CSF progranulin increases in the course of Alzheimer's disease and is associated with sTREM2, neurodegeneration and cognitive decline

Abstract

Progranulin (PGRN) is predominantly expressed by microglia in the brain, and genetic and experimental evidence suggests a critical role in Alzheimer's disease (AD). We asked whether PGRN expression is changed in a disease severity-specific manner in AD We measured PGRN in cerebrospinal fluid (CSF) in two of the best-characterized AD patient cohorts, namely the Dominant Inherited Alzheimer's Disease Network (DIAN) and the Alzheimer's Disease Neuroimaging Initiative (ADNI). In carriers of AD causing dominant mutations, cross-sectionally assessed CSF PGRN increased over the course of the disease and significantly differed from non-carriers 10 years before the expected symptom onset. In late-onset AD, higher CSF PGRN was associated with more advanced disease stages and cognitive impairment. Higher CSF PGRN was associated with higher CSF soluble TREM2 (triggering receptor expressed on myeloid cells 2) only when there was underlying pathology, but not in controls. In conclusion, we demonstrate that, although CSF PGRN is not a diagnostic biomarker for AD, it may together with sTREM2 reflect microglial activation during the disease.

Keywords: Alzheimer's disease; TREM2; biomarker; microglia; progranulin.

Figure 1. Association of CSF PGRN with mutation status, age and gender

1. CSF PGRN is increased in mutation carriers (MC) compared to non‐carriers (NC).

2. CSF PGRN is not associated with age in either NC or MC.

3. CSF PGRN is increased in males compared to females.

4. CSF PGRN levels do not differ among MC participants carrying a PSEN1, PSEN2 or APP mutation.

Data information: The blue or red bars in (A), (C) and (D) represent the mean and the standard deviation (SD). Group comparisons were assessed by a linear model adjusting by age, gender and APOE ε4 status. The solid lines in (B) indicate the regression line for each of the groups and the 95% confidence interval (CI) calculated by a linear model adjusting by gender and APOE ε4 status. The standardized regression coefficients (β) and the P‐values are also shown. In graph (B), the individual values are not shown in order to protect participants' confidentiality. All analysis and graphs are performed excluding 3 PGRN values outliers. Including the outliers in the analysis rendered similar results (Appendix Table S1). APP, amyloid precursor protein; CSF, cerebrospinal fluid; ns, non‐significant; PSEN1, presenilin 1; PSEN2, presenilin 2.

Figure 2. Changes in CSF PGRN as a function of EYO

1. CSF PGRN as a function of EYO in mutation carriers (MC, red) and non‐carriers (NC, blue). The solid lines indicate the regression line for each of the groups and the 95% confidence interval (CI) calculated by a linear model adjusting by gender. The interaction term of mutation status and EYO is significant (P = 0.041), also when including PGRN outliers and participants with EYO > +20 (P = 0.030). Individual data points are not displayed to prevent disclosure of mutation status.

2. The graph depicts the standardized differences in CSF PGRN between MCs and NCs as a function of EYO, in the context of other biomarker and cognitive changes. The curves were generated by the linear model that best fit each marker (see Statistical analysis section and Appendix Table S2). CSF PGRN is significantly increased in MC compared to NC 10 years before the expected symptom onset (shadowed area) after brain amyloidosis and brain injury (as measured by CSF T‐tau) have started, and shortly before CSF sTREM2 starts to increase.

Data information: Aβ_1–42: amyloid‐β 42; CSF, cerebrospinal fluid; MC, mutation carrier; MMSE, Mini‐Mental State Examination; NC, non‐carrier; T‐tau, total tau.

Figure 3. CSF PGRN levels across the Alzheimer's continuum

Scatter plot representing the levels of CSF PGRN in healthy controls (depicted in blue) and the different stages of the Alzheimer's continuum (depicted in red). The blue and the red bars represent the mean and the standard deviation (SD). The analysis and graphs were performed excluding CSF PGRN outliers (1 “healthy control”, 1 “Preclinical AD A+/TN−”, 4 “AD CDR = 0.5” and 1 “AD CDR = 1”). Including them yielded a similar result, and CSF PGRN was still significantly higher in the “AD CDR = 1” group compared to the “healthy controls” (P = 0.001) and “Preclinical AD A+/TN−” (P = 0.0001) groups. P‐values were assessed by a one‐way analysis of covariance adjusted for age, gender and APOE ε4, followed by Bonferroni corrected pair‐wise post hoc comparisons. A: amyloid‐β biomarker status; AD: Alzheimer's disease; CDR: clinical dementia rating; CSF, cerebrospinal fluid; N, neurodegeneration biomarker status; T: tau pathology biomarker status.

Figure 4. CSF PGRN as a function of cognitive function

Scatter plots representing the association of CSF PGRN with different cognitive tests. Only the subjects of the Alzheimer's continuum group (n = 474) were included. In all tests studied, higher levels of CSF PGRN were associated with worse cognitive performance (namely lower scores in ADNI‐Mem, ADNI‐EF and MMSE and higher scores in ADAS‐Cog11, ADAS‐Cog13 and CDR‐SB). The analysis and the graphs are excluding PGRN outliers; including them rendered similar results (Appendix Table S6). Each point depicts the value of CSF PGRN and the corresponding cognitive test score of a participant. The solid lines indicate the regression line and the 95% confidence interval (CI) calculated by a linear model (Model 1, unadjusted). Table 5 shows the standardized regression coefficients (β) and the P‐values calculated by different models. ADAS‐Cog, Alzheimer's disease Assessment Scale—cognitive subscale; ADNI‐Mem: ADNI memory composite score; ADNI‐EF: ADNI executive function composite score; CDR‐SB: clinical dementia rating sum of boxes; MMSE, Mini‐Mental State Examination.

