• Contact Us

Columbia University
Irving Medical Center
Neurological Institute

710 West 168th Street, 3rd floor
(212) 305-1818

TaubCONNECT Research Perspective:
March 2025

ANXA11 Biomolecular Condensates Facilitate Protein-Lipid Phase Coupling on Lysosomal Membranes

Peter H. St George-Hyslop, OC, MD, FRCPC, FMedSci, FRS

In collaboration with colleagues at the University of Cambridge, our latest study reveals an unexpected way membrane-less biomolecular condensates—like ribonucleoprotein (RNP) granules—can directly alter the physical properties of classical membrane-bound organelles such as lysosomes. We focused on the tethering protein annexin A11 (ANXA11), showing that it co=phase transitions with RNP granules. The ANXA11-RNP co-phase separation is driven by the disordered N-terminal low complexity domain (LCD) of ANXA11. We show that condensation of the ANXA11-RNP granule biomolecular condensate (BMC) can induce a matching phase shift in lipids in the lysosomal membrane. As the ANXA11-RNP BMC condenses, it transforms the lipids in the lysosome’s membrane from a more fluid, “oily” state to a firmer, “buttery” one. This stiffening appears to give the complex the mechanical resilience it needs to move through the densely packed axoplasm.

Figure 7. Protein-lipid phase coupling in the ANXA11-lysosome ensemble. The ARD of ANXA11 mediates binding to lysosomes in a Ca2+-dependent manner and causes a phase transition in lysosomal membrane lipids into a more ordered state. Condensation of the ANXA11 LCD can then act to tune the magnitude of this lipid phase transition in a coupled manner. ANXA11 interacting partners ALG2 and CALC either increase (ALG2) or decrease (CALC) ANXA11 LCD-based condensation to regulate phase coupled effects on lysosomal membrane lipids.

As recently reported in Nature Communications, we found that lipid ordering in lysosomal membranes increases in step with ANXA11 condensation. Remarkably, this effect persisted even when ANXA11’s native membrane-binding region was replaced with a chemical tag, showing that ANXA11 condensation alone can drive concomitant changes in membrane properties. We call this phenomenon “phase coupling”—where the state of a protein directly influences the physical state of its attached membrane – stiffening its nanomechanical properties. While similar effects have been observed at the plasma membrane in immune cells, we show BMC interaction with membranes has a much wider impact on cell biology, also occuring on intracellular membranes. We believe this stiffening helps the ANXA11-RNP-lysosome complex endure mechanical shear stress during axonal transport, particularly in neurons with long axons.

We also identified two regulatory proteins—ALG2 and CALC—that modulate this phase coupling. ALG2 enhances ANXA11 condensation and membrane stiffening, potentially priming the complex for movement. CALC does the opposite, reducing condensation and softening the membrane, which may aid in cargo release. These regulators also influence the complex’s ability to bind RNP granules, suggesting a finely tuned system where protein and membrane phase transitions coordinate intracellular transport.

Overall, our findings suggest a highly adaptable system where phase transitions in proteins in BMCs and lipids in cellular membranes work together – in this instance to control cargo movement inside cells. This work connects two major areas of cell biology—membrane-less and membrane-bound compartments—and could offer new insight into diseases like ALS. Indeed, several proteins in RNP granules (e.g. FUS, TDP43) and ANXA11 are the sites of ALS-causing mutations.

Peter H. St George-Hyslop, OC, MD, FRCPC, FMedSci, FRS
Belle and Murray Nathan Professor of Neurology (in the Taub Institute)
Co-Director, The Carol and Gene Ludwig Center for Research on Neurodegeneration
ps2764@cumc.columbia.edu

Local Genetic Covariance Analysis with Lipid Traits Identifies Novel Loci for Early-Onset Alzheimer's Disease

