 (metastasis associated lung adeno-carcinoma transcript 1) (Table 2), is differentially up-regulated in patients with dengue fever, dengue hemorrhagic fever and dengue shock syndrome [37]. Another identified lncRNA, LINC00339, is the subject of transcription read-through during herpes simplex virus 1 infection [46]. Much remains to be uncovered regarding these poorly understood DE tags, but two pieces of evidence provide support for standard expression of their loci: first, a parallel RNAseq analysis detects both MALAT1 and LINC00339 as DE genes with the same directionality as in RP (Table 2); and second, our RNAseq analysis was performed using polyA RNA. Similarly to the case of lncRNAs, miRNAs are beginning to emerge as potentially relevant actors during pathogenesis, and many viruses synthesize their own miRNAs [74]. Of the ten miRNAs identified in our dataset, six are novel products with no specified name or known function. The remaining three (miR21, miR142 and miR144) (Table 2) have documented associations to DENV, to related flavi-viruses like West-Nile virus (WNV) [75], and to other viruses like Epstein-Barr virus (EBV), which has been linked to the development of EBV-associated Burkitt’s lymphoma [80]. miR21 is particularly interesting because, in addition to being the most down-regulated gene in our entire dataset of 471 genes (with a logFC value of -16.23), others have shown that it is differentially expressed in DENV-infected patients and that its levels can be used to distinguish infected from non-infected patients with both sensitivity and specificity [81]. miR142 is also interesting because it is considered a hematopoietic-specific miRNA [82, 83]. As such, miR142 target sites have been artificially engineered into DENV-2 to show that these can be exploited to attenuate viral replication in dendritic cells and macrophages [47], two of the primary targets of DENV infection along with monocytes [84]. Our study in K-562s, a cell line with the capacity to differentiate into monocytes, complements those observations and suggests a mechanism by which DENV infection may induce one or more hematopoietic-specific miRNAs, like miR142, that in turn function to down-regulate DENV. Finally, mir144 does not have a known function during DENV infection. However, it has been shown as up-regulated after Crimean Congo virus (CCV) infection in human subjects [49]. Taken together, our data and the literature suggests that ncRNAs are far from being unlikely modulators of the interactions between viruses and their cellular hosts. A functional enrichment analysis of the protein codes genes performed with STRING v11 [19] revealed that changing the stringency settings of the minimum required interaction score from high to medium did not affect the total number of significant functional enrichments detected, just the confidence level of each. Given the inherent low signal-to-noise ratios expected of MGEP systems, we believe that setting this parameter to moderate confidence will be crucial to detect as many functional enrichments as possible in other MGEP studies. In our hands, STRING identified 7 of 37 components in the iron homeostasis gene set defined by Luo et al [50]. (Figure 7A and Table 3). The importance of iron homeostasis during DENV and other flavivirus infections has been demonstrated in mosquito vectors [54–56]. The analysis also identified genes in hypoxia [51, 52, 85] and apoptosis gene sets [59] (Figures 7B and C and Table 3). Both of these processes are hallmarks of DENV infection [57–59], and there is evidence to suggest that hypoxia in particular may constitute a ubiquitous signature during inflammatory and infectious disease processes [60]. We want to point out that because MGEP systems will inherently have low signal-to-noise ratios that likely result in a decreased number of DE genes identified, a functional enrichment analysis like the one performed with STRING will tend to recover poorly-to-moderately populated edges. Nevertheless, the functional associations identified can serve to predict additional DE genes of interest that could