Regional variability and genotypic and pharmacodynamic effects on PrP concentration in the CNS

Prion protein (PrP) concentration controls the kinetics of prion replication and is a genetically and pharmacologically validated therapeutic target for prion disease. In order to evaluate PrP concentration as a pharmacodynamic biomarker and assess its contribution to known prion disease risk factors, we developed and validated a plate-based immunoassay reactive for PrP across 6 species of interest and applicable to brain and cerebrospinal fluid (CSF). PrP concentration varied dramatically across different brain regions in mice, cynomolgus macaques, and humans. PrP expression did not appear to contribute to the known risk factors of age, sex, or common PRNP genetic variants. CSF PrP was lowered in the presence of rare pathogenic PRNP variants, with heterozygous carriers of P102L displaying 55%, and D178N just 31%, of the CSF PrP concentration of mutation-negative controls. In rodents, pharmacologic reduction of brain Prnp RNA was reflected in brain parenchyma PrP and, in turn in CSF PrP, validating CSF as a sampling compartment for the effect of PrP-lowering therapy. Our findings support the use of CSF PrP as a pharmacodynamic biomarker for PrP-lowering drugs and suggest that relative reduction from individual baseline CSF PrP concentration may be an appropriate marker for target engagement.


CVs comparing right vs. left brain hemispheres of the same animal, G) mean CVs between animals, and H) normalized response data for brains homogenized with the indicated detergents.
PrP in CSF exhibits enormous inter-individual variability if preanalytical variables are not properly controlled (1), and we hypothesized the same might be true for PrP in brain tissue. We therefore sought to establish conditions for brain homogenization that would enable reliable PrP quantification. We hemisected frozen brains from wild-type mice, and for each animal, both right and left hemispheres were homogenized at 10% wt/vol in either 0.2% or 0.03% wt/vol CHAPS, or RIPA buffer (Pierce 89900, 25 mM Tris HCl pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS). Homogenization in 0.2% CHAPS, just below the critical micelle concentration (2,3), resulted in tight agreement of PrP concentration between hemispheres (mean CV = 3.8%, Figure S1F) and between animals (mean CV = 1.5%, Figure S1G), with >2x higher PrP recovery ( Figure S1H) compared to 0.03% CHAPS or RIPA.
After establishing the final standard curve points and assay concentrations (see Methods and Appendices 1-2), we sought to characterize the assay's performance and determine whether it is fit for purpose for measuring PrP in mouse brain tissue in preclinical drug discovery experiments. We prepared quality control (QC) samples using mouse brain homogenized at 10% wt/vol in 0.2% CHAPS (Table S1), intended to represent brains with 100%, ~50%, 10%, and 0% wild-type levels of PrP (high, mid, low, and negative QCs respectively) and analyzed them at a final 1:200 dilution (1:20 dilution of 10% wt/vol homogenate). A non-GLP validation following FDA guidance (4) determined a dynamic range of 0.05 to 5 ng/mL, with acceptable precision for both calibrators and QCs across this range, except for the low QC sample, which had a high inter-plate CV (32.7%; Table S1). We further conducted a stability assessment for common preanalytical perturbations (Table S2). In contrast with CSF (1), brain homogenate did not disclose a decrease in PrP concentration upon transferring between plastic tubes (Table  S2). Instead, the most important variable was time the brain homogenate spent at room temperature or 4°C, with apparent PrP concentration increasing by 29-56% after 4 hours at either temperature.  We sought to determine across what dilutions the assay might exhibit the property of parallelism, meaning that a sample plated at different dilutions results in the same dilutionadjusted concentration. The adjusted concentrations for all QCs rose at progressively weaker dilutions, even up to the lower limit of quantification of the assay ( Figure S2A). However, the relative concentration of PrP in mid and low QC samples compared to the high QC remained constant regardless of dilution ( Figure S2B). This suggested that while progressive dilution of brain homogenate into assay buffer changes the apparent concentration of PrP in this assay, progressive dilution of endogenous PrP into brain homogenate does not. This was confirmed by preparing a 7-point dilution series of wild-type brain into PrP knockout mouse brain, which resulted in a linear response at a 1:200 final dilution ( Figure S2C). Thus, this assay exhibits a linear response to PrP concentration in brain tissue, provided that brain samples to be compared are plated at the same dilution into assay buffer. For three control human CSF samples, however, parallelism was observed over dilutions from 1:5 to 1:80 ( Figure S2D), in agreement with findings from a commercial PrP ELISA kit (1). Standard curves of five species' recombinant PrP reacted identically in our assay, while a sixth species, Syrian hamster, exhibited ~3-fold lower, but still dose-responsive, reactivity ( Figure S2E; see Figure S3 and Supplemental Discussion). For N=64 human CSF samples analyzed by both cross-species PrP ELISA and the commercially available BetaPrion ELISA kit, the rank order of concentrations was closely preserved (rho = 0.84, Spearman's correlation), while the absolute PrP concentration read out in cross-species PrP ELISA was ~6-fold lower ( Figure 2F; see "Discussion of assay validation" status below).  There are several possible explanations for the reduced reactivity observed for Syrian hamster PrP. The N terminus of our other five constructs contain a retained N-terminal methionine (12), while the Syrian hamster construct contains a cleaved (5) N-terminal methionine followed by a retained glycine ( Figure S3A, red). The 8H4 antibody (6) has been found nonreactive for squirrel monkey PrP, which contains an I182V substitution (human codon numbering; CNVNVTIKQ), as well as for the human mutations H187R and E196K (7), suggesting its epitope spans from at least residue 182 to 196. These residues are invariant among the six species studied here ( Figure S3B, bold). Syrian hamsters do harbor V203I and M205I substitutions (TETDIKIMERV) not found in any other species considered here ( Figure S3B, red), though in order for these to affect 8H4 binding, the epitope would have to be discontinuous, as our MRM data indicate that our ELISA assay shows undiminished activity for PrP with the E200K mutation. Mutation scanning showed that the EP1802Y epitope was disrupted by mutations from residues 218 to 227 (human codon numbering) (8). Syrian hamster PrP in this span is identical to both rat and mouse PrP ( Figure S3C, bold), however it does harbor a nearby I215T substitution not seen in any other species here ( Figure S3C, red). Finally, our Syrian hamster construct contains one additional residue of C-terminal sequence present in the other species' genomes but not included in the recombinant constructs used here.