Nicholas Ray, PhD		Christiane Reitz, MD, PhD

Early-onset Alzheimer’s disease (EOAD), which begins before age 65, accounts for about 10% of Alzheimer’s cases. While some cases are linked to mutations in APP, PSEN1, and PSEN2, most remain genetically unexplained. Building on previous work, our team—in collaboration with the laboratories of Drs. Caghan Kizil, Philip L. De Jager, and colleagues—explored whether lipid metabolism contributes to EOAD risk. Using genetic covariance analysis across five lipid traits—total cholesterol, HDL-C, LDL-C, non-HDL-C, and triglycerides—we identified three genomic regions with shared genetic architecture and five overlapping loci. These included known AD risk genes (APOE, TREM2, MS4A4E) and novel candidates (LILRA5, LRRC25).

To prioritize genes in the three regions identified by the genetic covariance analysis, we created composite scores for each gene. These scores combined data from multiple sources: gene-based analysis, AD risk scores from Agora (calculated using various multi-omic data, including GWAS, eQTL, transcriptomic, and proteomic data), MetaBrain eQTL data from the cortex and hippocampus, eQTL colocalization analyses across 61 GTEx datasets, ROSMAP brain methylation data, and single-cell RNA sequencing data from both humans and zebrafish. This novel prioritization approach identified ANKDD1B, CUZD1, and MS4A6A as key genes of interest.

Figure 1. Flowchart showing input data (EOAD and Lipids GWAS) used in the primary analyses (local genetic covariance analysis using SUPERGNOVA).

ANKDD1B encodes the ankyrin repeat and death domain containing 1 B protein and is linked to both dyslipidemia and diabetes. CUZD1 encodes a protein located in secretory granules in the pancreas that plays a role in lipid metabolism. This gene also contributes to the zymogen activation pathway, which is associated with AD. MS4A6 is a well-known AD-risk that encodes a member of the membrane-spanning 4A gene family and may also contribute to atherosclerosis. Published in PLOS Genetics, our findings provide evidence of shared genetic pathways between EOAD and lipid regulation, suggesting new avenues for understanding disease mechanisms and identifying therapeutic targets. Further work is needed to confirm causal variants and improve genetic risk prediction. Ongoing work includes the exploration of these genes using whole-genome-sequencing data from the Alzheimer's Disease Sequencing Project.

Nicholas Ray, PhD
Postdoctoral Research Scientist in the Gertrude H. Sergievsky Center
nrr2132@cumc.columbia.edu

Christiane Reitz, MD, PhD
Professor of Neurology and Epidemiology (in the Taub Institute and the Gertrude H. Sergievsky Center)
cr2101@cumc.columbia.edu

The Association of Multilingualism with Diverse Language Families and Cognition Among Adults with and Without Education in India

Miguel Arce Rentería, PhD

Most studies on multilingualism and cognitive aging treat multilingual adults as a single group when in fact they demonstrate substantial within-group variability. A key factor in which multilinguals differ is on which languages they use. Prior studies combined different language pairs within a multilingual group regardless of the linguistic similarity between the languages. The demands of cross-language interference are a potential mechanism that may strengthen cognitive control among multilinguals, suggesting that more similar languages may be more prone to interference and thus provide more frequent opportunities to strengthen cognitive control. In a new study led by me, with the help of postdoctoral fellow Dr. Iris Strangmann and collaborators from the University of Southern California and University of Michigan, we examined how these linguistic differences relate to cognitive function in older multilingual adults with and without formal education.

Published recently in Neuropsychology, our study used data from the Longitudinal Aging Study in India—Diagnostic Assessment of Dementia (LASI-DAD), a nationally representative sample of 4,088 Indian adults aged 60 and older, speaking 40 languages and dialects (54% without formal schooling). Participants were categorized based on whether their languages belonged to the same or different language families. This variety of languages was represented across five Indian language families: Indo-Aryan, Dravidian, Austro-Asiatic, Tibeto-Burmese, and Andamanese. In addition to these Indian languages, non-Indian language families which have been increasing in prevalence in India such as Indo-European (i.e., English) and Japonic (i.e., Japanese), among others, were also included.

