Resilience to autosomal dominant Alzheimer’s disease in a Reelin-COLBOS heterozygous man

We characterized the world’s second case with ascertained extreme resilience to autosomal dominant Alzheimer’s disease (ADAD). Side-by-side comparisons of this male case and the previously reported female case with ADAD homozygote for the APOE3 Christchurch (APOECh) variant allowed us to discern common features. The male remained cognitively intact until 67 years of age despite carrying a PSEN1-E280A mutation. Like the APOECh carrier, he had extremely elevated amyloid plaque burden and limited entorhinal Tau tangle burden. He did not carry the APOECh variant but was heterozygous for a rare variant in RELN (H3447R, termed COLBOS after the Colombia–Boston biomarker research study), a ligand that like apolipoprotein E binds to the VLDLr and APOEr2 receptors. RELN-COLBOS is a gain-of-function variant showing stronger ability to activate its canonical protein target Dab1 and reduce human Tau phosphorylation in a knockin mouse. A genetic variant in a case protected from ADAD suggests a role for RELN signaling in resilience to dementia.

We characterized the world's second case with ascertained extreme resilience to autosomal dominant Alzheimer's disease (ADAD). Side-by-side comparisons of this male case and the previously reported female case with ADAD homozygote for the APOE3 Christchurch (APOECh) variant allowed us to discern common features. The male remained cognitively intact until 67 years of age despite carrying a PSEN1-E280A mutation. Like the APOECh carrier, he had extremely elevated amyloid plaque burden and limited entorhinal Tau tangle burden. He did not carry the APOECh variant but was heterozygous for a rare variant in RELN (H3447R, termed COLBOS after the Colombia-Boston biomarker research study), a ligand that like apolipoprotein E binds to the VLDLr and APOEr2 receptors. RELN-COLBOS is a gain-of-function variant showing stronger ability to activate its canonical protein target Dab1 and reduce human Tau phosphorylation in a knockin mouse. A genetic variant in a case protected from ADAD suggests a role for RELN signaling in resilience to dementia.
Article https://doi.org/10.1038/s41591-023-02318-3 end-stage dementia at age 72 (see the pedigree in Supplementary  Fig. 1). According to the family, she had hypothyroidism, hypertension, depression and cognitive decline at age 58 and developed dementia at age 61. Although less protected than her brother, her MCI began 14 years and her dementia 12 years later than expected for this population. Dementia was preceded by ocular trauma and tibial fracture after a fall, which required surgery under general anesthesia. She died at age 73 of sepsis of pulmonary origin. Additional clinical details about the cases can be found in the supplementary results section of the Supplementary Information.
The male patient was enrolled in the Colombia-Boston biomarker research study (COLBOS) and underwent neuroimaging examinations at the Massachusetts General Hospital (MGH) when he was 73 years old (see Supplementary Table 2 for the demographic information). Amyloid positron emission tomography (PET), measured using cortical-to-cerebellar Pittsburgh compound B (PiB), revealed that the individual's levels of cortical amyloid beta (Aβ) plaque burden were higher (distribution volume ratio (DVR) = 1.77) compared to that of younger impaired carriers from this kindred with a typical age at onset (mean DVR = 1.51 ± 0.13; Fig. 1a,b). Tau tangle burden in the inferior temporal lobe, measured by flortaucipir (FTP), was similar to that seen in younger PSEN1-E280A impaired carriers with typical age at onset (standardized uptake value ratio (SUVR) = 1.78). However, he had relatively limited Tau pathology in the entorhinal cortex (ERC) (SUVR = 1.34; Fig. 1a,c) and in other neocortical regions, such as the posterior cingulate cortex (PCC) and precuneus (SUVR PCC = 1.51; SUVR who remained cognitively unimpaired for nearly 30 years after the expected age at clinical onset 2 .
In this article, we report the clinical, in vivo neuroimaging, genetic and neuropathological characteristics of a male case with the PSEN1-E280A mutation from the same population also presenting with an extreme phenotype of delayed age at clinical onset of ADAD.

