Learn Explainer
Lipoprotein(a) in your DNA: what two SNPs can say about an inherited level
Lipoprotein(a) is a cholesterol-carrying particle whose level in the blood is set mostly by one gene, LPA. A raw DNA file can carry two LPA variants, rs10455872 and rs3798220, that mark a high level. Neither reads the repeat inside the gene that does most of the work, so what they say runs in one direction only.
Key takeaways
- Lp(a) is an LDL particle with an extra protein, apo(a), bonded to it. Its level varies widely between people and is mostly inherited: 91 percent of the variation traced to the LPA gene in a 1992 family study[2].
- The main genetic driver is a repeat inside LPA, kringle IV type 2. Fewer copies make a smaller apo(a) and, on average, a higher level[2, 7]. Chips and variant files cannot count the copies.
- Two common variants, rs10455872 and rs3798220, mark high-Lp(a) versions of the gene in Europeans. Together they explained 36 percent of the variation in Lp(a) in the study that found them[3].
- In people of European ancestry, carrying one of the two variants went with 1.51 times the odds of coronary disease and carrying two or more with 2.57 times, as population averages[3].
- European guidelines describe measuring Lp(a) at least once in adulthood. A DNA file with neither variant does not show a low level[5, 15].
Lipoprotein(a), written Lp(a), is an LDL particle with a second protein attached[1]. Its level differs widely from person to person, and most of that difference is inherited: in a 1992 family study, the gene that makes the extra protein accounted for 91 percent of it[2]. Large genetic studies have since found that people whose genes give them a higher level have more coronary heart disease and more narrowing of the aortic valve, and European guidelines now describe measuring it[3, 4, 5].
An LDL particle with a second protein
Cholesterol travels in the blood inside particles. The best known, LDL, is a droplet of cholesterol and fat wrapped by one large protein, apolipoprotein B-100. Lp(a) is the same particle with a second protein, apolipoprotein(a), joined to the apoB-100 by a disulfide bond, a bridge between two sulfur atoms[1].
Apo(a) is the odd part. When its gene was first sequenced, in 1987, the protein turned out to be close kin to plasminogen, the precursor of the enzyme that breaks down blood clots[6]. It is built from folded loops called kringles. Apo(a) carries copies of plasminogen's fourth kringle in ten subtypes, numbered KIV-1 to KIV-10, one copy resembling its fifth, and a region resembling plasminogen's enzyme that does not work as one[1]. One of the ten, KIV-2, is repeated a different number of times on different copies of the gene[7].
Five terms this article relies on
- Lp(a)
- Lipoprotein(a): an LDL particle with apo(a) attached. A blood test reports its concentration.
- Apo(a)
- Apolipoprotein(a), the extra protein. LPA is the gene that makes it.
- KIV-2
- Kringle IV type 2, the part of apo(a) whose number of copies varies between people.
- Isoform
- One size of apo(a). Each copy of LPA makes an isoform whose size depends on its KIV-2 count.
- Mendelian randomisation
- Using genetic variants as a natural experiment: if people born with a higher level have more disease, the level is more likely to be a cause than a bystander.
A level set mostly by one gene
Studies of twins, families and siblings put its heritability at about 70 to over 90 percent, in European, Asian and African populations[1]. The figure most often quoted comes from 48 families of European descent, studied in 1992 with a gel method that could size each copy of the apo(a) gene. Siblings who had inherited the same two copies had closely matching Lp(a) levels; siblings who shared neither did not[2].
91%
of the variation in Lp(a) traced to the apo(a) gene in 48 families[2]
69%
explained by the number of kringle IV repeats alone, in the same study[2]
36%
explained by rs10455872 and rs3798220 together[3]
The repeat count does most of the work. Each KIV-2 copy is a stretch of about 5.6 thousand DNA letters, and a copy of the gene can carry anything up to more than 40 of them[7]. Fewer repeats make a smaller apo(a), which one review puts down to smaller apo(a) maturing more efficiently inside the liver cell, and people whose apo(a) isoforms have up to 22 kringle IV domains in all have, on average, markedly higher Lp(a) than people with only larger ones[7]. The 1992 study put the repeat's share of the variation at 69 percent and the rest of the gene's at 22[2]. Later estimates of how much the isoform size explains run from about 40 to 70 percent, depending on the method and the people studied[7].