Figure 1. Association of multilingual status and cognitive functioning among participants with education and without education. Note: All models control for: Age, sex, years of education, education, urbanicity, BMI, hypertension, diabetes, heart disease, smoking status, hearing loss, consumption quartile, and childhood SES (parental education).

After adjusting for covariates in the full and propensity-score matched samples, multilingual participants with formal education outperformed monolinguals across cognitive domains, regardless of language similarity. However, among those without formal schooling, only multilinguals speaking related languages showed a cognitive advantage. Our findings demonstrate potential for examining the relationship between multilingualism and cognition in large population-based cohort studies. Future measurement of the various attributes of multilingualism, such as language proficiency and frequency of use, will help us further explore how managing linguistic interference and exposure to similar languages enhances late-life cognition. For more information, listen to my interview on the Neuropsychology Meet the Authors podcast on Spotify or wherever you get your podcasts.

Miguel Arce Rentería, PhD
Assistant Professor of Neuropsychology (in Neurology)
ma3347@cumc.columbia.edu

Axonal Transport of CHMP2b Is Regulated by Kinesin-Binding Protein and Disrupted by CHMP2bintron5

Konner Kirwan, PhD		Clarissa Waites, PhD

Maintaining the neuronal proteome is essential for nervous system health. Indeed, many studies show that disruption of cellular pathways for protein synthesis, trafficking, and degradation lead to synaptic dysfunction and ultimately neurodegeneration. One such pathway is the ESCRT (endosomal sorting complex required for transport), a series of molecular complexes that sort proteins destined for degradation into multivesicular bodies (MVBs) for delivery to lysosomes. Mutations in the ESCRT protein CHMP2b, responsible for the final step of MVB formation, impair neuronal protein degradation and cause frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS).

Previous studies in mouse models expressing the most common of these mutations, CHMP2bintron5, show that this mutant also promotes early deficits in synaptic function that precede neurodegeneration. To understand how these synaptic defects arise, we performed live imaging experiments to a) characterize the axonal transport and synaptic localization of CHMP2b, and b) determine how these features are disrupted by the CHMP2bintron5 mutation. As recently reported in Life Science Alliance, we found that CHMP2b undergoes vesicular transport in axons, and that both its trafficking and recruitment to synapses are positively regulated by neuronal activity, consistent with the need for ESCRT protein delivery to synapses to catalyze MVB formation facilitating the use-dependent turnover of synaptic proteins. In contrast, our imaging studies of CHMP2bintron5 revealed that this mutant exhibits very little directional transport or synaptic localization under basal conditions, and that these features are not regulated by neuronal activity. Instead, CHMP2bintron5 transport vesicles exhibit oscillatory behavior reminiscent of a ‘tug-of-war’ between kinesin and dynein motor proteins.

Graphical Abstract.

In the second part of the study, we investigated why this phenotype occurs. We demonstrated that it is due to the reduced binding of CHMP2bintron5 to an inhibitor of kinesin-mediated axonal transport, kinesin family binding protein (KBP), leading to dysregulation of the mutant’s trafficking and recruitment to synapses. Additionally, we found that CHMP2bintron5 acts as a dominant-negative to prevent wild-type CHMP2b from properly localizing to synapses. The cumulative result of this mutation is thus a lack of CHMP2b delivery to synapses to facilitate the formation of MVBs for degradation of synaptic proteins, leading to their accumulation, altered synaptic vesicle cycling, and deficient neurotransmitter release.

These findings represent the first characterization of CHMP2b axonal transport dynamics, and implicate KBP as a key regulator of CHMP2b activity-dependent transport and synaptic localization. Moreover, our work provides novel mechanistic insight into how CHMP2bintron5 impairs the axonal transport and synaptic recruitment of CHMP2b, contributing to our understanding of how this ALS/FTD-causative mutant induces synaptic dysfunction.

Clarissa Waites, PhD
Associate Professor of Pathology and Cell Biology and Neuroscience
cw2622@cumc.columbia.edu