PREreview of Stochastic splicing and deterministic inclusion of variable exons promote diversification of Down Syndrome Cell Adhesion Molecule expression

Lassota and colleagues ask whether the selection of Dscam variable exons is really the stochastic process the field has assumed. They build a dual-amplification splicing reporter for the Drosophila exon 9 cluster, in which a single-nucleotide deletion places one variant in frame with LexA, so that inclusion of that variant drives a LexAop reporter through a GAL4 cascade. Applied to third instar tissues, the reporters give patterned rather than random output: segmentally reiterated rows of cells in the ventral nerve cord, enrichment in the mushroom body neuroblast region for exon 4, and repeated labelling of the same positional photoreceptor across ommatidia. Against this, the authors find that reporter expression is markedly heterogeneous between salivary gland cells even when all 33 variants are in frame, and that the terminal intron of the cluster is retained more in salivary glands than in the VNC. In honey bees, FISH against variable exons 4.5, 4.7, 10.10 and 10.15 shows broad uniform signal in newly emerged workers and compartmentalised signal in foragers, with patterns differing between the two mushroom bodies of the same brain, and a 2-phenylethanol exposure paradigm reproduces the shift within 24 to 48 hours.

The advance is conceptual as much as technical. The field has treated “stochastic” as a single property of exon choice. This work separates it into two: whether the cluster is spliced through productively, which varies between genotypically identical cells, and which variant is selected given that it is, which appears biased by lineage, cell type and compartment. The all-in-frame control is the key piece of evidence, since a dark cell there cannot be explained by a different variant having been chosen, and exon 9 variants preserve frame so skipping is also excluded. The bee data then add a layer the fly work cannot reach, showing that the spatial organisation of inclusion is not fixed at differentiation but is rewritten by sensory experience over a day or two. Together these results give a more tractable model than “each neuron rolls the dice,” and the reporter itself will be useful to others working on low-abundance isoforms.

Major issues

Absence of reporter signal is read throughout as absence of productive splicing, but the detection threshold of a two-step amplification system is never characterised. This is most consequential for the optic lobe claim. The authors concede the general point themselves when they attribute the larval to pupal photoreceptor difference to amplification delay. The qRT-PCR assay in Figure 2X–Y is the obvious fix: run it on dissected optic lobe against central brain, and the claim stops depending on a negative image.

The HA and FLAG logic does not work as drawn. In Figure 1B, FLAG sits downstream of the 2A peptide and in frame with LexA, so FLAG positivity is itself conditional on productive splicing and cannot serve as an independent check that the transgene is expressed. “All FLAG-positive cells expressed the reporter” is close to circular. The eye disc result then describes the opposite relationship, with FLAG ubiquitous and HA tracking the reporter, which is inconsistent both with the salivary glands and with the construct.

Figure 2W compares variance between two different reporter systems. elav>H2B::YFP involves one transcriptional step; the exon 9 reporter involves two, with a different fluorophore and different protein stability. These will differ in variance for reasons unrelated to splicing. The appropriate comparison is a UAS-LexA construct lacking the variable cluster, read out through the same LexAop line. If that line does not exist, the argument should rest on the qRT-PCR, which is not vulnerable to this.

The NMJ experiment lacks its genetic controls. Only wild type is shown. Driver-alone and RNAi-alone heterozygotes on RU486 are needed, knockdown efficiency should be documented at the two day timepoint, and bouton counts should be normalised to muscle area. There is also a flat contradiction in the text: line 258 reports a significant reduction in bouton number, while the heading, the figure, the legend and the abstract all report an increase. This inverts the conclusion and needs fixing.

Age and experience are confounded in Figures 5 and 6, since naïve bees are dissected within 24 h of emergence and foragers are weeks older. The authors half acknowledge this and the odour experiment does most of the work of separating them, but the wording in those sections should reflect it. Relatedly, “compartmentalised” is scored by eye from representative images. Defining the compartments anatomically, extracting intensity profiles normalised to the constant region probe, and having a blinded scorer classify the images would convert an impression into a measurement without much additional work.

One probe control is missing. The cross-species negative and the complexin comparison are both good, but the real risk with variable exons is cross-hybridisation between paralogous variants of the same cluster, which nothing currently addresses. Showing that two variant probes give non-overlapping patterns in the same brain would settle it.

Minor issues

• The bee exon 10 cluster and the fly exon 9 cluster are orthologous; say so at first mention, because readers will otherwise read it as an error.

• Methods list a probe against exon 10.2 (line 547) that never appears in the results. Presumably 10.15.

• Figure 2W marks p = 0.0059 with two asterisks defined as p ≤ 0.001.

• Supplementary Figure 4, the three schematic models, frames the entire paper and would do more work in the main text.

• For the salivary gland comparison, give n for both nuclei and glands, and state whether nuclei were nested within glands in the analysis.

• The claim that reduced Dscam in fly motor neurons is “consistent with enhanced memory” links two species and two paradigms fairly loosely. Flag it as speculation.

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.