CSF Biomarkers & the ATN Framework
Cerebrospinal fluid biomarkers have become indispensable in the diagnosis of neurodegenerative dementias, particularly Alzheimer disease (AD). The ATN framework — originally proposed in 2018 to classify individuals by amyloid (A), tau (T), and neurodegeneration (N) biomarker status — shifted dementia diagnosis from syndromic classification to a biologically defined model. In 2024, the Alzheimer’s Association expanded this framework to AT1T2NISV, adding inflammation, synaptic, and vascular categories, and formally including blood-based biomarkers for the first time. The clinical urgency of accurate biomarker diagnosis has intensified with FDA approval of anti-amyloid immunotherapies (lecanemab, donanemab), which require confirmation of amyloid pathology before initiation. CSF also remains essential for diagnosing Creutzfeldt-Jakob disease (CJD) via RT-QuIC and confirming synucleinopathies via alpha-synuclein seed amplification assay (SAA).
Bottom Line
- ATN framework (2024 revision): Expanded to AT1T2NISV — A (amyloid), T1 (early tau phosphorylation, Core 1), T2 (later tau aggregation on PET, Core 2), N (neurodegeneration), I (inflammation), S (synapse), V (vascular); a single abnormal Core 1 biomarker is sufficient to establish an AD diagnosis
- Core 1 CSF biomarkers: CSF Aβ42/40 ratio, p-tau181/Aβ42 ratio, t-tau/Aβ42 ratio, and p-tau217 each achieve ~90% concordance with amyloid PET; p-tau217 has the highest diagnostic accuracy among all tau species
- Pre-analytical factors: CSF must be collected in polypropylene (soft plastic) tubes — Aβ adheres to polystyrene, producing artifactually low values; standardized centrifugation and handling are essential
- Non-AD CSF biomarkers: RT-QuIC for prion protein has excellent sensitivity/specificity for CJD; alpha-synuclein SAA is commercially available for Lewy body disease; 14-3-3 protein is elevated in CJD but nonspecific; NfL reflects neurodegeneration across multiple conditions
- CSF vs blood vs PET: CSF provides the most comprehensive single-sample evaluation (AD, CJD, LBD simultaneously); plasma p-tau217 is emerging as a screening tool with near-CSF accuracy; PET remains the gold standard for spatial staging
The ATN Framework: Original and Expanded
Original 2018 ATN Framework
The 2018 NIA-AA Research Framework introduced a purely biological definition of AD, classifying individuals by three categories of biomarkers rather than clinical syndrome:
- A (Amyloid): CSF Aβ42 or amyloid PET — reflects brain amyloid plaque burden
- T (Tau phosphorylation): CSF p-tau181 or tau PET — reflects neurofibrillary tangle pathology
- N (Neurodegeneration): CSF total tau (t-tau), FDG-PET, or structural MRI atrophy — reflects neuronal injury
Under this framework, an individual classified as A+T+N+ has biomarker evidence of amyloid pathology, tau pathology, and neurodegeneration, consistent with AD across its biological continuum regardless of clinical symptoms. The system enabled research classification of individuals along the AD pathological spectrum even before symptom onset.
2024 Expanded Framework: AT1T2NISV
The 2024 Alzheimer’s Association revised criteria (Jack et al.) significantly expanded the framework to seven categories, reorganized biomarkers into Core 1 (diagnostic, early-changing) and Core 2 (staging, later-changing), and formally included blood-based biomarkers as diagnostic tools for the first time. The expanded framework also added categories for inflammation (I), synaptic dysfunction (S), and vascular pathology (V), reflecting the multifactorial nature of neurodegeneration.