later be confirmed using alternative methods, like quantitative PCR. For example, STRING identified a functional enrichment centered on CBX4 (Chromobox Homolog 4) (Figure 7D and Table 3), but the protein was not detected as a DE gene with edgeR. However, CBX4 is down-regulated in patients with dengue shock syndrome relative to patients with classical dengue fever or dengue hemorrhagic fever grades I/II [45], indicating that CBX4 has potential as protein of interest during DENV infection. Somewhat related to this finding, the TEXTMINING function of STRING was particularly useful during our analysis because it produced accurately targeted literature digests that enabled us to focus on papers that proved to be highly relevant to the identified DE proteins (Supplementary Table 4). In summary, RP data along with edgeR analysis has enabled identification of a number of individual genes that together constitute signatures of DENV infection in an MGEP model system where the signal of interest derives from less than 3% of the cells examined. Furthermore, when the data is further analyzed with tools like STRING, the resulting functional association studies can be used as a gene discovery tool to shed light on particular genes that may not be initially identified as DE genes. This is particularly relevant in MGEP model systems, where by definition, the total number of DE genes identified will be significantly lower than in an equivalent non-MGEP system, thus potentially missing important players that are expressed at very low levels, and therefore, undetectable. Our results also show that RP can identify DE genes that are not identified as DE using RNAseq analysis only (Table 2). This is likely due to the fact a correlation between transcription and translation does not always hold throughout eukaryotic organisms [7], and indicates that the DE gene phenotypes obtained from RP studies, which include both transcriptomic and proteomic measurements, are probably more accurate than those obtained from RNAseq studies alone, which use transcriptomic data to extrapolate and infer proteomic behavior. Our proof-of-concept study shows that RP is a powerful tool for the study of gene expression in the context of MGEP systems. Nevertheless, our analysis has also uncovered potential avenues for improvement. First, we found that sequencing depth was critical for successful DE gene identification in MGEP systems. Initial trials with ~30M reads/sample, which have been sufficient in other set-ups where low signal-to-noise ratios are not an issue, failed to produce detectable DE gene signals (not shown). DE gene signal detection in MGEP systems required increased sequencing depth. This was achieved using high capacity PIv2 sequencing chips (see Materials and Methods), which contain 165M wells each, allowing us to generate an average of 104M usable reads/run (Table 1). Another area with room for improvement was the presence of elevated numbers (30–55% of the total usable reads) of contaminating rRNAs (Figure 2), even after one round of physical depletion using an oligo depletion pool (Supplementary Table 5). Contaminating rRNAs can occupy valuable real estate in the sequencing chip, which can significantly lower the amount of usable data per experiment. To address this, efforts to reduce them by any means possible should be made, including through the use of additional depletion oligos to eliminate known contaminants (Supplementary Table 5) and through additional rounds of physical depletion (not shown). We also identified a number of signal sources (ncRNAs) that were initially concerning because they are considered non-mRNA signal sources. As such, we worried they may represent an artifact linked to the sucrose cushions used for monosome isolation. Although there is now clear and ample precedent in the literature for their presence in RP datasets, one way to address them directly is to eliminate the sucrose cushion and instead use a sucrose gradient [5]. However, both sucrose cushions and gradients require ultracentrifugation, which is not always available. Since this study we