The term “small non-coding RNA” broadly refers to diverse RNA species ~15–150 nucleotides (nt) in size that fulfill biological functions without being translated into proteins. While the involvement of small RNAs in cellular housekeeping processes such as transcript splicing and protein translation has been known since the 1960s, research over the past twenty years has revealed that small RNAs play vital roles in the regulation of gene expression, via both transcriptional and post-transcriptional mechanisms (Choudhuri, 2010).

Of the small RNAs involved in gene regulation, the most well-studied are microRNAs (miRNAs; ~22 nt in size), which facilitate post-transcriptional gene silencing by binding specific target mRNAs via base-pair complementarity, and either blocking translation or triggering transcript degradation (Ha and Kim, 2014). Another group of small RNAs that have been well characterized are Piwi-interacting RNAs (piRNAs), which silence transposons using a miRNA-like mechanism, in addition to inducing epigenetic modifications that influence the transcription of both transposons and protein-coding genes (Weick and Miska, 2014).

Tremendous progress has been made in the identification and characterization of small RNAs, and the current rate of discovery in this field suggests that much more remains to be elucidated. The development of next-generation sequencing (NGS) technology has proven instrumental to this progress, in part because it allows for identification of small RNAs without prior knowledge of their existence (in contrast with array-based or qPCR methods), and can discriminate between small RNA variants that differ by a single nucleotide. However, small RNA-seq library preparation is not without its challenges, which may include time-consuming enrichment steps prior to cDNA synthesis, and sample misrepresentation due to biases in small RNA end modification, reverse transcription, and PCR amplification.

A major source of bias in small RNA-seq data involves the manner in which small RNAs are captured during library construction (reviewed in Raabe et al., 2014). The most common method involves using a T4 RNA ligase (T4Rnl) to attach adapters to RNA 5′ and 3′ ends. However, T4Rnl exhibits sequence-specific substrate preferences, such that certain combinations of adapters and small RNAs are more readily incorporated than others, leading to sample misrepresentation in small RNA-seq libraries (Jayaprakash et al., 2011; Hafner et al., 2011). An alternative to adapter ligation is RNA 3′ polyadenylation, in which a poly(A) polymerase is used to add a stretch of repeated nucleotides to RNA 3′ ends. In contrast with adapter ligation, RNA polyadenylation occurs in a sequence-independent manner. While RNA 3′ polyadenylation was previously reported to generate small RNA-seq libraries (Berezikov, et al., 2006), this approach involved ligation of RNA 5′ ends, and was still susceptible to sequence-specific biases.

Here we present data from the SMARTer smRNA-Seq Kit for Illumina, which employs RNA 3′ polyadenylation and SMART (Switching Mechanism at the 5′ end of RNA Template) technology (Chenchik et al., 1998) to generate sequencing libraries in a ligation-independent manner. Rather than ligating adapters to small RNAs, this method incorporates adapters at both ends of nascent cDNAs during first-strand synthesis (Figure 1). Following polyadenylation of input RNA, first-strand cDNA synthesis is dT-primed (3′ smRNA dT Primer) and performed by the MMLV-derived PrimeScript Reverse Transcriptase (RT), which adds non-templated nucleotides upon reaching the 5′ end of each RNA template. The SMART smRNA Oligo then anneals to the non-templated nucleotides, and serves as a template for the incorporation of an additional sequence of nucleotides to the first-strand cDNA by the RT. Sequences incorporated at the 5′ and 3′ ends of each cDNA molecule serve as primer-annealing sites for PCR, which is performed using oligos that incorporate Illumina-compatible adapters and indexes during library amplification.

Assessing the accuracy of the SMARTer smRNA-Seq Kit for Illumina

To measure the accuracy of the SMARTer approach, the SMARTer smRNA-Seq Kit for Illumina was used to generate a sequencing library from the miRXplore Universal Reference (Miltenyi Biotec Inc.), an equimolar pool of 963 synthetic miRNAs. For comparison purposes, a sequencing library was prepared in parallel from the same starting material using an adapter-ligation-based method from Competitor N. Following sequencing and mapping, the number of reads corresponding to each miRNA was counted and normalized such that each miRNA in the library was expected to have a normalized value of "1". Given that the various miRNAs are present in equal quantities, each should be represented equally in the sequencing output, whereas observed differences in expression are conceivably due to biases imposed during library construction.