Figure S2. Parallelism, specificity, cross-reactivity, and comparison with BetaPrion
Although characteristics of this protein looked similar to the other batches employed here ( Figure S3), we also considered technical explanations for the reduced reactivity of our Syrian hamster recombinant PrP. However, its elution curve was typical ( Figure S4A), high purity by Coomassie ( Figure S4B) was confirmed by size exclusion chromatography ( Figure S4C), and identity was confirmed by LC/MS ( Figure S4D). Despite all this, the lower reactivity compared to mouse PrP replicated identically across two plates ( Figure S4E-F).

Discussion of assay validation status
Bioanalytical methods used in drug development should be "fit for purpose," with standards and expectations differing depending on the intended use case (4). The data presented here indicate that our cross-species PrP ELISA is suitable for quantifying target engagement of PrPlowering therapeutics in mouse brain tissue, with certain caveats. Preanalytical variablesparticularly time spent above freezing -must be properly controlled, samples are best compared at the same dilution, and inter-plate variability at the lower end of the dynamic range may be higher than desired, leading to a need for within-plate comparisons or additional technical replicates. PrP in brain homogenate, unlike CSF, does not appear highly sensitive to plastic exposure, perhaps because the high protein, lipid, and detergent content mitigate sticking. Surprisingly, for reasons not yet understood, measurable PrP in brain homogenate does appear to rise with increased time spent above freezing. Based on recombinant PrP binding curves, the assay appears applicable across at least six species of interest for prion research, although we did not perform full validation for all of them. Our data also support analysis of CSF in this assay, though we did not perform full validation in the final assay configuration for this matrix. Importantly, our assay uses a frozen recombinant PrP calibrator curve quantified by amino acid analysis (AAA). The one commercially available PrP ELISA, BetaPrion, uses lyophilized calibrators which appear to have PrP concentrations substantially lower than advertised (1), which limits that assay's capacity for absolute quantification of PrP (Dr. Ashutosh Rao, FDA, Oct 31, 2019). Our assay may be suitable for quantification of PrP in human CSF in a clinical trial setting, but because we are not a GLP laboratory, we did not pursue a formal validation for this use case. One important limitation is that the manufacturer (Abcam) recommends short-term storage at +4°C for the EP1802Y antibody, whereas long-term banking of a single lot of antibody at -80°C would be desirable for long-term analysis of clinical trial samples. We did not assess stability of either of our antibodies at -80°C. Finally, while we demonstrated target engagement of ASOs in prion-infected animals, we have not investigated whether our assay exhibits equal reactivity to PrP Sc as it does to PrP C . Some PrP antibodies, including 8H4, have been reported to exhibit diminished reactivity for PrP Sc depending upon both the prion strain and the capture antibody employed (13).

Quality control of PrP MRM.
Among the five short-term test-retest CSF pairs analyzed, two peptides had high CVs (>30%), but these were peptides that also had high technical replicate CVs (>15%) among these samples (Table S4), perhaps because overall recovery (both of light and 15 N-labeled peptides) was relatively low. For the four peptides with low technical replicate CVs, test-retest CV was also low, supporting the analysis of just one CSF sample from each individual in Figure 3.

