Figure 6 - qPCR analysis for possible target glycosylation
genes of miR-30b OE (B1), miR-449a OE (G1) and mock cell pool.
Discussion
Along with our increased understanding of miRNAs and miRNA correlations,
it has become feasible to make at least rough predictions of the targets
of miRNAs. Furthermore, due to the ability of the miRNA to degrade or to
silence multiple mRNAs in a highly specific and potent way, makes miRNA
a highly interesting target for cell engineering; in particular of
complex traits such as N-glycosylation, which involves a large number of
enzymes working in concert. Hence, in this study we have tried to adopt
a composite prediction and experimental strategy to closer examine on
how miRNA regulating the N-glycan pathway in CHO cells
In this study, the strategy – in addition to sequence-based prediction
– was to use expression data of mRNA and miRNA to link miRNA with mRNA
based on correlation analysis. As we were interested in picking only
genes which are inhibited by miRNA, we have filtered for only those
mRNA-miRNA combinations that were negatively co-expressed. In
combination with a pathway analysis, to filter for N-glycan pathways
such as fucosylation, mannosylation, sialylation, galactosylation and
branching of N-glycan, we could focus on relevant genes. In the end,
this combination of approaches gave a list of likely miRNA-mRNA sets.
Furthermore, the experimental approach and analysis of mRNA and miRNA
sequencing has generated a lot of new knowledge on which miRNAs are
active in these cell lines, and provide a long list of hypotheses on
co-expressed genes. In total, we saw that four out of our ten miRNA
targets had an effect on N-glycosylation in the tested conditions,
showing that this field has an enormous potential for future
applications.
For instance, in our combined analysis, we saw that miR-128 should
theoretically combine with Alg9 and Man2a1, based on that mRNA
expression of Alg9 and Man2a1 was down-regulated and miR-128 was
up-regulated in EPO-producing cell lines as compared to parental cell
lines. Suggesting knock-out of miR-128 in EPO producing cell lines would
ideally increase the expression level of Alg9 and Man2a1, increasing
alpha 1,3 mannosyltransferase and mannosidase alpha class2 member1
activities. However, this was not the phenotype that was seen directly,
but instead, we saw a shift towards more mature (tri- and tetra-
antennary) N-glycans. This fitted well with the miRwalk predictions,
predicting binding to Mgat1 and Mgat4b (Glc-NAc transferases, involved
in increasing antennarity). A similar effect was found for miR-34c.
Here, Alg1 and B4galt2 were found in combination with miR-34c, where
expression levels supported an inverse correlation, and we noticed that
even though expression level of miR-34c was down regulated; it was
significantly expressed in EPO producing cell lines suggesting complete
functional knock-out of miR-34c to see increase in the activity of alpha
1,3 mannosyltransferase and beta1,4 galactosyltransferase. Furthermore,
Mgat1 and Mgat4 and St3gal1 (sialyltransferases) were predicted by
miRwalk to be targets (Table 1). In this cases as well, KD of the miRNA
increased the maturation. It thus seems - based on the prediction of the
functional targets - that increasing availability of early
N-glycosylation pathway enzymes improves maturation of the N-glycans.
This is very important for proteins such as EPO, as mature antennae and
sialylation is critical for serum half-life of
EPO(Tsuda et al. 1990).
Similarly, for antibodies, mannose residues can increase antibody
clearance rates
(Kaufman et
al. 1991; K. H. Lee et al. 2019; Fabian 2010; J. Yang et al. 2015).
Further, sialylation of native and recombinant EPO increases itsin vitro biological activity, possibly by increasing its affinity
to receptor on target
cells(Kaufman et al. 1991).
It was previously shown that making more sites available for adding
sialic acid by overexpressing B4gal1, also in tandem with
St3gal1(Kaufman et al.
1991; K. H. Lee et al. 2019). Terminal N-linked glycosylation is
directly related to N-linked sialylation. Thus, the prediction that
miR-128 and miR-34c inhibits early maturation enzymes and Mgat1 and
Mgat4, fits well with increased number of antennae, and will be
beneficial for certain recombinant protein production, e.g. EPO
production.
For the overexpression studies, only miR30b and miR-449a out of the
seven tested miRNAs gave a phenotype, again increased maturation of the
N-glycans. This was interesting, as these miRNAs did not have predicted
targets in the correlation analysis, but were predicted by miRwalk to
have a large number of gene targets in glycosylation.
In general, our study also shows that it is hard to predict the
phenotype of a miRNA perturbation solely based on sequence analysis,
even when supplemented with miRNA- and mRNA-seq. One reason for this may
be that miRNAs show as little complementarity as 2-8nt to their gene
targets, making them hard to identify using sequence alignment.
Furthermore previous
studies(Kaufman et
al. 1991; K. H. Lee et al. 2019; Fabian 2010) have shown that higher
complementarity with the mRNA with their targets degrade the transcripts
and lower complementarity leads to translational repression , thus
making both the prediction more complicated, and the regulation even
more complex. Even so, we managed to find a desirable phenotype in 40%
of our predicted targets. Due to limited capacity, we were only able to
test 10 targets, but the full data set (Supplementary Table 19) contains
even more interesting future targets, for examples, miRNAs with high
correlation with FUT8 fucosyltransferase, galactotransferases B4galT1-8,
sialylases St3gal1-6, and sialic acid transporters SLC35A1-A5.
Conclusion
We have adopted a bifurcated comparative approach using both
computational and experimental elements to identify miRNAs with an
impact on N-glycan maturation. In total, 656 miRNAs were predicted, ten
were experimentally characterized and four were found to have beneficial
phenotypes for N-glycosylation. The present study is one among few
studies aiming to assess the impact of overall miRNAs on N-glycan
pathway of various CHO cell lines. The data generated from miRNA- and
mRNA-seq is an asset for the scientific community as it provides the
valuable understanding of some of the useful expression targets of the
genes. Finally, identification of CHO-specific miRNAs presented in the
study might serve as a tool to improve cell line engineering effort to
increase the overall productive robustness of the cell lines.