| Category | What It Measures | Biomarker Examples | Classification |
|---|---|---|---|
| A | Amyloid pathology | Amyloid PET, CSF Aβ42/40, plasma Aβ42/40 | Core 1 |
| T1 | Early tau phosphorylation | CSF p-tau181, CSF/plasma p-tau217 | Core 1 |
| T2 | Later tau aggregation | Tau PET (flortaucipir), biofluid markers (p-tau205) | Core 2 |
| N | Neurodegeneration | CSF/plasma NfL, CSF t-tau, FDG-PET, MRI volumetrics | Non-specific |
| I | Inflammation | CSF/plasma GFAP, sTREM2, YKL-40 | Non-specific |
| S | Synaptic dysfunction | CSF neurogranin, SNAP-25, NPTX2 | Non-specific |
| V | Vascular pathology | MRI white matter hyperintensities, microbleeds | Non-specific |
Key Changes in the 2024 Revised Criteria
- Core 1 biomarkers (diagnostic): Any single abnormal Core 1 biomarker is sufficient to establish an AD diagnosis — amyloid PET, CSF Aβ42/40 ratio, CSF p-tau181/Aβ42 ratio, CSF t-tau/Aβ42 ratio, or accurate plasma p-tau217 assays
- Core 2 biomarkers (staging): Tau PET and biofluid markers of tau aggregation provide disease staging, prognosis, and increased confidence that AD is contributing to symptoms
- Blood-based biomarkers formally included for the first time — plasma p-tau217 now recognized alongside CSF and PET
- Biological definition: AD defined as a biological process beginning with neuropathologic change while asymptomatic; clinical symptoms inform staging, not diagnosis
- Biomarker-only diagnosis: AD can be diagnosed in asymptomatic individuals with positive Core 1 biomarkers (biological Stage A)
Core CSF Biomarkers for Alzheimer Disease
CSF Aβ42/40 Ratio
Patients with AD have lower CSF levels of Aβ1-42 compared with age-matched controls, presumably due to sequestration of Aβ42 within amyloid plaques. The finding was less intuitive than the elevation of tau in AD but is equally informative: CSF Aβ42 is inversely correlated with brain amyloid plaque burden. Using the Aβ42/Aβ40 ratio rather than absolute Aβ42 values corrects for individual differences in total amyloid-beta production and significantly improves diagnostic accuracy. In CSF, there is a ~50% difference in the Aβ42/40 ratio between AD patients and healthy controls (vs only ~20% in plasma, which amplifies the need for precise measurements). The CSF Aβ42/40 ratio achieves approximately ~90% concordance with amyloid PET and is a Core 1 biomarker in the 2024 revised criteria. It is available on both the Lumipulse and Elecsys automated platforms.
Phosphorylated Tau (p-tau) Species
Phosphorylated tau in CSF reflects tau pathology relatively specific to AD, unlike total tau which rises nonspecifically with any neuronal injury. The amyloid-tau index (ATI) combined with absolute p-tau181 is a much stronger indicator of AD than either Aβ or tau alone.
| p-tau Species | Diagnostic Role | Key Characteristics | Category |
|---|---|---|---|
| p-tau181 | Established CSF/plasma AD biomarker | Reflects mature NFT pathology; p-tau181/Aβ42 ratio ~90% concordant with amyloid PET; most widely used clinically | Core 1 |
| p-tau217 | Highest diagnostic accuracy | Better correlated with both amyloid and tau PET than p-tau181; plasma p-tau217 equivalent to CSF biomarkers for predicting amyloid PET status; AUC 0.93–0.96 | Core 1 |
| p-tau205 | Marker of tau aggregation | Better correlated with tau PET than p-tau181; reflects later-stage tau spreading; emerging biofluid alternative to tau PET for staging | Core 2 |
| p-tau231 | Earliest-changing tau biomarker | Predominant in pre-tangles; abnormal at lowest amyloid load; plateaus and does not track progression; best for preclinical screening | Research |
Total Tau (t-tau)
CSF total tau is a nonspecific marker of neuronal injury. Elevations are seen in AD but also in stroke, traumatic brain injury, encephalitis, and non-AD neurodegenerative diseases. It is important to note that elevated CSF tau is not specific to AD because any brain injury that damages neurons will release the normally intracellular tau protein into CSF. Markedly elevated CSF t-tau (often >1,000–1,200 pg/mL) is characteristic of CJD, where rapid neuronal destruction produces exceptionally high levels. In the context of the ATN framework, the t-tau/Aβ42 ratio is a Core 1 biomarker when used to assess amyloid status, and standalone CSF t-tau is classified under the N (neurodegeneration) category.
