Lipoprotein(a), abbreviated Lp(a) and often pronounced simply as L-P-little-a, is a low-density lipoprotein particle to which an unusual protein called apolipoprotein(a) is covalently attached. Its concentration in blood is overwhelmingly dictated by the LPA gene, whose size polymorphisms vary dramatically across ethnic groups, and it has emerged over the past decade as one of the most powerful independent genetic risk factors for atherosclerotic cardiovascular disease, aortic valve stenosis and, in some studies, heart failure. Unlike LDL cholesterol, which responds robustly to diet, exercise and most lipid-lowering drugs, Lp(a) has historically resisted every lifestyle intervention thrown at it, fuelling the long-standing assumption that a single lifetime measurement is sufficient for risk stratification.

That assumption is precisely what the new study set out to test. A team of cardiologists and data scientists led by Ming-Lung Tsai of New Taipei Municipal TuCheng Hospital and Dong-Yi Chen of Linkou Chang Gung Memorial Hospital mined the Chang Gung Research Database, one of Taiwan’s largest multi-hospital electronic health record systems, for adults who had at least two Lp(a) measurements taken at least one year apart between January 2004 and December 2019. The retrospective cohort design allowed the researchers to observe how the biomarker drifted, or failed to drift, across a mean follow-up of 5.2 years, with some individuals contributing data across nearly the entire fifteen-year window.

The headline result is one of striking stability. Among the 17,363 participants, who had a mean age of 42.1 years and were 73.3 percent male, the median Lp(a) concentration actually declined slightly over follow-up, falling from 9.3 milligrams per deciliter at baseline to 7.4 milligrams per deciliter at the final measurement. Only 479 participants, a mere 2.8 percent of the cohort, showed an increase of 10 milligrams per deciliter or more, the threshold the investigators pre-specified as a clinically meaningful elevation. For the overwhelming majority, the genetic determinism model of Lp(a) held firm: what was measured in the first blood draw was, within analytical noise, what was measured years later.

But the small minority who did cross the 10-milligram threshold was not random, and this is where the study earns its clinical relevance. Using multivariable logistic regression, the team identified five independent predictors of a substantial Lp(a) rise: a higher baseline Lp(a) concentration, initiation of statin therapy between measurements, a diagnosis of diabetes mellitus, female sex, and higher triglyceride levels. Each of these factors points to a different biological thread, and together they sketch a picture of Lp(a) as a biomarker that, while genetically anchored, can be nudged by metabolic state, medication and perhaps hormonal influences.

The statin finding is likely to draw the most attention, and the most debate. Statins are the workhorse of cardiovascular prevention, yet their relationship with Lp(a) has been contentious for years, with some trials reporting modest increases in the particle among treated patients. The new cohort cannot disentangle whether statin-associated Lp(a) elevation reflects a direct pharmacological effect, a marker of worsening underlying atherosclerosis that prompted statin prescription, or some combination of both, and the authors are careful not to overinterpret the association. What it does suggest is that clinicians who start a patient on a statin and later observe a rising Lp(a) should not automatically assume the measurement is erroneous or that the drug is to blame without considering the broader clinical context.

The diabetes and triglyceride associations are equally intriguing from a mechanistic standpoint. Diabetes is known to influence several apolipoprotein-containing particles, and insulin resistance reshapes hepatic lipid handling in ways that could plausibly alter the assembly or clearance of Lp(a), which is produced almost exclusively by the liver. Higher triglycerides, a hallmark of metabolic syndrome, may similarly mark a hepatic environment in which Lp(a) production or catabolism shifts. The female sex signal, meanwhile, echoes prior observations that Lp(a) concentrations can rise after menopause, hinting at estrogen-related regulation, although the relatively young cohort, with a mean age of just over 42, means the study captured only the early edge of that transition.

For a field racing toward Lp(a)-lowering therapies, these longitudinal data arrive at a propitious moment. RNA-targeted drugs such as pelacarsen, olpasiran and lepodisiran are in late-stage trials and promise, for the first time, to suppress Lp(a) by 80 percent or more. Trial designers and clinicians alike have assumed that a single baseline measurement suffices to identify candidates and to judge treatment response. The new study broadly supports that assumption, since 97 percent of participants remained below the meaningful-change threshold, but it also flags the minority in whom a second measurement could change the risk calculus. The authors conclude that repeat Lp(a) testing may be warranted in selected high-risk individuals, particularly those with elevated baseline values, new diabetes, or newly initiated statin therapy, to sharpen cardiovascular risk stratification.