have adopted size-exclusion columns (not shown), which have the added advantage of addressing the issue at hand while reducing processing times and eliminating the need for expensive equipment like ultracentrifuges. In summary, any action aimed at increasing the sequencing depth, such as improvements in chip capacity and/or efforts to enhance the representation of bona fide reads, are expected to improve the desired DE gene signal. Finally, we knew going in that because the number of replicates is a crucial variable for the identification of factors of unwanted variation in RNAseq experiments, that by extension, replicates would also be important in an MGEP RP set-up. We opted to run the minimum number of replicates required for statistical inference, namely three [28, 86, 87], for two reasons. First, we wanted our MGEP model system set-up to be as relevant as possible to real-life situations, where multiple samples from the same patient, or multiple samples of the same type, are not always available for analysis. Second, we wanted to control costs given that many laboratories often face scarce funding situations that significantly limit their choices at the design table. With only three replicates we encountered significant sources of unwanted variation. Nevertheless, we were still able to normalize the data and identify DE gene profiles of interest. That said, an increased number of biological replicates will enhance the false discovery rate and the power of DE gene analysis more significantly than an expansion in sequencing depth alone [87–90]. Thus, if given a choice, MGEP studies should prioritize additional replicates over generating more sequence depth per sample. In conclusion, we show that RP is a powerful tool for gene discovery in the context of MGEP systems. Furthermore, we point out clear avenues that should lead to significant improvements in and diversification of applications, particularly in the context of MGEP scenarios. Some of these improvements are already being implemented. For example, we are currently conducting a study to evaluate the immunogenicity of two different pertussis vaccines in healthy infants in Peru [9], where improvements discussed above have significantly impacted the quality of the data in a positive way. We believe that RP has enormous potential for the study of human disease, and in particular for neglected or difficult to study infectious diseases. Properly applied, this tool can also be employed for the development of new therapeutic targets and for the evaluation of existing drugs and vaccines. Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Notes Acknowledgments. We gratefully acknowledge Karina Leiva, Lucy Espinoza and Christian Baldeviano at NAMRU-6 for technical support with FACS analysis; Gary Wessel at Brown University for providing access to his fluorescent microscope and image analysis software; Mark Robinson at the University of Zurich for advice on EdgeR Robust; Davide Risso at the University of California Berkeley for advice on RUVseq; and Thomas J. Hardcastle at the University of Cambridge for advice on RiboSeqR. Disclaimer. The views expressed in this article are those of the authors and do not reflect the official policy or position of the Department of the Navy, Department of Defense, nor the U.S. Government. Some of the authors were employees of the U.S. Government at the time the work was performed, and the work was prepared as part of their official duties. Title 17 U.S.C. §105 provides that ‘Copyright protection under this title is not available for any work of the United States Government.’ Title 17 U.S.C. §101 defines U.S. Government work as a work prepared by military service members or employees of the U.S. Government as part of that person’s official duties. Financial support. The work was supported by an award to ML from the In-House Laboratory Independent Research Program of the Naval Medical Research Center (Project ID: ILIR-4514). Supplement sponsorship. This supplement is sponsored by WRAIR, LANL, USAMRIID, PUCP (Pontificia Universidad Catolica del Peru), USAFSAM, NIH. Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. References 1. Ingolia NT , Brar GA, Rouskin S, McGeachy AM, Weissman JS. The ribosome profiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments . Nat Protoc 2012 ; 7 : 1534 – 50 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Michel AM , Baranov PV. Ribosome profiling: a Hi-Def monitor for protein synthesis at the genome-wide scale . Wiley Interdiscip Rev RNA 2013 ; 4 : 473 – 90 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Ingolia NT . Ribosome profiling: new views of translation, from single codons to genome scale . Nat Rev Genet 2014 ; 15 : 205 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Wang Z , Gerstein M, Snyder M. RNA-Seq: a revolutionary tool for transcriptomics . Nat Rev Genet 2009 ; 10 : 57 – 63 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Ingolia NT , Ghaemmaghami S, Newman JR, Weissman JS. Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling . Science 2009 ; 324 : 218 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Saayman S , Roberts TC, Morris KV, Weinberg MS. HIV Latency and the noncoding RNA therapeutic landscape . Adv Exp Med Biol 2015 ; 848 : 169 – 89 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Vogel C , Marcotte EM. Insights into the regulation of protein abundance from proteomic and transcriptomic analyses . Nat Rev Genet 2012 ; 13 : 227 – 32 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Liu TY , Huang HH, Wheeler D, et al. Time-resolved proteomics extends ribosome profiling-based measurements of protein synthesis dynamics . Cell Syst 2017 ; 4 : 636 – 644.e9 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Leguia M , Juarez D, Loyola S, et al. Use of ribosome profiling and RNA-seq in a systems biology-based assessment of immune responses to whole cell and acellular pertussis vaccines in healthy infants in Peru . Am J Trop Med Hyg . [67th Annual Meeting of the American Society for Tropical Medicine and Hygiene, New Orleans, LA, 29 October–1 November 2018]. 2018 ; 99 . OpenURL Placeholder Text WorldCat 10. van Knippenberg PH , Duijts GA. Polyribosomes of Escherichia coli. II. In vivo and in vitro studies on DNA bound and cytoplasmic polysomes . Mol Gen Genet 1971 ; 112 : 208 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Martin M . Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J 2011; 17:10–12. Available at: https://journal.embnet.org/index.php/embnetjournal/article/view/200. Accessed 15 November 2019. 12. Langmead B , Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome . Genome Biol 2009 ; 10 : R25 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Trapnell C , Pachter L, Salzberg SL. TopHat: discovering splice junctions with RNA-Seq . Bioinformatics 2009 ; 25 : 1105 – 11 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Langmead B , Salzberg SL. Fast gapped-read alignment with Bowtie 2 . Nat Methods 2012 ; 9 : 357 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Risso D , Ngai J, Speed TP, Dudoit S. Normalization of RNA-seq data using factor analysis of control genes or samples . Nat Biotechnol 2014 ; 32 : 896 – 902 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Zhou X , Lindsay H, Robinson MD. Robustly detecting differential expression in RNA sequencing data using observation weights . Nucleic Acids Res . 2014 ; 42 : 20 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Robinson MD , McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data . Bioinformatics 2010 ; 26 : 139 – 40 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Hardcastle TJ. riboSeqR: analysis of sequencing data from ribosome profiling experiments. R package version 1.6.0 . 2014 . http://bioconductor.org/packages/release/bioc/html/riboSeqR.html. Accessed September 2019. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 19. Szklarczyk D , Gable AL, Lyon D, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets . Nucleic Acids Res . 