Non-AD CSF Biomarkers
14-3-3 Protein and Prion RT-QuIC for CJD
CSF 14-3-3 protein was among the earliest CJD biomarkers, but as a normal intraneuronal protein that spills into CSF during any significant neuronal injury, elevated levels must be interpreted cautiously. Real-time quaking-induced conversion (RT-QuIC) has transformed CJD diagnosis — this seed amplification assay exploits the ability of misfolded prion "seeds" to recruit and convert recombinant prion protein, amplifying signal to detectable levels. RT-QuIC demonstrates sensitivity of 92%–96% and specificity of 99%–100% for sporadic CJD.
Alpha-Synuclein SAA for Lewy Body Disease
The amplification principles behind prion RT-QuIC have been successfully applied to alpha-synuclein, the protein aggregated in Lewy bodies in Parkinson disease (PD) and related disorders. Although absolute levels of alpha-synuclein in CSF do not reliably distinguish patients from controls, the seeded amplification assay for aggregated alpha-synuclein has shown excellent performance. A commercially available CSF SAA (available since 2022) reports results in a dichotomous fashion as positive or negative, with 92%–95% sensitivity and 95%–97% specificity for PD, LBD, and multiple system atrophy. Sensitivity and specificity have also been established for autopsy-confirmed cases with and without Lewy body pathology.
The assay also detects prodromal synucleinopathy, including isolated REM sleep behavior disorder and asymptomatic PD gene carriers. Although several studies have demonstrated that particular elements of SAA may distinguish synuclein strains in PD compared with LBD or MSA, the currently available clinical assay does not provide that distinction. Importantly, care must be taken to minimize blood contamination during CSF collection, as alpha-synuclein is expressed in red blood cells and can produce false-positive results from traumatic taps.
Neurofilament Light Chain (NfL)
NfL is a structural component of large-caliber myelinated axons that is released into CSF and blood during neuronal injury. CSF NfL is elevated in AD, frontotemporal dementia, ALS, MS, traumatic brain injury, and many other neurological conditions. It is not disease-specific but correlates with the rate of cognitive decline in AD, MCI, and even asymptomatic aging. NfL is classified under the N (neurodegeneration) category in the expanded ATN framework. In anti-amyloid clinical trials, NfL has not consistently shown a treatment effect, suggesting it may reflect neurodegeneration independent of amyloid clearance. NfL is measured via the Simoa platform (single molecule array) or Lumipulse and is being translated from research to routine clinical practice.
GFAP (Glial Fibrillary Acidic Protein)
GFAP is a marker of astrocyte activation and neuroinflammation classified under I (inflammation) in the expanded ATN framework. CSF and plasma GFAP are elevated in multiple neurodegenerative diseases but become abnormal earlier than tau biomarkers in AD progression. GFAP discriminates amyloid-positive from amyloid-negative individuals with high sensitivity and may serve as an early indicator of AD-related pathology. In anti-amyloid trials, plasma GFAP decreased with both lecanemab and donanemab treatment, suggesting potential utility as a treatment response marker. GFAP is measured via single-molecule array (Simoa) technology but is not yet FDA-cleared for AD diagnosis.
Oligoclonal Bands
CSF oligoclonal bands (OCBs) are not a marker of neurodegenerative disease but are essential in the differential diagnosis of cognitive decline when inflammatory or autoimmune etiologies are suspected. Their presence in CSF but not serum indicates intrathecal immunoglobulin synthesis and is characteristic of multiple sclerosis, neurosarcoidosis, CNS vasculitis, and autoimmune encephalitis. OCBs are not expected in typical AD, LBD, FTD, or CJD.