The study also fills an important geographic gap. Most longitudinal Lp(a) data come from European and North American cohorts, where median concentrations and LPA isoform distributions differ substantially from those in East Asian populations. The Taiwanese cohort’s median baseline value of 9.3 milligrams per deciliter sits well below the roughly 30 to 50 milligrams per deciliter often reported in populations of European descent, a disparity rooted in the frequency of small apolipoprotein(a) isoforms. Demonstrating that Lp(a) stability holds in an Asian population strengthens the case that lifetime genetic determination of this biomarker is a universal feature of human biology, not an artifact of any one ancestry.

Caveats remain, as they do with any retrospective database study. The cohort was predominantly male and middle-aged, the assay methodology was consistent within the health system but may not translate directly to other platforms, and unmeasured factors such as kidney function, thyroid status and inflammatory disease, all of which are known to modulate Lp(a), could not be fully adjusted. The work was funded by Novartis Pharmaceuticals, though the funder had no role in data acquisition, analysis or interpretation, and the authors declare no competing interests. Even with these limitations, the message is clear and reassuring in equal measure: for most people, one well-timed Lp(a) measurement tells you nearly everything you need to know, but for a definable minority, the story written in their genes can still acquire new chapters, and a repeat blood test may be the only way to read them.

Subject of Research: Long-term variability and clinical determinants of lipoprotein(a) concentrations

Article Title: Longitudinal patterns and determinants of lipoprotein(a) variability: a 15-year multi-center study

Article References: Tsai, M.-L., Chen, C.-C., Hsieh, M.-J., Chang, C.-Y., Chen, S.-W., Chen, T.-H., Chu, P.-H., Hsieh, I.-C., Hung, K.-C., Wen, M.-S., & Chen, D.-Y. (2026). Longitudinal patterns and determinants of lipoprotein(a) variability: a 15-year multi-center study. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02631-1

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02631-1

Keywords: lipoprotein(a), cardiovascular risk, statins, diabetes mellitus, triglycerides, biomarkers, risk stratification, longitudinal cohort study, dyslipidemia, Taiwan, LDL cholesterol, genetic risk factors

The research, led by Bernard Mulvey, Heena R. Divecha, and colleagues at the Lieber Institute for Brain Development working with collaborators at Johns Hopkins and the University of Cambridge, applied spatially resolved transcriptomics, a technique that maps gene expression across intact tissue sections while preserving information about where each gene is active, to postmortem locus coeruleus tissue from 33 neurotypical middle-aged human brain donors. Crucially, the donors were stratified by major Alzheimer’s disease risk factors, including biological sex, African or European genetic ancestry, and apolipoprotein E haplotype, specifically carriers of the APOE E4 allele, which raises disease risk, versus carriers of the APOE E2 allele, which appears protective. This design allowed the investigators to ask a subtle but important question: do the genes expressed in and around the locus coeruleus already differ, in healthy middle age, depending on which Alzheimer’s risk variants a person carries?

The answer, in several respects, is yes, and the details are illuminating. When the researchers compared gene expression patterns across APOE haplotypes, they found reduced expression of astrocytic genes, meaning genes characteristically active in star-shaped support cells called astrocytes, in the tissue immediately surrounding locus coeruleus neurons in E4 carriers. Astrocytes are far more than passive glue; they regulate synapse formation, supply neurons with cholesterol and metabolic support, control local blood flow, buffer neurotransmitters, and respond to norepinephrine signaling from the locus coeruleus itself. A diminishment of astrocytic gene expression near these noradrenergic neurons suggests that E4 carriers may have subtly weakened local support infrastructure around one of the brain’s most Alzheimer’s-vulnerable cell populations, potentially lowering the threshold at which tau pathology or other stresses become damaging.

The study also uncovered ancestry-specific differences in locus coeruleus gene expression, a finding with real public health significance given well-documented disparities in dementia incidence and outcomes across populations. Follow-up analyses using in situ sequencing at single-cell resolution demonstrated that the APOE-related differences in regional gene expression were partly attributable to astrocytes themselves, and that the haplotype effects were more pronounced in donors of European ancestry. This interaction between genetic ancestry and APOE haplotype echoes earlier work showing that the relationship between APOE E4 and Alzheimer’s pathology differs across admixed populations, and it underscores the importance of including ancestrally diverse cohorts in neuroscience research rather than extrapolating from studies of predominantly European-ancestry subjects.