2018 ; 47 ( D1 ): D607 – 13 . Google Scholar Crossref Search ADS WorldCat 20. Lozzio CB , Lozzio BB. Human chronic myelogenous leukemia cell-line with positive Philadelphia chromosome . Blood 1975 ; 45 : 321 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Choudhury MA , Lott WB, Aaskov J. Distribution of fitness in populations of dengue viruses . PLoS One 2014 ; 9 : e107264 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Giard DJ , Aaronson SA, Todaro GJ, et al. In vitro cultivation of human tumors: establishment of cell lines derived from a series of solid tumors . J Natl Cancer Inst 1973 ; 51 : 1417 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Chiu HC , Hannemann H, Heesom KJ, Matthews DA, Davidson AD. High-throughput quantitative proteomic analysis of dengue virus type 2 infected A549 cells . PLoS One 2014 ; 9 : e93305 . Google Scholar Crossref Search ADS PubMed WorldCat 24. De Maio FA , Risso G, Iglesias NG, et al. The dengue virus NS5 protein intrudes in the cellular spliceosome and modulates splicing . PLoS Pathog 2016 ; 12 : e1005841 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Fink J , Gu F, Ling L, et al. Host gene expression profiling of dengue virus infection in cell lines and patients . PLoS Negl Trop Dis 2007 ; 1 : e86 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Umareddy I , Tang KF, Vasudevan SG, Devi S, Hibberd ML, Gu F. Dengue virus regulates type I interferon signalling in a strain-dependent manner in human cell lines . J Gen Virol 2008 ; 89 : 3052 – 62 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Yu CY , Chang TH, Liang JJ, et al. Dengue virus targets the adaptor protein MITA to subvert host innate immunity . PLoS Pathog 2012 ; 8 : e1002780 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Conesa A , Madrigal P, Tarazona S, et al. A survey of best practices for RNA-seq data analysis . Genome Biol 2016 ; 17 : 13 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Guo H , Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAs predominantly act to decrease target mRNA levels . Nature 2010 ; 466 : 835 – 40 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Ingolia NT , Brar GA, Stern-Ginossar N, et al. Ribosome profiling reveals pervasive translation outside of annotated protein-coding genes . Cell Rep 2014 ; 8 : 1365 – 79 . Google Scholar Crossref Search ADS PubMed WorldCat 31. Martens AT , Taylor J, Hilser VJ. Ribosome A and P sites revealed by length analysis of ribosome profiling data . Nucleic Acids Res 2015 ; 43 : 3680 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Leek JT , Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments . Bioinformatics 2012 ; 28 : 882 – 3 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Kalita J , Chauhan PS, Mani VE, Bhoi SK, Misra UK. VEGF and its receptors in dengue virus infection . J Med Virol 2015 ; 87 : 1449 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Del Moral-Hernández O , Martínez-Hernández NE, Mosso-Pani MA, et al. Association DENV1 and DENV2 infection with high serum levels of soluble thrombomodulin and VEGF in patients with dengue fever and dengue hemorrhagic fever . Int J Clin Exp Med 2014 ; 7 : 370 – 8 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 35. van de Weg CA , Pannuti CS, van den Ham HJ, et al. Serum angiopoietin-2 and soluble VEGF receptor 2 are surrogate markers for plasma leakage in patients with acute dengue virus infection . J Clin Virol . 2014 ; 60 : 328 – 35 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Hernández-García R , Iruela-Arispe ML, Reyes-Cruz G, Vázquez-Prado J. Endothelial RhoGEFs: a systematic analysis of their expression profiles in VEGF-stimulated and tumor endothelial cells . Vascul Pharmacol 2015 ; 74 : 60 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Kwissa M , Nakaya HI, Onlamoon N, et al. Dengue virus infection induces expansion of a CD14(+)CD16(+) monocyte population that stimulates plasmablast differentiation . Cell Host Microbe 2014 ; 16 : 115 – 27 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Schweitzer CJ , Zhang F, Boyer A, Valdez K, Cam M, Liang TJ. N-Myc downstream-regulated gene 1 restricts hepatitis C virus propagation by regulating lipid droplet biogenesis and viral assembly . J Virol . 