Figure 5. CSF PGRN as a function of neuroimaging biomarkers

1. A, B

   Scatter plots representing the association of CSF PGRN with temporo‐parietal FDG‐PET uptake (A) and total hippocampal volume (B) within the subjects of the Alzheimer's continuum group (n = 474). Each point depicts the value of CSF PGRN and the corresponding neuroimaging biomarker of a participant. The solid lines indicate the regression line and the 95% confidence interval (CI). The regression coefficients (β) and the P‐values calculated by a linear model adjusted for age, gender, APOE ε4 and education. FDG‐PET: fludeoxyglucose positron emission tomography; SUVR: standardized uptake value ratio.

Figure EV1. CSF PGRN levels in the A/T/N framework

1. Scatter plot representing the levels of CSF PGRN for each of the four biomarker profiles within each clinical staging, as defined by CDR. CDR = 1 stage comprises some biomarker profiles that do not contain enough participants to perform statistical analysis but are nevertheless shown in the figure for completeness. Each biomarker category is represented in a different colour. Healthy controls are depicted in blue, Alzheimer's continuum category in red and SNAP category in green. Purple depicts biomarker profiles not assigned in any category in the present study.

2. Scatter plot grouping the three biomarker categories: healthy controls, all the participants belonging to the Alzheimer's continuum category and the suspected non‐Alzheimer's pathophysiology (SNAP) category.

   Solid bars represent the mean and the standard deviation (SD). P‐values were assessed by a one‐way analysis of covariance adjusted for age, gender and APOE ε4, followed by Bonferroni corrected pair‐wise post hoc comparisons. The analysis and graphs were performed excluding PGRN outliers (1 “healthy control”, 1 “Preclinical AD A+/TN−”, 4 “AD CDR = 0.5”, 3 “CDR = 0.5 A−TN−”, 1 “CDR = 0.5 A+TN−”, 1 “AD CDR = 1”). Including them yielded a similar result.

Data information: A: amyloid‐β biomarker status; AD: Alzheimer's disease; CDR: clinical dementia rating; CSF, cerebrospinal fluid; N: neurodegeneration biomarker status; SNAP: suspected non‐Alzheimer's pathophysiology; T: tau pathology biomarker status.

Figure EV2. Diagnostic accuracy of CSF PGRN as an AD biomarker

ROC analysis was performed to test the accuracy to discriminate between ADAD mutation carriers and non‐carriers of the DIAN study (A) and late‐onset AD with mild dementia (CDR = 1) from healthy controls of the ADNI study (B). The areas under the curve (AUCs), their 95% confidence interval (CI) intervals and the P‐values are reported. Optimal cut‐offs were derived based on the Youden index, and the sensitivity and specificity were calculated based on these cut‐offs. AD: Alzheimer's disease; ADAD: autosomal dominant Alzheimer's disease; AUC: area under the curve; CDR: clinical dementia rating; HC: healthy controls; MC: mutation carriers; NC: non‐carriers.

Figure 6. Association of CSF PGRN with AD core CSF biomarkers in ADAD (DIAN)

1. A–H

   Scatter plots representing the associations of CSF PGRN with CSF sTREM2 and each of the AD CSF core biomarkers (T‐tau, P‐tau_181P and Aβ₁₋₄₂) in non‐carriers (NC, blue; A, C, E and G) and in mutation carriers (MC, red; B, D, F and H). Each point depicts the value of CSF PGRN and the corresponding biomarker of a subject and the solid lines indicate the regression line and the 95% confidence interval (CI) for each of the groups. The standardized regression coefficients (β) and the P‐values are shown and were computed using a linear model adjusting for age, gender and APOE ε4. The sample contained some outliers (defined as 3 SDs below or above the group mean) of the CSF core markers of AD. The results shown in the figure are excluding these outliers. We also performed the analysis including these outliers which yielded similar results (Appendix Table S8). Aβ_1–42: amyloid‐β 42; T‐tau: total tau; P‐tau: tau phosphorylated at threonine 181.

Figure 7. Association of CSF PGRN with AD core CSF biomarkers in late‐onset AD (ADNI)

1. A–L

   Scatter plots representing the associations of CSF PGRN with CSF sTREM2 and each of the AD CSF core biomarkers (T‐tau, P‐tau_181P and Aβ_1–42) in healthy controls (blue; A, D, G and J), Alzheimer's continuum (red; B, E, H and K) and “suspected non‐Alzheimer's pathophysiology (SNAP)” groups (green; C, F, I and L). Each point depicts the value of CSF PGRN and the corresponding biomarker of a subject, and the solid lines indicate the regression line and the 95% confidence interval (CI) for each of the groups. The standardized regression coefficients (β) and the P‐values are shown and were computed using a linear model adjusting for age, gender and APOE ε4. The sample contained some outliers (defined as 3 SDs below or above the group mean) of the CSF core markers of AD, and the analysis including these outliers yielded similar results (Appendix Table S8). The Aβ_1–42 values used for the association test are those based on an extrapolation curve since the upper technical limit is 1,700 pg/ml. We also tested the associations with Aβ_1–42 values truncated at the upper technical limit and the result was similar. Aβ_1–42: amyloid‐β 42; T‐tau: total tau; P‐tau: tau phosphorylated at threonine 181; SNAP: suspected non‐Alzheimer's pathophysiology.