Perhaps the most visually and conceptually striking component of the study concerns neuromelanin, the dark pigment that gives the locus coeruleus its name, which means “blue spot” in Latin. Neuromelanin accumulates in certain catecholaminergic neurons, including the noradrenergic neurons of the locus coeruleus and the dopaminergic neurons of the substantia nigra, as a byproduct of oxidative metabolism of neurotransmitters. It is sequestered within specialized autolysosomal organelles that bind potentially toxic metals such as iron, and its loss from these neurons is a hallmark of both Alzheimer’s and Parkinson’s disease. Neuromelanin-sensitive magnetic resonance imaging has emerged as a promising biomarker, with reduced pigment signal in the locus coeruleus predicting clinical severity and future progression in Alzheimer’s patients. What has remained unclear is the precise molecular relationship between a neuron’s pigment content and its gene expression program.

To address this, the team quantified neuromelanin content directly in the tissue and related it to spatial gene expression. They found that higher APOE gene expression correlated with reduced neuromelanin, and that genes whose expression tracked with local pigment levels were enriched for aging-related biological pathways. Taking the analysis to its finest resolution, the investigators used in situ sequencing data to examine individual locus coeruleus neurons, measuring neuromelanin in each cell and validating its associations with the expression of APOE itself, genes involved in norepinephrine metabolism, and components of the autophagy machinery, the cellular recycling system that clears damaged proteins and organelles. This constellation of pigment-linked genes paints a coherent picture: neuromelanin content reflects, at the single-cell level, the interplay of neurotransmitter handling, lipid biology, and protein quality control, all processes implicated in neurodegeneration.

The significance of these findings is best appreciated against the backdrop of what is already known about the locus coeruleus in Alzheimer’s disease. Postmortem studies stretching back decades, including the classic staging work of Heiko Braak and colleagues, established that phosphorylated tau appears in locus coeruleus neurons exceptionally early, often in individuals who died with no cognitive impairment. Autopsy studies have documented substantial neuronal loss in this nucleus in mild cognitive impairment and early Alzheimer’s disease, and neuroimaging research has repeatedly linked reduced locus coeruleus integrity to tau burden, memory decline, neuropsychiatric symptoms, and sleep disruption. Animal work has shown, conversely, that noradrenergic depletion exaggerates the inflammatory response to amyloid-beta pathology, while pharmacological enhancement of norepinephrine signaling can suppress neuroinflammation. The locus coeruleus, in other words, is not merely a passive victim of Alzheimer’s pathology; its noradrenergic output actively modulates the brain’s response to it.

The new study adds two crucial molecular threads to this narrative. First, it demonstrates that Alzheimer’s genetic risk acts on the locus coeruleus ecosystem, not just on its neurons in isolation. The astrocytic deficit observed near E4-carrier neurons is particularly intriguing in light of recent evidence that norepinephrine signals through astrocytes to modulate synaptic function, and that astrocytes undergo stereotyped transcriptomic changes across the spatiotemporal progression of Alzheimer’s disease. If E4 carriers begin adulthood with less robust astrocytic support around their noradrenergic neurons, this could help explain why these neurons succumb early, and why E4 homozygosity has recently been characterized as a distinct, nearly deterministic genetic form of Alzheimer’s disease. Second, by tying neuromelanin, a biomarker measurable in living patients with specialized MRI, to concrete gene expression programs involving autophagy and catecholamine metabolism, the study strengthens the mechanistic bridge between what clinicians can image and what is happening molecularly inside vulnerable neurons.

Technically, the study represents a tour de force of modern spatial genomics applied to a notoriously difficult brain region. The locus coeruleus is small, deeply located in the dorsal pons, and densely pigmented, making it easy to miss in standard postmortem dissection and challenging to analyze with conventional bulk methods that average away spatial relationships. The team’s spatially resolved transcriptomics data captured tissue architecture at spot-level resolution, while their in situ sequencing pushed the analysis to individual cells, using the pigment itself as an additional segmentation cue to define neuromelanin-rich cellular compartments. The authors have deposited all data in public repositories and built interactive web portals, using visualization tools such as Samui and spatialLIBD, allowing any researcher to explore the spatial domain assignments, gene expression maps, and stained tissue sections, an unusually transparent approach that should accelerate follow-up work across the field.

