Reply: The spliceosomal component RBP45d regulates pre-mRNA splicing and gene silencing for flowering time control in Arabidopsis

Dear Editor,

Two recent studies including one from our group (Chang et al., 2022) and another from Wang et al. (2022) reported the function of an RNA-binding protein RBP45d in flowering time control of Arabidopsis. Our data showed that RBP45d is a novel component of U1 small nuclear ribonucleoprotein (U1 snRNP; Chang et al., 2022). With strong evidence supporting the role of RBP45d in pre-mRNA splicing, we focused on its functions in regulating alternative splicing (AS) of flowering regulatory genes, especially FLOWERING LOCUS M (FLM), to explain the late-flowering phenotype in mutants of RBP45d and its partner PRP39a. We also found that the expression of another flowering repressor gene, FLC, was upregulated in the rbp45d mutant. We proposed that FLC regulation by RBP45d could be another unknown mechanism to control flowering time.

Wang et al. (2022) also reported findings with some experimental data similar to ours. In their study, RBP45d was identified from a genetic screen for components involved in RNA-directed DNA methylation (RdDM). Besides having similar observations that FLC was upregulated in the rbp45d mutant and FLM was differentially spliced, Wang et al. proposed that RBP45d functions in transgene silencing. We are excited to see the results from another group that supports our findings and were inspired by their hypothesis and surprised about the complex functions a splicing factor can have.

Tatjana Kleine, the corresponding author of Wang et al. (2022), recently questioned some of the results reported in our paper (Chang et al., 2022) in a letter to the editor (Kleine, 2022). In the following sections, we try to address the comments raised. Our aim is not to dismiss the argument of Kleine and colleagues but to emphasize what we believe is the most important fundamental of the findings for all scientists in the field.

According to Kleine (2022), the major concern is whether RBP45d directly binds to FLM mRNA. One concern mentioned was that the use of formaldehyde for the crosslinking process in the RNA-immunoprecipitation (RIP) assay may result in the cross-linking of RNA targets and proteins that are indirectly associated with RBP45d. Kleine compared the RIP experimental processes of the two studies and claimed that the washing conditions of Wang et al. (2022) were more stringent than ours, which might be the main reason for the different numbers of target RNAs identified in the two studies. We are not sure whether the fewer targets identified in Wang et al. (2022) versus our study can be explained by the difference in washing conditions in the experimental design. First, in our RIP experiments, we used 167 mM sodium chloride for the washing step. This is a concentration similar to that of the plant cell and should be able to provide enough ion strength to remove nonspecific, protein–protein and protein–RNA associations not crosslinked by formaldehyde. Second, the negative controls used in the two studies were different. Rather than using extracts from wild-type plants as in our RIP-seq experiment, Wang et al. used a transgenic line expressing “another unrelated protein-Myc fusion” as the negative control. Third, the RIP and library construction differed greatly at several major steps. For example, the plant extracts used for IP and the RNA pretreatment for library construction are different. In our experiments, we used nuclear extracts for IP, whereas Wang et al. used total extracts. Our immunoprecipitated RNAs were directly subjected to library construction, whereas Wang et al. used an rRNA depletion kit to remove rRNA. More importantly, the data analysis and filtering criteria the two groups used for RIP-seq were fundamentally different, which will obviously result in a distinct list of RNA targets. Therefore, we do not think it is appropriate to make any claim simply by comparing the number of targets identified between two experiments with distinct designs.

Kleine (2022) was also concerned by the identification of several rRNAs, chloroplast, and mitochondria-specific transcripts as targets of RBP45d, claiming that “RBP45d should be active only in the nucleus.” However, we have observed RBP45d to some extent in the cytoplasm, although mostly distributed in the nucleus. Similar observations were also reported by Wang et al. (2022), which suggests that the activity of RBP45d may not be restricted to the nucleus. We acknowledge that the nuclear extraction process might include contamination from the cytoplasm and other organelles, so we performed the experiments three times, and only those identified in at least two of the three biological repeats were considered as potential targets of RBP45d. The entire set of raw reads from the RIP-seq is also available for researchers who may have different viewpoints on the filtering criteria for the dataset.

Kleine (2022) further expressed concerns about the input data for our RIP-seq, claiming that the reads of all genes, including FLM, PRP39a, and RBP45d itself, had low coverage, although the libraries were sequenced to the same depth as the mock and IP sample. We believe this is actually a common and reasonable phenomenon. With the same amount of RNA being sequenced, the RNAs in input samples consisted of all RNAs existing in the nucleus, whereas the RNAs in mock and IP samples consisted of only those RNAs with affinity to the beads, antibody, or RBP45d-cMyc. Consequently, the read numbers of the input will be low and those of the RNAs enriched from the immunoprecipitation will be greater in the IP sample than in mock samples. It will be difficult to see as illustrated in Figure 1, A and B of the letter (Kleine, 2022).

Performing qRT-PCR on potential targets identified from RIP-seq results can be considered a “semi-quantitative” method to validate the association between the protein and its putative RNA targets. We have used this technique to validate the interaction between RBP45d and several of the putative targets of RBP45d, including U1 snRNA, U2 snRNA, and PRP39a as well as the FLM transcripts, all using U3 snRNA as an internal control. Note that the fold enrichment of FLM was comparable to that of U1 snRNA, which Wang et al. also identified as one of the strongest targets of RBP45d on RIP-seq. Therefore, at least in our experimental conditions, transcripts of FLM are the major targets of RBP45d.

Lastly, Kleine pointed out the identification of RBP45d by mutagenesis screening for factors involved in transgene silencing in two independent systems as a hint that RBP45d functions in other mechanisms to regulate flowering time. For clarification, studies by Kanno et al. (Kanno et al., 2017, 2020) were a genetic screen for mutants showing modified splicing of an intron-containing GFP reporter gene in Arabidopsis. Actually, the partner of RBP45d, PRP39a, was also identified in the screen by Kanno et al. (2017), which further supports the interaction between RBP45d and PRP39a with genetic evidence.

Although we agree with the conclusions in the letter and are excited to see some other functions of the RBP45d uncovered, we still think that the involvement of RBP45d, a component of U1 snRNP, in temperature-responsive AS regulation is no doubt the main activity of RBP45d in flowering time control. The temperature-triggered changes in AS pattern of FLM largely rely on the existence of a functional RBP45d-PRP39a complex in plants. Even Wang et al. (2022) reported that the AS of FLM was affected in the rbp45d-ko line in their study, which supports the involvement of RBP45d in regulating the AS of FLM. According to the results of our experiments, we have high confidence that the AS of FLM is regulated by a direct or at least an association with U1 snRNP as a complex, most possibly via the RNA binding activity of RBP45d.

Funding

Support during the preparation of this letter was provided by the Ministry of Science and Technology (MOST 109-2311-B-001-029-MY3) and Academia Sinica of Taiwan (to S.-L.T.).

Conflict of interest statement. None declared.