For the sequencing library generated using the adapter ligation kit from Competitor N, only ~22% of the miRNAs presented an expression level within a 2-fold cutoff of the expected expression level (Figure 3), similar to previously reported data (Fuchs et al., 2015). In contrast, with the SMARTer approach, expression levels for ~55% of the miRNAs fell within the cutoff range. Furthermore, whereas a 7.3 x 10⁴—fold range of expression levels was observed using adapter ligation, the SMARTer approach yielded a 1.5 x 10³—fold range, an improvement of greater than one order of magnitude. These results suggest that libraries generated with the SMARTer smRNA-Seq Kit for Illumina much more accurately portray the expression of miRNA species relative to libraries produced using adapter ligation.

Evaluating performance across total RNA sources and input amounts

To gauge the performance of the SMARTer approach, the SMARTer smRNA-Seq Kit for Illumina was used to generate sequencing libraries from 1 ng and 2 µg inputs of human total RNA obtained from brain, placenta, and spleen tissue. These three different tissues exhibit a range of small RNA (<200 nt) expression levels, varying from 2–13% as shown in the table below (Table 1). Consistent with its relatively high proportion of small RNAs (~13%), placental total RNA yielded the most robust libraries, with the lowest number of reads lost in trimming and the best overall mapping to the transcriptome of all three RNA sample types tested (mapping to GENCODE). In addition, libraries generated from placental total RNA demonstrated the most overlap between 1 ng and 2 µg input amounts in terms of which miRNAs were identified. Nevertheless, consistent sequencing metrics were also obtained for libraries generated from brain and spleen total RNA (Table 1 and Figure 4), indicating that the SMARTer smRNA-Seq Kit for Illumina generates reliable small RNA-seq data for RNA samples containing various proportions of small RNAs. Lastly, the data also demonstrate that the SMARTer approach is capable of capturing diverse small RNA species other than miRNAs, including but not limited to piRNAs, small nucleolar RNAs (snoRNAs), and small nuclear RNAs (snRNAs) (Table 1), while keeping representation of less desirable ribosomal RNAs (rRNAs) relatively low (typically between 10–20%).

Table 1. Evaluating the performance of the SMARTer approach across total RNA input types and amounts. Sequencing libraries were generated from 1 ng and 2 µg of human placenta, brain, and spleen total RNA using the SMARTer smRNA-Seq Kit for Illumina, and size selected using the BluePippin system for further enrichment of the miRNA fraction. Following trimming of 3′ A-tail sequences attributed to polyadenylation and 5′ sequences derived from template switching, reads shorter than 15 nt were eliminated, resulting in trimming of 15–38% of reads. Remaining reads were mapped either to GENCODE for overall mapping, or specific small RNA datasets, as indicated. Only miRNAs represented by at least five reads were included in count data for the number of miRNAs detected.

Measuring data reproducibility with the SMARTer approach

To specifically evaluate the reproducibility of the SMARTer smRNA-Seq Kit for Illumina, miRNA expression levels measured from different technical replicates were plotted and statistical analyses were performed (Figure 5). The results indicate a strong correlation in miRNA expression levels between replicates performed using 1 ng of human brain total RNA (Pearson coefficient = 0.99, Spearman coefficient = 0.86), indicating that data generated using the SMARTer smRNA-Seq Kit for Illumina is highly reproducible. Consistent with this result and data presented above, miRNA expression levels for 1 ng and 2 µg input amounts of human brain total RNA were also found to be highly correlated (Pearson = 0.94, Spearman = 0.87; (Figure 5).

The development of NGS technology has greatly accelerated the discovery and characterization of diverse small RNA species. However, the ability to accurately measure small RNA expression levels on a genome-wide scale remains elusive, largely due to biases imposed during sequencing library construction. Furthermore, study of small RNAs typically requires large amounts of starting material. To help researchers overcome these hurdles, we have developed the SMARTer smRNA-Seq Kit for Illumina, a new library-preparation kit that leverages RNA 3′ polyadenylation and SMART template-switching technology during cDNA synthesis. Through the application of these two techniques, our approach avoids biases associated with adapter ligation methods, and provides the sensitivity necessary to generate high quality sequencing libraries from as little as 1 ng of RNA.

Input material

To assess the accuracy of the SMARTer approach (Figure 3), sequencing libraries were generated from the miRXplore Universal Reference (Miltenyi Biotech Inc., Cat. No. 130-093-521) using a 1 ng input for the SMARTer approach, and a 100 ng input for the adapter-ligation approach. Starting material for other experiments (Table 1, Figures 4 and 5) consisted of the following commercially available preparations of total RNA: Human Brain Total RNA, Human Placenta Total RNA, and Human Spleen Total RNA (Thermo Fisher; Cat. Nos. AM7962, AM7950, and AM7970).