2018 ; 92 : e01166 - 17 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 39. Olagnier D , Peri S, Steel C, et al. Cellular oxidative stress response controls the antiviral and apoptotic programs in dengue virus-infected dendritic cells . PLoS Pathog 2014 ; 10 : e1004566 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Warke RV , Martin KJ, Giaya K, Shaw SK, Rothman AL, Bosch I. TRAIL is a novel antiviral protein against dengue virus . J Virol 2008 ; 82 : 555 – 64 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Hsu YL , Shi SF, Wu WL, Ho LJ, Lai JH. Protective roles of interferon-induced protein with tetratricopeptide repeats 3 (IFIT3) in dengue virus infection of human lung epithelial cells . PLoS One 2013 ; 8 : e79518 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Diamond MS , Farzan M. The broad-spectrum antiviral functions of IFIT and IFITM proteins . Nat Rev Immunol 2013 ; 13 : 46 – 57 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Shirato K , Ujike M, Kawase M, Matsuyama S. Identification of CCL2, RARRES2 and EFNB2 as host cell factors that influence the multistep replication of respiratory syncytial virus . Virus Res 2015 ; 210 : 213 – 26 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Rouillard AD , Gundersen GW, Fernandez NF, et al. The harmonizome: a collection of processed datasets gathered to serve and mine knowledge about genes and proteins . Database 2016 ; 2016 : baw100 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Devignot S , Sapet C, Duong V, et al. Genome-wide expression profiling deciphers host responses altered during dengue shock syndrome and reveals the role of innate immunity in severe dengue . PLoS One 2010 ; 5 : e11671 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Hennig T , Michalski M, Rutkowski AJ, et al. HSV-1-induced disruption of transcription termination resembles a cellular stress response but selectively increases chromatin accessibility downstream of genes . PLoS Pathog 2018 ; 14 : e1006954 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Pham AM , Langlois RA, TenOever BR. Replication in cells of hematopoietic origin is necessary for dengue virus dissemination . PLoS Pathog 2012 ; 8 : e1002465 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Kanokudom S , Vilaivan T, Wikan N, Thepparit C, Smith DR, Assavalapsakul W. miR-21 promotes dengue virus serotype 2 replication in HepG2 cells . Antiviral Res 2017 ; 142 : 169 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Demir ZC , Bastug A, Bodur H, Ergunay K, Ozkul A. MicroRNA expression profiles in patients with acute Crimean Congo hemorrhagic fever reveal possible adjustments to cellular pathways . J Med Virol 2017 ; 89 : 417 – 22 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Luo XQ , Zhang CY, Zhang JW, et al. Identification of iron homeostasis genes dysregulation potentially involved in retinopathy of prematurity pathogenicity by microarray analysis . J Ophthalmol 2015 ; 2015 : 584854 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 51. Chan MC , Ilott NE, Schödel J, et al. Tuning the transcriptional response to hypoxia by inhibiting hypoxia-inducible factor (HIF) prolyl and asparaginyl hydroxylases . J Biol Chem 2016 ; 291 : 20661 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Chen C , Lou T. Hypoxia inducible factors in hepatocellular carcinoma . Oncotarget 2017 ; 8 : 46691 – 703 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 53. Yang J , Cheng D, Zhu B, Zhou S, Ying T, Yang Q. Chromobox homolog 4 is positively correlated to tumor growth, survival and activation of HIF-1α signaling in human osteosarcoma under normoxic condition . J Cancer 2016 ; 7 : 427 – 35 . Google Scholar Crossref Search ADS PubMed WorldCat 54. Chisenhall DM , Londono BL, Christofferson RC, McCracken MK, Mores CN. Effect of dengue-2 virus infection on protein expression in the salivary glands of Aedes aegypti mosquitoes . Am J Trop Med Hyg 2014 ; 90 : 431 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Tsujimoto H , Anderson MAE, Myles KM, Adelman ZN. Identification of candidate iron transporters from the ZIP/ZnT gene families in the