The implications reach toward prevention and early detection. Because the donors in this study were neurotypical and middle-aged, the observed molecular differences, diminished astrocytic gene expression in E4 carriers, ancestry-dependent expression patterns, pigment-linked autophagy and norepinephrine genes, represent the brain’s baseline state before overt pathology, a snapshot of vulnerability rather than of damage. Interventions aimed at supporting astrocytic function, enhancing autophagic clearance of tau, or modulating noradrenergic tone might therefore be most effective when targeted at these early molecular states rather than at established dementia. Meanwhile, the confirmation that APOE expression itself inversely tracks neuromelanin suggests that pigment imaging could serve not only as a marker of neuronal loss but as a window into the lipid-handling pathways that connect APOE biology to neurodegeneration. As the authors conclude, Alzheimer’s risk factors appear to modulate locus coeruleus vulnerability through molecular processes intrinsic to both the noradrenergic neurons and their astrocytic partners, a reminder that the neurons that wake us, focus us, and help us remember may hold some of the earliest clues to when and how Alzheimer’s disease takes hold.

Multiple sclerosis has long been recognized as a disease of extraordinary variability. Some patients lose the ability to walk within a decade of diagnosis, while others remain ambulant for decades. Under the microscope, the variability is just as striking: some brains are riddled with actively inflamed demyelinated lesions, while others show mostly burned-out, inactive scars. For years, this heterogeneity has frustrated efforts to draw clean lines between what pathologists see in post-mortem tissue and what clinicians observed in the patient while alive. The new research argues that the missing link may lie in donor-specific pathological features—persistent, recurring characteristics of an individual’s inflammatory and reparative biology that color everything the disease does in that person’s brain.

The research team, led by Lukas Lütje and J. Q. Alida Chen along with colleagues including Jörg Hamann, Joost Smolders, Inge Huitinga, and Aletta M. R. van den Bosch, systematically assessed brain tissue from 287 donors with confirmed MS pathology, collected between 1990 and 2021. The scale of the sampling effort is remarkable: on average, nearly 22 tissue blocks and more than 35 lesions were dissected per donor, taken from standardized locations in the brainstem, with lesions identified macroscopically or guided by post-mortem magnetic resonance imaging. Each tissue block was formalin-fixed and paraffin-embedded, then subjected to double immunostaining for human leukocyte antigen proteins (HLA-DP-DR-DQ) and proteolipid protein (PLP)—a technique that simultaneously reveals activated microglia and macrophages, the inflammatory workhorses of the disease, and the myelin they attack. Iron deposition was visualized using a DAB-enhanced Turnbull Blue staining method, adding another layer of pathological detail.

The four donor-specific features at the heart of the study each capture a different dimension of MS biology. Perivascular cuffs are exaggerated accumulations of lymphocytes—both B and T cells—crowding into the perivascular spaces of the brain, visible when at least one such space shows the hallmark buildup in any examined tissue block. Microglial nodules are small clusters of at least four ramified microglia, the brain’s resident immune cells, appearing in tissue that otherwise looks normal—the so-called normal-appearing white matter. Broad rim lesions, or BRLs, are a particularly dramatic subtype of chronic active lesion, defined by a hypercellular HLA-positive rim at least one millimeter wide surrounding a hypocellular, fully demyelinated core. And remyelination efficiency divides donors into efficient remyelinating donors and poorly remyelinating donors, based on the proportion of lesions showing repair, with a threshold set at a remyelinated lesion proportion of at least 0.27.

When the researchers crossed these pathological features with the donors’ genetic profiles, striking associations emerged. Carriers of the HLA-DRB115:01 allele—by far the strongest known genetic risk factor for multiple sclerosis, tagged by the SNP rs3135388—showed a higher prevalence of both perivascular cuffs and microglial nodules. This finding is biologically provocative: HLA-DRB115:01 encodes a molecule that presents antigens to T cells, and its association with lymphocytic cuffing suggests that the classic genetic gateway to MS risk may shape the very character of the inflammatory infiltrates in the brain decades later. Meanwhile, carriers of a severity-associated variant in the DYSF–ZNF638 locus (rs10191329), which prior genome-wide studies have linked to faster disability progression, showed enrichment of both broad rim lesions and perivascular cuffs—hinting that genetic determinants of severity may act partly by promoting more aggressive, more heavily inflamed lesion architectures.

The clinical correlations were equally revealing. Perivascular cuffs were associated with an increased microglia/macrophage activation score across the brain and with a younger age at death, suggesting that donors whose brains harbored these lymphocyte accumulations died earlier and with more widespread innate immune activation. Microglial nodules in the normal-appearing white matter—the tissue that appears spared at first glance—were associated with a higher proportion of actively inflamed lesions elsewhere, implying that these subtle clusters of immune cells may be a harbinger of a brain in which inflammatory lesion formation remains in full swing. The presence of these nodules may thus serve as an indicator of ongoing, brain-wide inflammatory activity even in regions that would never be flagged as lesions on conventional examination.