When to Perform a Lumbar Puncture for CSF Biomarkers
- Suspected AD requiring treatment confirmation: CSF Aβ42/40, p-tau181, and t-tau to confirm amyloid pathology before anti-amyloid immunotherapy when PET is unavailable
- Rapidly progressive dementia: 14-3-3, RT-QuIC, t-tau, and alpha-synuclein SAA — a single LP can simultaneously assess AD, CJD, and LBD biomarkers
- Diagnostic uncertainty between AD and LBD: Alpha-synuclein SAA differentiates synucleinopathies; concomitant AD biomarkers assess for mixed pathology
- Suspected autoimmune encephalopathy: CSF cell count, protein, glucose, OCBs, and autoimmune antibody panels
- Intermediate plasma biomarker results: CSF serves as confirmatory testing when blood-based screening yields equivocal results
CSF Collection and Handling
The reliability of CSF biomarker results depends critically on pre-analytical factors. Failure to adhere to standardized collection and processing protocols is a leading cause of discordant or uninterpretable results, particularly for Aβ measurements which are highly sensitive to tube material and handling conditions.
Critical Pre-Analytical Factors
- Collection tubes: Must use polypropylene (soft plastic) — Aβ adheres to polystyrene (hard plastic), producing artifactually low Aβ42 values and potential false-positive results for amyloid pathology
- Volume: Typically 10–15 mL of CSF is sufficient for a comprehensive biomarker panel including AD, CJD, and alpha-synuclein markers; at least 1–2 mL per analyte is recommended
- Centrifugation: Samples centrifuged within 1–2 hours at 2,000 × g for 10 minutes at room temperature to remove cells and cellular debris
- Aliquoting: Use polypropylene tubes; avoid repeated freeze-thaw cycles, which degrade Aβ and tau proteins
- Blood contamination: Traumatic taps introduce blood-derived alpha-synuclein (expressed in RBCs), potentially causing false-positive SAA results; grossly bloody samples should not be used for alpha-synuclein testing
- Storage and transport: Process and freeze at –80°C within 2 hours if not analyzed immediately; ship on dry ice for remote laboratory analysis
Automated Assay Platforms
| Feature | Lumipulse G (Fujirebio) | Elecsys (Roche) |
|---|---|---|
| Technology | Chemiluminescent enzyme immunoassay (CLEIA) | Electrochemiluminescence immunoassay (ECLIA) |
| CSF analytes | Aβ42, Aβ40, t-tau, p-tau181 | Aβ42, Aβ40, t-tau, p-tau181 |
| Plasma capability | FDA-cleared plasma p-tau217/Aβ1-42 ratio (May 2025) — first FDA-cleared blood test for AD | Plasma p-tau181 and NfL available for research |
| Amyloid PET concordance | ~90% for CSF t-tau/Aβ42 ratio | ~90% for CSF Aβ42/40 ratio |
| Commercial availability | Multiple reference labs; plasma test at Labcorp and Quest (2025) | ARUP, Mayo, and other reference laboratories |
Mass spectrometry-based assays (e.g., PrecivityAD2 by C2N Diagnostics) offer an alternative approach. The PrecivityAD2 test combines plasma %p-tau217 and Aβ42/40 ratio in an algorithm optimized for detecting amyloid pathology, achieving 92% sensitivity and 96% specificity for amyloid PET positivity. A two-cutoff approach has emerged for plasma biomarkers: patients below a low cutoff are likely amyloid-negative, those above a high cutoff are likely amyloid-positive (PPV >96%), and those in the intermediate zone require confirmatory testing (PET or CSF). Cutoff values are platform-specific and not interchangeable between assays.
