Reply: PLD3 and spinocerebellar ataxia Free

Sir,

With the current advances in genetic technologies, we are now able to rapidly identify many rare variants throughout the genome. However, as we often lack insight into their biological consequences, it remains uncertain which variants are disease-causing. Geneticists therefore often perform gene co-expression studies and functional experiments in cell model systems, as well as in silico prediction studies, to look for supplementary evidence for variant ‘pathogenicity’. Here, pathogenicity refers to a functional variant that may lead to abnormal protein function. Determining pathogenicity remains challenging because the cellular function is not well established for most proteins encoded by the genes carrying the rare variants and/or their function may be cell-type specific. Moreover, the corresponding proteins are often overexpressed in cell model systems, and this may lead to misinterpretation of results. The letter from Gonzalez et al. (2018) addresses this issue.

Gonzalez et al. (2018) write that they are surprised by the suggested causative role of PLD3 for spinocerebellar ataxia (SCA) as, among other lines of argument that we discuss below, their cellular localization data for PLD3-WT is different than that reported by Nibbeling et al. (2017). They point out that untagged PLD3-WT is barely detected in the endoplasmatic reticulum (ER) and claim that the GFP-tagged PLD3 used by Nibbeling et al. is an ER-resident protein. However, as reported by Nibbeling et al., this is incorrect: the Pearson’s correlation coefficient of GFP-tagged PLD3-WT with the ER marker calnexin is 0.25, which indicates a weak positive correlation. Gonzalez et al. show a similar correlation using ER marker KDEL (Figure 1A in Gonzalez et al., 2018). This implies that GFP-tagged PLD3-WT is only marginally localized to the ER and a similarly weak correlation coefficient was observed for GFP-tagged PLD3-L308P. Also, given our GFP-tag, we are not able to discern PDL3’s luminal site from its N-terminal site. The differences in localization that Gonzalez et al. observe could therefore be masked. Importantly, in Nibbeling et al. we did not draw our conclusions with respect to the causality of PLD3 in SCA from our cellular localization experiment. Since the lysosomal localization of PLD3 was only recently discovered by Gonzalez and colleagues (Gonzalez et al., 2018), and was unknown at the time of our publication, we did not investigate this in-depth.

Furthermore, Gonzalez et al. are unable to demonstrate direct PLD activity for PLD3, and they observe no major changes in lipid species associated with PLD activity. The authors use this as an argument to uncouple PLD3 as causal gene underlying SCA46, a conclusion that we do not grasp. As Nibbeling et al. was able to show PLD activity of PLD3 in COS-7 cells, we wonder whether Gonzalez et al.’s experiment simply failed. It is also possible that PDL3’s activity is cell-type-specific, an idea recently reinforced by a study describing how the various PLD isoforms differentially regulate leucocyte response to acute lung injury (Abdulnour et al., 2018). We speculate that this might be true for other cell types as well.

Gonzalez et al. go a step further in challenging the interpretation that PLD3 mutations can lead to cerebellar ataxia by stating that the lowest PLD3 expression levels were observed in cerebellum compared to all other brain regions. This seems to be untrue as no significant changes are seen in PLD3 expression levels in the cerebellum compared to pons/medulla, midbrain, olfactory bulb, and the rest of the brain (Figure 2A in Gonzalez et al., 2018). However, even if true, this would not exclude PLD3 as the causative gene for SCA46. Several established SCA genes have been shown to have relatively low expression levels in the cerebellum including PPP2R2B, PDYN, and TGM6, and only 40% of SCA genes have highest expression in the cerebellum (Bettencourt et al., 2014; Nibbeling et al., 2017). We also wonder why Gonzalez et al. did not show any immunohistochemistry data on PLD3 expression in Purkinje cells, as these are the main cell type affected in SCA pathogenesis. It might be that PLD3 is more abundantly expressed in Purkinje cells than in other cell types of the cerebellum, which may explain why Gonzalez et al. find PLD3 ‘hardly’ expressed in total cerebellar extracts.

To make their final point that PLD3 cannot be the causative SCA46 gene, Gonzalez et al. generated PLD3 knockout (KO) mice. Complete loss of PLD3 did not cause apparent cerebellar atrophy or Purkinje cell loss at 20 months of age. We wonder why Gonzalez et al. did not study Purkinje cell climbing fibre innervation in more detail, as this has been shown to underlie other SCA types prior to the onset of Purkinje cell degeneration (Barnes et al., 2011; Fogel et al., 2015; Smeets et al., 2015; Smeets and Verbeek, 2016). Furthermore, Gonzalez et al.’s PLD3 KO mouse developed no motor dysfunction at 9 months of age, which is not very surprising given the absence of clear pathology at 20 months of age. Finally, the rotarod test they use is not the most sensitive assay to study ataxic features, an automated catwalk analysis investigating gait behaviour would have been more conclusive. Although Nibbeling et al. did observe a loss of PLD activity for PLD3-L308P, it is not surprising that complete loss of PLD3 does not cause SCA. The proposition by Gonzalez et al. that loss of PLD activity due to the p.Leu308Pro variant reflects PLD3 KO in SCA patients seems a simplistic conclusion given other evidence that the disease mechanisms of SCA can be complex. Other examples exist where the mutant allele exhibits a loss of function, but where loss of both alleles does not lead to disease, such as SCA19 (Niwa et al., 2008; Duarri et al., 2012). Moreover, redundancy of the other PLD isoforms may compensate for the loss of PLD3. Or an unknown gain of function might also contribute the pathogenesis as, undoubtedly, not all functions of PLD3 have yet been investigated. Gonzalez et al.’s finding that the p.Leu308Pro variant leads to defects in the exonuclease activity of PLD3 is a clear example of this.

In conclusion, our study demonstrated that PLD3 is the causative SCA46 gene using a combination of supporting experiments including (i) thorough genetic studies; (ii) functional support showing that the p.Leu308Pro variant reduced the PLD activity of PLD3 and is thus a variant with functional consequences; and (iii) gene co-expression analysis showing functional classification of PLD3 with other SCA genes. While we agree with Gonzalez et al. that the identification of additional patients carrying PLD3 mutations will further strengthen the role of PLD3 in SCA, we do not question the pathogenicity of the p.Leu308Pro variant in PLD3 causing SCA46. Nevertheless, we highly appreciate the efforts of colleagues to further understand the molecular underpinnings of the disease variants leading to SCA.

Data availability

Data sharing is not applicable to this article as no new data were created or analysed in this study.

Acknowledgement

We would like to thank Kate McIntyre for editing of the letter.

Funding

This work is funded by a Rosalind Franklin Fellowship from the University of Groningen to D.S.V. and a Graduate School of Medical Science scholarship from the University of Groningen to K.Y.M.

Competing interests

The authors report no competing interests.

References