Library preparation

Sequencing libraries were generated as specified in the user manual for the SMARTer smRNA-Seq Kit for Illumina. Preparation of a sequencing library using an adapter ligation approach was performed using a small RNA-seq kit from another vendor, according to its accompanying protocol.

Library purification

As specified in the SMARTer smRNA-Seq Kit for Illumina user manual, following PCR amplification, all libraries were purified using the Macherey-Nagel NucleoSpin Gel and PCR Clean-Up kit (Takara Bio, sold as part of the SMARTer smRNA-Seq Kit for Illumina; also sold separately as Cat. Nos. 740609.50 and 740609.250).

Library size selection

Libraries included in the accuracy analysis (Figure 3) were not size-selected following PCR amplification. For all other libraries, post-PCR size selection was performed using the BluePippin Size Selection System (Sage Science, BLU0001) and 3% Agarose Gel Cassettes (Sage Science, BDF3010), as specified in the SMARTer smRNA-Seq Kit for Illumina user manual, but with a size selection range of 148–184 bp.

Library quantification and validation

Prior to validation and sequencing, all libraries were quantified using a Qubit Fluorometer (Thermo Fisher Scientific) and a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Cat. No. 32851). Validation was performed using an Agilent Bioanalyzer and the Agilent High Sensitivity DNA Kit (Agilent, Cat. No. 5067-4626).

Illumina sequencing

All libraries were sequenced on the Illumina MiSeq® platform using single end reads (50 bp), generating at least 1 million reads per library.

Data analysis

Sequencing reads for all libraries were trimmed and annotated using CLC Genomics Workbench 8.5.1 (Qiagen) and Small RNA Analysis tools, allowing no more than one mismatch during mapping. Overall mapping was performed against GENCODE (GRCH38). miRNA sequences were mapped to miRBase (release 21), while all other small RNA species were analyzed by mapping to the NONCODE v3.0 dataset. For analysis of the miRXplore libraries, reads were annotated using a reference file containing sequences present in the synthetic pool. Normalization was performed by determining, for each miRNA, the ratio between the observed number of reads and the predicted number of reads.

Berezikov, E., Cuppen, E., & Plasterk, R. H. Approaches to microRNA discovery. Nat. Genet. 38, S2–S7 (2006).

Chenchik, A., Zhu, Y.Y., Diatchenko, L., Li, R., Hill, J., & Siebert, P.D. Generation and use of high quality cDNA from small amounts of total RNA by SMART PCR. Gene Cloning and Analysis of RT-PCR. 305–319 (1998).

Choudhuri, S. Small noncoding RNAs: biogenesis, function, and emerging significance in toxicology. J. Biochem. Mol. Toxic. 24, 195–216 (2010).

Fuchs, R. T., Sun, Z., Zhuang, F., & Robb, G. B. Bias in ligation-based small RNA sequencing library construction is determined by adaptor and RNA structure. PloS One 10, e0126049 (2015) .

Ha, M., & Kim, V.N. Regulation of microRNA biogenesis. Nat. Rev. Mol. Cell Bio. 15, 509–524 (2014).

Hafner, M., Renwick, N., Brown, M., Mihailović, A., Holoch, D., Lin, C., Pena, J.T., Nusbaum, J.D., Morozov, P., Ludwig, J. and Ojo, T. RNA-ligase-dependent biases in miRNA representation in deep-sequenced small RNA cDNA libraries. RNA 17, 1697–1712 (2011).

Jayaprakash, A. D., Jabado, O., Brown, B. D., & Sachidanandam, R. Identification and remediation of biases in the activity of RNA ligases in small-RNA deep sequencing. Nucleic Acids Res. 39, e141 (2011).

Raabe, C. A., Tang, T. H., Brosius, J., & Rozhdestvensky, T. S. Biases in small RNA deep sequencing data. Nucleic Acids Res. 42, 1414–1426 (2014).

Weick, E.M., Sarkies, P., Silva, N., Chen, R.A., Moss, S.M., Cording, A.C., Ahringer, J., Martinez-Perez, E. & Miska, E.A. PRDE-1 is a nuclear factor essential for the biogenesis of Ruby motif-dependent piRNAs in C. elegans. Genes Dev. 28, 783–796 (2014).