Broad rim lesions carried perhaps the weightiest clinical associations of all. Donors with BRLs showed increased proportions of both active and mixed active/inactive lesions, a higher lesion rate in the brainstem, and—a crucial finding—a higher age-related Multiple Sclerosis Severity Score, a metric that adjusts disability for age to capture how aggressively the disease has progressed. The broad rim itself is thought to represent a smoldering zone of myelin-phagocytosing macrophages at the lesion edge, and the new data reinforce the idea that these structures are engines of ongoing damage rather than passive scars. The association with brainstem lesions is particularly notable, as brainstem involvement often correlates with severe, disabling disease courses.

On the repair side of the ledger, poor remyelination efficiency painted its own distinctive portrait. Donors classified as poorly remyelinating showed a higher proportion of mixed active/inactive and inactive lesions and, tellingly, a shorter disease duration. The link with shorter disease duration is intriguing and suggests that inefficient repair may accelerate the trajectory toward severe disability and death, compressing the disease course. It may also reflect a vicious cycle in which active inflammation impairs oligodendrocyte precursor cells’ ability to rebuild myelin sheaths, leaving lesions unrepaired and vulnerable axons exposed to further injury. The loss of the protective, insulating myelin sheath ultimately leads to neurodegeneration, and the study’s authors and their predecessors have consistently shown that inflammatory lesion activity persists until death and correlates closely with that neurodegeneration.

The lesion classification system underpinning the work is itself a technical achievement. Lesions in white matter and deep grey matter were categorized by integrating innate inflammatory activity, assessed through HLA staining, with myelin status assessed by PLP staining. Reactive sites—aggregates of microglia and macrophages without demyelination—were distinguished from active lesions with dense microglial accumulation and partial myelin loss, mixed lesions with a rim of myeloid cells surrounding a quiescent core, inactive hypocellular lesions, and remyelinated lesions showing sparse inflammation and partial myelination restoration. Cortical lesions were classified by anatomical location as leukocortical, intracortical, or subpial. Against this standardized backdrop, the donor-specific features stand out as modifiers—variables that reshape the statistical landscape of lesion types and clinical outcomes rather than merely adding noise.

What makes this study transformative is its implication for how post-mortem MS research is conducted and interpreted. Rather than treating all donors with MS as members of a single pathological population, the findings argue for a stratified approach in which cuff status, nodule status, BRL presence, and remyelination efficiency are recorded alongside lesion classifications. Two donors with identical lesion loads could, in this framework, represent fundamentally different biological states—one with brisk lymphocyte-driven inflammation tied to HLA risk alleles, the other with smoldering broad-rimmed lesions tied to severity loci and a failing repair program. Pooling such donors in a typical case-control analysis would wash out precisely the signals that matter most for understanding disease mechanisms and for designing targeted therapies.

The work also resonates with the broader movement in MS research toward recognizing the disease as a spectrum of interacting pathological dimensions rather than a single entity. Prior analyses from the same cohort demonstrated that lesion load, the proportion of mixed active/inactive lesions, and microglial activation scores associate with clinical severity while also varying enormously between individuals. By showing that this inter-individual variability has genetic roots and clinical consequences, the new study closes the loop: it connects the genome to the lesion microenvironment to the bedside. Patients carrying risk or severity alleles may be pre-programmed, in a sense, to build different kinds of lesions and to repair them differently, with predictable consequences for how their disease unfolds.

For the estimated 2.9 million people worldwide living with multiple sclerosis, the clinical translation of such findings remains on the horizon rather than at hand—autopsy tissue, by definition, comes from those who have died, and the features studied here cannot yet be visualized reliably in living patients. But the biological logic they reveal is actionable. Therapies aimed at enhancing remyelination, for example, may need to be tailored differently for patients whose reparative capacity is genetically or immunologically compromised, while anti-inflammatory strategies might be stratified by whether a patient’s pathology is dominated by lymphocytic cuffing or by smoldering myeloid rims. Biomarker development, too, could take cues from these pathological dimensions, seeking surrogate markers in blood or advanced imaging that recapitulate what the microscope sees.

In the end, the study’s message is one of biological individuality made rigorous. The 287 donors of the Netherlands Brain Bank, through their extraordinary gift, have shown that multiple sclerosis is not one disease stamped identically upon every brain it touches, but a process profoundly shaped by each person’s genetic inheritance and intrinsic inflammatory and reparative character. Reading those individual signatures in the tissue—and understanding how they link risk variants, lesion architecture, and clinical fate—may finally provide the framework needed to explain, and one day to predict, why this disease takes such different courses in different lives.