Two things follow for anyone reading a genome file. The level is largely set by the genes a person was born with. And the thing that sets it, a count of long, near-identical repeats, is exactly what genotyping chips and standard variant files are not built to record.
What rs10455872 and rs3798220 capture, and what they miss
The two variants consumer files can carry were found in 2009 by the PROCARDIS consortium, which tested 48,742 variants in 2,100 candidate genes in 3,145 people with coronary disease and 3,352 without, all of European ancestry, then checked the results in 4,846 more cases and 4,594 more controls. The LPA region gave the strongest signal, through two variants that each raised the level of Lp(a)[3].
| Variant | GRCh38 position | Change | High-Lp(a) allele frequency, Europeans | Odds of coronary disease, per variant |
|---|---|---|---|---|
| rs10455872 | 6:160589086 | A to G | about 7% | 1.70 (1.49 to 1.95) |
| rs3798220 | 6:160540105 | T to C | about 2% | 1.92 (1.48 to 2.49) |
Source: Frequencies and odds ratios from the PROCARDIS study, with 95 percent confidence intervals[3]. Positions and alleles as the report reads them[8, 9].
Neither variant needs to be the cause. Both turned up more often on copies of the gene with fewer KIV-2 repeats and smaller apo(a)[3], so they act as tags: letters a chip can read that tend to travel with a short repeat a chip cannot. rs3798220 does change the protein, swapping one amino acid in the protease-like region[3]. Combined into a score, one tag went with 1.51 times the odds of coronary disease and two or more with 2.57 times. Once the analysis allowed for each person's measured Lp(a), the score's link to disease disappeared, which is what you would expect if the tags matter only through the level[3].
Their reach is the problem. Together they explained 36 percent of the variation in Lp(a) in the PROCARDIS participants[3], so most of it lies elsewhere, mainly in the repeat. Many people with a short repeat, and a high level, will carry neither tag.
What the level has been linked to
High Lp(a) has gone with heart disease in observational studies, but a blood level measured once can be a bystander, raised by whatever else is wrong. Genetics gives a cleaner test, because the LPA gene a person inherits was fixed at conception. This is Mendelian randomisation, and Lp(a) has passed it several times.
In Copenhagen, people in the top 5 percent of measured Lp(a) had 2.6 times the rate of heart attack of people below the 22nd percentile[10]. The study then used KIV-2 repeat counts as the instrument: each doubling of Lp(a) predicted by genotype went with a 22 percent higher rate, and the authors called the result consistent with a causal link[10]. The PROCARDIS score reached the same conclusion from the two tags[3]. rs10455872 is also one of the 241 variants in the report's coronary artery disease polygenic score.
The heart valve shows the same pattern. In 6,942 people of European ancestry, rs10455872 was the only variant to reach genome-wide significance for calcium in the aortic valve, with about twice the odds per copy. In a Swedish cohort the same variant predicted later aortic stenosis, a narrowing of the valve, at 1.68 times the rate per copy, and the finding was repeated in a Danish one[4].
460,506
UK Biobank participants with a measured Lp(a)[11]
1.11
hazard ratio for atherosclerotic disease per 50 nmol/L of Lp(a)[11]
101.5 mg/dL
fall in Lp(a) with the genetic effect of a 1 mmol/L fall in LDL cholesterol[12]
The largest single cohort is UK Biobank. Among 460,506 middle-aged participants, the rate of atherosclerotic cardiovascular disease rose in a straight line with Lp(a), by 11 percent for each 50 nmol/L[11]. That is a population average across the whole range, and most people sit low on it: the medians in the next section run from 16 to 75 nmol/L[11].