CSF Profile Interpretation
| Diagnosis | Aβ42 | Total Tau | p-tau181 | 14-3-3 | Prion RT-QuIC | α-syn SAA |
|---|---|---|---|---|---|---|
| Alzheimer disease | Low | High | High | Normal | Negative | Negative |
| Creutzfeldt-Jakob disease | Normal | Very high | Normal | High | Positive | Negative |
| Lewy body dementia | Variable | Variable | Variable | Normal | Negative | Positive |
| Mixed AD/LBD | Low | High | High | Normal | Negative | Positive |
| Frontotemporal dementia | Normal | Normal–mild ↑ | Normal | Normal | Negative | Negative |
| Normal aging | Normal | Normal | Normal | Normal | Negative | Negative |
Interpreting CSF Biomarker Results
- Classic AD profile: Low Aβ42/40 + elevated p-tau181 + elevated t-tau; low ATI with high p-tau181 is strong evidence for AD pathology
- Indeterminate results: Discordant biomarkers (e.g., low Aβ42 with normal p-tau) may indicate early AD, copathology, or pre-analytical variability — consider confirmatory PET
- Elevated t-tau with normal p-tau and normal Aβ42: Suggests non-AD neuronal injury; markedly elevated t-tau (>1,000–1,200 pg/mL) raises concern for CJD
- Mixed pathology: ~80% of LBD cases have concomitant AD pathology at autopsy; positive alpha-synuclein SAA with abnormal AD biomarkers suggests mixed AD/LBD
- Racial/ethnic variation: CSF t-tau, p-tau181, and NfL are lower in African American participants vs non-Hispanic White participants, though Aβ42 and Aβ40 levels are similar; cutoffs may need adjustment
CSF vs Blood vs PET: Comparative Approach
| Feature | CSF Biomarkers | Blood-Based Biomarkers | PET Imaging |
|---|---|---|---|
| Invasiveness | Lumbar puncture required | Simple blood draw | IV radiotracer injection |
| Cost | Moderate ($500–$1,500) | Low–moderate ($250–$800) | High ($3,000–$6,000) |
| Accessibility | Most medical centers | Widely accessible via commercial labs | Limited to PET-capable centers |
| Diagnostic breadth | AD, CJD, LBD from single sample | Primarily AD (p-tau217, Aβ42/40) | AD-specific (amyloid or tau PET) |
| Spatial information | None (global measure) | None (global measure) | Regional pathology distribution (staging) |
| Amyloid PET concordance | ~90% for Aβ42/40 and p-tau ratios | AUC 0.93–0.96 for plasma p-tau217 | Gold standard |
| Serial monitoring | Impractical (repeated LPs) | Feasible (serial blood draws) | Feasible but expensive |
| Unique role | CJD diagnosis (RT-QuIC), alpha-synuclein SAA | Population screening; triage before PET/CSF | Tau staging (Braak), amyloid clearance monitoring |
Biomarkers in Anti-Amyloid Therapy Trials
Anti-amyloid antibody trials have evaluated CSF and plasma biomarker changes to identify potential treatment monitoring markers:
| Agent | CSF Aβ42 | CSF t-tau | CSF p-tau181 | Plasma p-tau | Plasma GFAP | NfL |
|---|---|---|---|---|---|---|
| Aducanumab | Increased | Decreased | Decreased | ↓ (181) | NR | NR |
| Lecanemab | Increased | Decreased | Decreased | ↓ (181) | Decreased | No change |
| Donanemab | NR | NR | NR | ↓ (217) | Decreased | No effect |
Serial CSF examination is impractical for monitoring; the validation of plasma biomarkers as treatment response markers remains a critical unmet need. The consistent decrease in plasma p-tau and GFAP with effective anti-amyloid therapy is encouraging, but these markers are not yet validated for guiding clinical treatment decisions. CSF biomarkers that reflect ongoing synaptic damage — including neurogranin, SNAP-25, and NPTX2 — are also being evaluated as outcome measures, since synaptic loss is the best neuropathologic correlate of dementia severity.
Emerging Diagnostic Workflow
The convergence of blood-based biomarkers, CSF assays, and PET imaging has led to a practical stepwise diagnostic algorithm for evaluating suspected AD in the era of disease-modifying therapy:
- Step 1 — Clinical assessment: Cognitive complaint with objective testing; rule out reversible causes with standard labs and structural imaging
- Step 2 — Blood-based biomarker screening: Plasma p-tau217 and/or Aβ42/40 ratio; if clearly positive or negative, may inform decisions without further testing; if intermediate, proceed to confirmatory testing
- Step 3 — Confirmatory testing: Amyloid PET or CSF biomarkers (Aβ42/40, p-tau181, t-tau); APOE genotyping for ARIA risk stratification if anti-amyloid therapy is being considered
- Step 4 — Disease staging: Tau PET for biological staging if available; clinical staging based on functional assessment
- Step 5 — Treatment decision: Anti-amyloid therapy eligibility assessment, risk-benefit discussion (APOE4 status, ARIA risk), baseline MRI for monitoring, and lifestyle intervention recommendations
This workflow reflects the emerging paradigm of blood-first screening, reserving more invasive or expensive confirmatory testing for cases where screening results are indeterminate or where treatment decisions require the highest diagnostic confidence.