mosquito Aedes aegypti . Front Physiol 2018 ; 9 : 380 . Google Scholar Crossref Search ADS PubMed WorldCat 56. Duchemin JB , Paradkar PN. Iron availability affects West Nile virus infection in its mosquito vector . Virol J 2017 ; 14 : 103 . Google Scholar Crossref Search ADS PubMed WorldCat 57. Gan ES , Cheong WF, Chan KR, et al. Hypoxia enhances antibody-dependent dengue virus infection . EMBO J 2017 ; 36 : 1348 – 63 . Google Scholar Crossref Search ADS PubMed WorldCat 58. Frakolaki E , Kaimou P, Moraiti M, et al. The role of tissue oxygen tension in dengue virus replication . Cells 2018 ; 7: E241 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Courageot MP , Catteau A, Desprès P. Mechanisms of dengue virus-induced cell death . Adv Virus Res 2003 ; 60 : 157 – 86 . Google Scholar Crossref Search ADS PubMed WorldCat 60. Santos SAD , Andrade DR Júnior. HIF-1alpha and infectious diseases: a new frontier for the development of new therapies . Rev Inst Med Trop Sao Paulo 2017 ; 59 : e92 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 61. Holt SH , Darash-Yahana M, Sohn YS, et al. Activation of apoptosis in NAF-1-deficient human epithelial breast cancer cells . J Cell Sci 2016 ; 129 : 155 – 65 . Google Scholar Crossref Search ADS PubMed WorldCat 62. Peixoto L , Risso D, Poplawski SG, et al. How data analysis affects power, reproducibility and biological insight of RNA-seq studies in complex datasets . Nucleic Acids Res 2015 ; 43 : 7664 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Karyala P , Metri R, Bathula C, et al. DenHunt - a comprehensive database of the intricate network of dengue-human interactions . PLoS Negl Trop Dis 2016 ; 10 : e0004965 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Kokame K , Agarwala KL, Kato H, Miyata T. Herp, a new ubiquitin-like membrane protein induced by endoplasmic reticulum stress . J Biol Chem 2000 ; 275 : 32846 – 53 . Google Scholar Crossref Search ADS PubMed WorldCat 65. Jheng JR , Ho JY, Horng JT. ER stress, autophagy, and RNA viruses . Front Microbiol 2014 ; 5 : 388 . Google Scholar Crossref Search ADS PubMed WorldCat 66. Fung TS , Liao Y, Liu DX. The endoplasmic reticulum stress sensor IRE1α protects cells from apoptosis induced by the coronavirus infectious bronchitis virus . J Virol 2014 ; 88 : 12752 – 64 . Google Scholar Crossref Search ADS PubMed WorldCat 67. Benezra R , Davis RL, Lockshon D, Turner DL, Weintraub H. The protein Id: a negative regulator of helix-loop-helix DNA binding proteins . Cell 1990 ; 61 : 49 – 59 . Google Scholar Crossref Search ADS PubMed WorldCat 68. Bazin J , Baerenfaller K, Gosai SJ, Gregory BD, Crespi M, Bailey-Serres J. Global analysis of ribosome-associated noncoding RNAs unveils new modes of translational regulation . Proc Natl Acad Sci U S A 2017 ; 114 : E10018 – 27 . Google Scholar Crossref Search ADS PubMed WorldCat 69. Zeng C , Fukunaga T, Hamada M. Identification and analysis of ribosome-associated lncRNAs using ribosome profiling data . BMC Genomics 2018 ; 19 : 414 . Google Scholar Crossref Search ADS PubMed WorldCat 70. Hsu PY , Calviello L, Wu HL, et al. Super-resolution ribosome profiling reveals unannotated translation events in Arabidopsis . Proc Natl Acad Sci U S A 2016 ; 113 : E7126 – 35 . Google Scholar Crossref Search ADS PubMed WorldCat 71. Popa A , Lebrigand K, Barbry P, Waldmann R. Pateamine A-sensitive ribosome profiling reveals the scope of translation in mouse embryonic stem cells . BMC Genomics 2016 ; 17 : 52 . Google Scholar Crossref Search ADS PubMed WorldCat 72. Ji Z , Song R, Regev A, Struhl K. Many lncRNAs, 5’UTRs, and pseudogenes are translated and some are likely to express functional proteins . Elife 2015 ; 4 : e08890 . Google Scholar Crossref Search ADS PubMed WorldCat 73. Gantier MP . New perspectives in MicroRNA regulation of innate immunity . J Interferon Cytokine Res 2010 ; 30 : 283 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 74. Tycowski KT , Guo YE, Lee N, et al. Viral noncoding RNAs: more surprises . Genes Dev 2015 ; 29 : 567 – 84 . Google Scholar Crossref Search ADS PubMed WorldCat 75. Bavia L , Mosimann