Size matters for any drug. A 2018 Mendelian randomisation analysis estimated that Lp(a) would need to fall by 101.5 mg/dL, with a 95 percent confidence interval of 71.0 to 137.0, to match the genetic effect on coronary disease of lowering LDL cholesterol by 1 mmol/L[12]. Lowering it by a few mg/dL in someone whose level is already modest would, on this estimate, do little.
Finland supplied a natural experiment of its own. Comparing the exomes of 3,000 Finns with 3,000 other Europeans, a 2014 study found Finnish-enriched variants that break the LPA gene at a splice site. Carriers had lower Lp(a), and across 36,262 Finns with national health records they had 0.84 times the odds of cardiovascular disease, which the authors described as the first demonstration in humans of a link between very low Lp(a) and protection[13].
Ancestry moves the level and the tags
Typical Lp(a) differs several-fold between populations. In UK Biobank the median was 19 nmol/L in White participants, 31 in South Asian, 75 in Black and 16 in Chinese participants[11]. A study of 4,464 African Americans in the Jackson Heart Study starts from the observation that people of African descent have roughly twice the Lp(a) of European Americans[14].
The risk that came with each step up looked similar across groups: per 50 nmol/L, a hazard ratio of 1.11 in White, 1.10 in South Asian and 1.07 in Black participants[11]. What changes is how many people sit high on the scale, and how well the two tags find them. Both were discovered and measured in Europeans[3]. The Jackson Heart Study traced most of the African American difference to other variants in LPA[14]. A file from someone of African, Asian or mixed ancestry with neither tag therefore says even less than a European one.
What the European guidelines say about measuring it
Because the level is mostly inherited, European cardiology bodies treat one measurement in adult life as informative. Three documents set out the current position. None of them involves DNA: all three describe a blood test.
| Document | On measuring Lp(a) | Levels it singles out |
|---|---|---|
| ESC/EAS dyslipidaemia guidelines, 2019 | Measurement should be considered at least once in each adult's lifetime (class IIa, level of evidence C) | Above 180 mg/dL (430 nmol/L), a level the guideline links to a lifetime risk like that of heterozygous familial hypercholesterolaemia |
| EAS consensus statement, 2022 | Lp(a) should be measured at least once in adults and read alongside overall cardiovascular risk | Below 30 mg/dL (75 nmol/L) to rule out and above 50 mg/dL (125 nmol/L) to rule in raised risk, with a grey zone between |
| ESC/EAS focused update, 2025 | Elevated Lp(a) should be considered a risk-enhancing factor in all adults (class IIa, level of evidence B) | Above 50 mg/dL (105 nmol/L) |
Source: ESC/EAS 2019[15]; EAS consensus[5]; ESC/EAS focused update[16].
The EAS consensus describes a causal and continuous relationship between Lp(a) and cardiovascular outcomes, aortic stenosis included, and an elevated level as a risk factor even when LDL cholesterol is very low[5]. It does not support Lp(a) as a cause of venous thrombosis[5].
What our report says, file by file
Every report built from a chip export, a VCF or a gVCF has an Lp(a) card. It reads the two tags at the positions in the table above and counts the high-Lp(a) alleles: one from each copy of each variant, so anywhere from none to four. The card groups the count the way the PROCARDIS score did[3].
| Tag alleles found | Label on the card | PROCARDIS odds of coronary disease |
|---|---|---|
| None | Not flagged (not ruled out) | 1, the comparison group |
| One | Likely elevated | 1.51 (1.38 to 1.66) |
| Two or more | Likely markedly elevated | 2.57 (1.80 to 3.67) |
Source: Labels as of October 2026. Odds ratios with 95 percent confidence intervals, people of European ancestry[3].