Common variants in PRNP.
We possessed only a small sample size of carefully handled CSF samples, and lacked genomewide SNP data to control for population stratification. Nonetheless, in the interest of thoroughness, we chose to ask whether genotypes at two common PRNP variants with high prior probabilities for association with PrP expression showed any obvious correlation with CSF PrP concentration.
The coding variant rs1799990 (M129V) has dramatic effects on prion disease risk, duration, age of onset, clinical presentation, and histopathology across many subtypes of sporadic, acquired, and genetic prion disease (14). For example, the heterozygous genotype is strongly protective against sporadic CJD in a genotypic model (OR = 0.39, P = 1e-135) (15). It is the lead SNP for an eQTL for PRNP in several peripheral tissues but not in any brain region ( Figure S5A). Our cohort contained only one VV individual, and there was no significant difference between CSF PrP in MM and MV individuals, whether all individuals or only mutation-negative controls were included (P = 0.06 or P=0.18, Kolmogorov-Smirnov test; Figure S5B).
Non-coding variant rs17327121, located 72 kb upstream of PRNP, is the lead SNP for an eQTL in cerebellum and cerebellar hemisphere, with no evidence of association with PRNP expression in any other brain region ( Figure S5C). This SNP has not been reported to associate with prion disease risk, although neither it nor any SNP in tight linkage disequilibrium (r 2 > 0.5 in CEU, computed using LDlink (16)) was genotyped or imputed in the largest sporadic CJD GWAS to date. None of the pairwise differences in CSF PrP between genotypes were significant (P > 0.2 for all pairs, Kolmogorov-Smirnov test; Figure S4D).

Figure S5. Common PRNP SNPs and CSF PrP. A) PRNP multi-tissue eQTL data for rs1799990 reproduced from the GTEx browser (gtexportal.org). Positions to the right of the zero indicate that the 129V haplotype is associated with higher PRNP RNA expression in certain tissues than the reference 129M haplotype. The x axis is normalized effect size (NES), which is performed on normalized expression values with no direct biological interpretation (17).
Empirical thresholds for significance (17) in GTEx v8 vary by tissue down to 1e-5; symbols displayed here are as follows: * P < 1e-5, ** P < 1e-6, *** P < 1e-7. B) rs1799990 genotype and CSF PrP for all individuals in our MGH cohort. C) As panel A but for rs17327121. Positions to the left of the zero indicate that the reference allele, C, is associated with higher expression in cerebellum and cerebellar hemisphere than the alternate allele, T. D) rs17327121 genotype and CSF PrP in our MGH cohort. 3. Centrifuge the column device at 1000xG for 2 mins. Flow-through is discarded and the device was placed back into the same falcon tube. 4. 1 mL 1X PBS was added directly on top of the resin. The device is centrifuged at 1000 RCF for 2 mins and the flow-through was discarded. This step is repeated two more times for a total of 3 washes. 5. After the last wash step, the column is removed from the conical tube. Keeping the column upright, the bottom of the column is blotted off with a Kimwipe and is transferred to a clean 15 mL falcon tube. 6. 200 µL of 8H4 Ab is applied directly on top of the resin. After 1 min, 40 µL of 1X PBS is applied as a stacker. 7. The device is centrifuged at 1000xG for 2 mins. The column is discarded and the flowthrough is kept on ice. The volume collected from the device is measured using a pipette and recorded. Purification of Conjugated Protein 11. Remove the bottom closure on a new Zeba column and place into a clean 15 mL falcon tube. The column is kept upright and the cap loosened. 12. Following similar steps in the Material Buffer Exchange section, centrifuge the column device at 1000xG for 2 mins. The flow-through is discarded and the device was placed back into the same falcon tube. 13. 1 mL 1X PBS is added directly on top of the resin. The device is centrifuged at 1000xG for 2 mins and the flow-through was discarded. This step is repeated two more times for a total of 3 washes. 14. After the last wash step, the column is removed from the falcon tube. Keeping the column upright, the bottom of the column is blotted off with a Kimwipe and was transferred to a clean 15 mL falcon tube. 15. The biotinylated 8H4 Ab is applied directly on top of the resin. After 1 min, 40 µL of 1X PBS is applied as a stacker. 16. The device is centrifuged at 1000xG for 2 mins. The column is discarded and the flowthrough is kept on ice. 17. The purified biotinylated 8H4 Ab solution is transferred into a clean 1.5 mL microtube, covered with foil and placed in the 4°C fridge. The final volume collected is measured using a pipette and recorded.