AL, Aoki MN, Duarte Dos Santos CN. A glance at subgenomic flavivirus RNAs and microRNAs in flavivirus infections . Virol J 2016 ; 13 : 84 . Google Scholar Crossref Search ADS PubMed WorldCat 76. Guttman M , Russell P, Ingolia NT, Weissman JS, Lander ES. Ribosome profiling provides evidence that large noncoding RNAs do not encode proteins . Cell 2013 ; 154 : 240 – 51 . Google Scholar Crossref Search ADS PubMed WorldCat 77. Rice AP . Roles of microRNAs and long-noncoding RNAs in human immunodeficiency virus replication . Wiley Interdiscip Rev RNA 2015 ; 6 : 661 – 70 . Google Scholar Crossref Search ADS PubMed WorldCat 78. Ouyang J , Hu J, Chen JL. lncRNAs regulate the innate immune response to viral infection . Wiley Interdiscip Rev RNA 2016 ; 7 : 129 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 79. Ding YZ , Zhang ZW, Liu YL, Shi CX, Zhang J, Zhang YG. Relationship of long noncoding RNA and viruses . Genomics 2016 ; 107 : 150 – 4 . Google Scholar Crossref Search ADS PubMed WorldCat 80. Lu F , Weidmer A, Liu CG, Volinia S, Croce CM, Lieberman PM. Epstein-Barr virus-induced miR-155 attenuates NF-kappaB signaling and stabilizes latent virus persistence . J Virol 2008 ; 82 : 10436 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 81. Ouyang X , Jiang X, Gu D, et al. Dysregulated serum MiRNA profile and promising biomarkers in dengue-infected patients . Int J Med Sci 2016 ; 13 : 195 – 205 . Google Scholar Crossref Search ADS PubMed WorldCat 82. Brown BD , Venneri MA, Zingale A, Sergi Sergi L, Naldini L. Endogenous microRNA regulation suppresses transgene expression in hematopoietic lineages and enables stable gene transfer . Nat Med 2006 ; 12 : 585 – 91 . Google Scholar Crossref Search ADS PubMed WorldCat 83. Landgraf P , Rusu M, Sheridan R, et al. A mammalian microRNA expression atlas based on small RNA library sequencing . Cell 2007 ; 129 : 1401 – 14 . Google Scholar Crossref Search ADS PubMed WorldCat 84. Blackley S , Kou Z, Chen H, et al. Primary human splenic macrophages, but not T or B cells, are the principal target cells for dengue virus infection in vitro . J Virol 2007 ; 81 : 13325 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat 85. Soler-Torronteras R , Lara-Chica M, García V, Calzado MA, Muñoz E. Hypoximimetic activity of N-acyl-dopamines. N-arachidonoyl-dopamine stabilizes HIF-1α protein through a SIAH2-dependent pathway . Biochim Biophys Acta 2014 ; 1843 : 2730 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 86. Auer PL , Doerge RW. Statistical design and analysis of RNA sequencing data . Genetics 2010 ; 185 : 405 – 16 . Google Scholar Crossref Search ADS PubMed WorldCat 87. Liu Y , Zhou J, White KP. RNA-seq differential expression studies: more sequence or more replication? Bioinformatics 2014 ; 30 : 301 – 4 . Google Scholar Crossref Search ADS PubMed WorldCat 88. Lamarre S , Frasse P, Zouine M, et al. Optimization of an RNA-seq differential gene expression analysis depending on biological replicate number and library size . Front Plant Sci 2018 ; 9 : 108 . Google Scholar Crossref Search ADS PubMed WorldCat 89. Ching T , Huang S, Garmire LX. Power analysis and sample size estimation for RNA-Seq differential expression . RNA 2014 ; 20 : 1684 – 96 . Google Scholar Crossref Search ADS PubMed WorldCat 90. Schurch NJ , Schofield P, Gierliński M, et al. How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use? RNA 2016 ; 22 : 839 – 51 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes " A. V.-S. and D. J. contributed equally to the study. © The Author(s) 2020. Published by Oxford University Press for the Infectious Diseases Society of America. All rights reserved. For permissions, e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Minority Gene Expression Profiling: Probing the Genetic Signatures of Pathogenesis Using Ribosome Profiling
JF - The Journal of Infectious Diseases
DO - 10.1093/infdis/jiz565
DA - 2020-03-28
UR - https://www.deepdyve.com/lp/oxford-university-press/minority-gene-expression-profiling-probing-the-genetic-signatures-of-jtzf16G74f
SP - S341
VL - 221
IS - Supplement_3
DP - DeepDyve
ER -