There is deliberately no normal or low label. A file with no tag allele gets Not flagged, with a note that the repeat behind most of the variation cannot be read from a variant file and that the tags were characterised mainly in Europeans. What the card can see depends on the file.
| Chip export | VCF or gVCF | BAM or CRAM | |
|---|---|---|---|
| rs10455872 and rs3798220 | Read when the chip has a row for each. Neither is an A/T or C/G pair, so the strand of the row does not matter | Read from the file's records | The Deep Read panel does not read LPA; the card comes from the report's VCF or gVCF |
| KIV-2 repeat count | Not read | Not read | Not read. A reads-based count has been designed but is not built |
| Neither tag found | Not flagged. The card does not say whether the chip carried both rows | Not flagged. A plain VCF cannot prove the positions were covered | As for the VCF or gVCF |
| File with no readable genotypes | Not applicable | A gVCF whose blocks call nothing, or a VCF of structural variants only, gets a card saying the markers were not read | As for the VCF or gVCF |
Source: The report's Lp(a) module as of October 2026. The letters at each position are as in the table of the two variants above[8, 9].
The sample report shows the card in full, and the free file check tells you what kind of file you have before anything is paid for. A missing line in a variant file has its own explainer, and what each genome file can read covers why a BAM or CRAM holds more than the VCF made from it. Three genes behind other blood measurements, GLS2, SLC39A8 and SLC5A11, have a guide of their own.
What Aimosti would (and wouldn't) show you
The report's Lp(a) card reads rs10455872 and rs3798220 from a chip export, a VCF or a gVCF and counts the tag alleles it finds: none, one, or two or more. It labels them Not flagged (not ruled out), Likely elevated and Likely markedly elevated, says the result is a genetic proxy and not a measurement, and explains that the repeat behind most of the variation cannot be read from a variant file. A file with no readable genotypes gets a card that says the markers were not read.
What we won't claim
We won't read a file with neither tag variant as a normal or low Lp(a), estimate a level in mg/dL or nmol/L from DNA, or turn the card into a personal chance of heart disease. A blood measurement is the only thing that says what the level is.
Bottom line. Lp(a) is the rare blood lipid whose level is mostly written in the genes, and two common variants can flag part of that inheritance from a raw DNA file. The part they cannot see, the number of KIV-2 repeats in LPA, is the larger part, so a flag means something and its absence settles nothing.
Questions people ask
Can a 23andMe or AncestryDNA file tell me my Lp(a) level?
No. A chip export can carry the two tag variants, if the chip has rows for them, and those can flag a high-Lp(a) version of the gene[3]. It cannot count the KIV-2 repeats that set most of the level, and it holds no measurement. The level itself comes from a blood test, reported in mg/dL or nmol/L[5].
If my DNA file has neither variant, is my Lp(a) low?
It is not shown to be. The two tags explained about a third of the variation in Lp(a) in the study that found them, and the repeat they stand in for explains more[3, 2]. Plenty of people with a high level will carry neither tag, and the tags were found and measured in Europeans[3].
Is Lp(a) the same as LDL cholesterol?
No. Lp(a) is an LDL particle with an extra protein attached, and a blood test reports it on its own. Its level is mostly inherited[1, 2]. In Mendelian randomisation, it took a fall of about 100 mg/dL in Lp(a) to match the genetic effect of a 1 mmol/L fall in LDL cholesterol[12].
Why do Lp(a) results come in two units?
Some assays report the particles' mass (mg/dL) and others their number (nmol/L). Apo(a) comes in many sizes, so there is no single conversion factor, and the EAS consensus advises against using one[5]. That is why 50 mg/dL sits beside 125 nmol/L in the 2022 consensus and 105 nmol/L in the 2025 guideline update[16].
Is there a drug that lowers Lp(a)?
Drugs designed to lower it have been in trials. One of them, pelacarsen, lowered Lp(a) but missed its main goal: Novartis announced in September 2026 that, in 8,323 people with cardiovascular disease and high Lp(a), it did not reduce cardiovascular events compared with placebo[17]. The Finnish gene-breaking variants are evidence of a different kind: a lifelong low level that went with less disease[13].