Important Caveats for Biomarker-Based Diagnosis
- Renal insufficiency: Plasma p-tau181 and p-tau217 may be artifactually elevated in patients with CKD, potentially causing false-positive results
- Racial/ethnic variability: Plasma p-tau markers are less predictive of brain amyloid positivity among African American individuals; however, longitudinal change in plasma Aβ42/40 is equally predictive across racial groups
- Platform-specific cutoffs: Values are not interchangeable between Lumipulse, Elecsys, and mass spectrometry platforms; results must be interpreted using platform-specific reference ranges
- Biology-symptom dissociation: A positive Core 1 biomarker establishes biological AD but does not predict the rate of cognitive decline — cognitive reserve, copathologies, and resilience modulate clinical expression
- Direct-to-consumer testing: Blood-based Aβ42/40 ratio testing has been available directly to consumers since July 2023; isolated abnormal values do not constitute an AD diagnosis
References
- Quinn JF, Gray NE. Fluid biomarkers in dementia diagnosis. Continuum (Minneap Minn). 2024;30(6):1790-1800.
- Jack CR Jr, Andrews JS, Beach TG, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimers Dement. 2024;20(8):5143-5169.
- Barthelemy NR, Saef B, Li Y, et al. CSF tau phosphorylation occupancies at T217 and T205. Nat Aging. 2023;3(4):391-401.
- Barthelemy NR, Salvado G, Schindler SE, et al. Highly accurate blood test for Alzheimer’s disease. Nat Med. 2024;30(4):1085-1095.
- Janelidze S, Bali D, Ashton NJ, et al. Head-to-head comparison of 10 plasma phospho-tau assays in prodromal AD. Brain. 2023;146(4):1592-1601.
- Mielke MM, Dage JL, Frank RD, et al. Performance of plasma phosphorylated tau 181 and 217 in the community. Nat Med. 2022;28(7):1398-1405.
- Meyer MR, Kirmess KM, Eastwood S, et al. Clinical validation of the PrecivityAD2 blood test. Alzheimers Dement. 2024;20(5):3179-3192.
- Kaplow J, Vandijck M, Gray J, et al. Concordance of Lumipulse CSF t-tau/Aβ42 ratio with amyloid PET status. Alzheimers Dement. 2020;16(1):144-152.
- Schindler SE, Gray JD, Gordon BA, et al. CSF biomarkers measured by Elecsys assays compared to amyloid imaging. Alzheimers Dement. 2018;14(11):1460-1469.
- Chatzikonstantinou S, Kazis D, Karantali E, et al. Meta-analysis on RT-QuIC for sporadic CJD. Acta Neurol Belg. 2021;121(2):341-349.
- Concha-Marambio L, Pritzkow S, Shahnawaz M, et al. Seed amplification assay for pathologic alpha-synuclein in CSF. Nat Protoc. 2023;18(4):1179-1196.
- Siderowf A, Concha-Marambio L, Lafontant DE, et al. Alpha-synuclein seed amplification in the PPMI cohort. Lancet Neurol. 2023;22(5):407-417.
- Schindler SE, Karikari TK, Ashton NJ, et al. Effect of race on prediction of brain amyloidosis by plasma biomarkers. Neurology. 2022;99(3):e245-e257.
- Brum WS, Cullen NC, Janelidze S, et al. Two-step workflow based on plasma p-tau217 for amyloid screening. Nat Aging. 2023;3(9):1079-1090.
- Wojdala AL, Bellomo G, Gaetani L, et al. Trajectories of CSF and plasma biomarkers across the AD continuum. Neurobiol Dis. 2023;189:106356.
- Arslan B, Zetterberg H, Ashton NJ. Blood-based biomarkers in AD. Clin Chem Lab Med. 2024;62(6):1063-1069.