SFPQ condensates keep neurons' two-megabase genes transcribing to the end

The scaffold is the interesting part. SFPQ does not build these compartments out of protein alone; it binds newly synthesizing RNA and uses the strands as structural material ^[2]. So the transcript being made is part of what holds together the compartment making it. Cells partition their interiors this way, without lipid membranes, through liquid-liquid phase separation ^[12].

Pairing the methods made sense: super-resolution microscopy shows where things sit, and proximity-dependent biotin labeling with mass spectrometry reports what is close to SFPQ inside a nucleus ^[3]. Inside the condensates the team found chromatin remodelers, transcriptional elongation factors and splicing regulators concentrated together, working on genes from over 100 kb to more than 2 Mb ^[4]. "These condensates act as a shared 'workspace' bringing together several processes required for exceptionally long genes to function properly," the authors wrote ^[9].

A two-million-base gene transcribed in a single day means polymerase covering roughly 1,390 bases a minute for 24 hours without detaching ^[13], and the source gives hours to days as the range for these genes ^[5]. The stated size range spans at least twentyfold ^[14]. Neurons sit at the long end by necessity, since synaptic connectivity depends on giant structural and signaling proteins ^[11].

I would slow down on the disease claim. Disrupting the condensates produced premature transcription arrest and splicing errors in critical synaptic genes, many of which appear on autism and ALS risk lists ^[16]. SFPQ and its partner FUS have long-standing genetic links to both conditions ^[8]. But an overlap between two gene sets that are both biased toward long, neuron-expressed genes is close to what you would expect before any mechanism is tested, and the Ehime University summary reports the overlap without a count of the shared genes or a statistical test of the enrichment ^[15].

In the cells studied, removing stable SFPQ condensates broke long-gene transcription and splicing, and expression fell sharply ^[6]. Whether a patient's variant does the same thing in a patient's neurons is a separate experiment.

Takeuchi's group proposes "long-gene transcriptopathies" as a category, and stabilizing nuclear architecture as a therapeutic aim ^[10]. The test that would move the hypothesis forward is straightforward to state: express an autism-associated variant of a long synaptic gene in neurons, then ask whether SFPQ condensates still form and whether the transcript reaches its far end.