References
- Schmidt K, Noureen A, Kronenberg F, Utermann G. Structure, function, and genetics of lipoprotein (a). Journal of Lipid Research, 2016. doi:10.1194/jlr.R067314
- Boerwinkle E, Leffert CC, Lin J, Lackner C, Chiesa G, Hobbs HH. Apolipoprotein(a) gene accounts for greater than 90% of the variation in plasma lipoprotein(a) concentrations. Journal of Clinical Investigation, 1992. doi:10.1172/JCI115855
- Clarke R, Peden JF, Hopewell JC, et al. Genetic variants associated with Lp(a) lipoprotein level and coronary disease. New England Journal of Medicine, 2009. doi:10.1056/NEJMoa0902604
- Thanassoulis G, Campbell CY, Owens DS, et al. Genetic associations with valvular calcification and aortic stenosis. New England Journal of Medicine, 2013. doi:10.1056/NEJMoa1109034
- Kronenberg F, Mora S, Stroes ESG, et al. Lipoprotein(a) in atherosclerotic cardiovascular disease and aortic stenosis: a European Atherosclerosis Society consensus statement. European Heart Journal, 2022. doi:10.1093/eurheartj/ehac361
- McLean JW, Tomlinson JE, Kuang WJ, et al. cDNA sequence of human apolipoprotein(a) is homologous to plasminogen. Nature, 1987. doi:10.1038/330132a0
- Coassin S, Kronenberg F. Lipoprotein(a) beyond the kringle IV repeat polymorphism: the complexity of genetic variation in the LPA gene. Atherosclerosis, 2022. doi:10.1016/j.atherosclerosis.2022.04.003
- rs10455872. dbSNP, US National Center for Biotechnology Information. GRCh38 6:160589086 A>G, the position and alleles the report's calling code uses.
- rs3798220. dbSNP, US National Center for Biotechnology Information. GRCh38 6:160540105 T>C, the position and alleles the report's calling code uses.
- Kamstrup PR, Tybjaerg-Hansen A, Steffensen R, Nordestgaard BG. Genetically elevated lipoprotein(a) and increased risk of myocardial infarction. JAMA, 2009. doi:10.1001/jama.2009.801
- Patel AP, Wang M, Pirruccello JP, et al. Lp(a) (lipoprotein[a]) concentrations and incident atherosclerotic cardiovascular disease: new insights from a large national biobank. Arteriosclerosis, Thrombosis, and Vascular Biology, 2021. doi:10.1161/ATVBAHA.120.315291
- Burgess S, Ference BA, Staley JR, et al. Association of LPA variants with risk of coronary disease and the implications for lipoprotein(a)-lowering therapies: a Mendelian randomization analysis. JAMA Cardiology, 2018. doi:10.1001/jamacardio.2018.1470
- Lim ET, Würtz P, Havulinna AS, et al. Distribution and medical impact of loss-of-function variants in the Finnish founder population. PLOS Genetics, 2014. doi:10.1371/journal.pgen.1004494
- Deo RC, Wilson JG, Xing C, et al. Single-nucleotide polymorphisms in LPA explain most of the ancestry-specific variation in Lp(a) levels in African Americans. PLOS ONE, 2011. doi:10.1371/journal.pone.0014581
- Mach F, Baigent C, Catapano AL, et al. 2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk. European Heart Journal, 2020. doi:10.1093/eurheartj/ehz455
- Mach F, Koskinas KC, Roeters van Lennep JE, et al. 2025 Focused Update of the 2019 ESC/EAS Guidelines for the management of dyslipidaemias. European Heart Journal, 2025. doi:10.1093/eurheartj/ehaf190
- Novartis announces Lp(a)HORIZON Phase III topline results for pelacarsen in patients with elevated Lp(a) and established cardiovascular disease. Novartis media release, 2026. 4 September 2026: the primary endpoint, a composite of cardiovascular death, non-fatal heart attack, non-fatal stroke and urgent revascularisation, was not met in 8,323 patients.
Last reviewed . Every number on this page links to the source it comes from; if one of them has moved, tell us.