Case report
We identified a male carrier of the PSEN1-E280A mutation who remained cognitively intact until age 67. He completed 5 years of formal education in his home country (Colombia) and worked until he retired in his early 60s. He was married and had two children. First cognitive assessment at age 67 revealed limited verbal learning skills and language difficulties in the context of functional independence. The patient was diagnosed with MCI, characterized by short-term memory and verbal fluency decline at age 70.
At age 72, his language had deteriorated further, progressing to mild dementia (Supplementary Table 1). Cognitive decline was preceded by a urinary tract infection-related episode of septic shock. At age 73, he required assistance with basic and instrumental activities of daily living, and met criteria for moderate dementia. He died at the age of 74 years from aspiration pneumonia; his relatives agreed to a brain donation for neuropathological study.
His sister also carried the PSEN1-E280A mutation, had severe dementia when she was first evaluated at age 64 and progressed to 0. 5 Fig. 1a), which usually show greater levels of Tau pathology in PSEN1-E280A carriers who develop MCI and dementia at a typical age 3 (Fig. 1, Supplementary Fig. 2 and Extended Data Figs. 1 and 2). Sparing of the ERC from Tau pathology is a salient feature in the case with RELN-COLBOS that could be critical for the protection phenotype. Measurements of metabolic rate for glucose in the precuneus and whole brain region using 18 F-fluorodeoxyglucose (FDG) PET showed a slightly higher level of glucose metabolism compared to the mean levels of typical MCI carriers from the kindred, who were much younger (Fig.1a,b). He had brain atrophy, measured by magnetic resonance imaging (MRI)-based hippocampal and whole brain volume typical of MCI carriers. These imaging findings suggest that in this patient and in the APOECh homozygote case 2 , protection against ADAD dementia occurred even in the face of high amyloid burden (Fig. 1a,b). Additional imaging and biomarker analyses are reported in the supplementary results of the Supplementary Information.
Our genetic analyses confirmed that the individual was a heterozygote carrier of the PSEN1-E280A mutation (confirmed by single-cell RNA sequencing (scRNA-seq); Supplementary Tables 3 and 4), ruled out the presence of the APOECh mutation (the individual was APOE3/APOE3 and had a normal blood lipid profile; Supplementary Table 5) and identified a heterozygous variant in RELN (chromosome 7:g.103113302T>C, H3447R; Supplementary Fig. 1b), which we named 'RELN-COLBOS', as the most promising missense variant potentially contributing to the phenotype in the protected individual. The RELN-COLBOS variant was only found in the individual and his sister (Supplementary Fig. 1), who also had late-onset cognitive decline. Both were APOE3/APOE3. We focused on the RELN-COLBOS variant because it ranked in the top three candidate genes in the Genomizer priority score analysis and because the encoded protein RELN modulates Tau phosphorylation 4 and is functionally closely related to APOE, the gene mutated in the other case with extreme protection against ADAD 2 . A more detailed description of the genetic analysis is reported in the genetic analysis section of the Supplementary Information.
Our genetic analyses also identified other variants of potential interest in the male individual including: (1) a noncoding variant in the amyloid beta precursor protein (APP) gene (3′-untranslated region, chromosome 21:g.27253263T>C; and (2) a noncoding variant in the calmodulin 2 (CALM2) gene (intronic, chromosome 2:g.47394764C>G). The levels of expression of APP in peripheral blood mononuclear cells from RELN H3447R carrier and noncarriers were very low and not substantially different. Therefore, the significance of the APP variant is uncertain, although it is unlikely to be implicated considering the robust amyloid pathology observed in the individual (Fig.1a,b). This APP variant (g.27253263T>C) was not seen in the APOECh case based on whole-genome sequencing (WGS). The CALM2 variant is predicted as intronic for the canonical transcript encoding a protein with 149 amino acids (P0DP24) and it would result in a Cys9Ser in a predicted longer transcript encoding for a protein with 196 amino acids (E7EMB3). This CALM2 variant has not been reported in the ClinVar database and has low in silico pathogenicity scores. We also could not find a per-variant regulatory score in the Encode database for this variant 5 . The significance of this variant is uncertain, although it is unlikely to be implicated because there is no proteomic evidence of the region of the variant producing any detectable protein in the PeptideAtlas 6 . Further studies may be conducted to examine the extent to which these or other variants contribute to the observed phenotype.
To strengthen our understanding of the genetic implications of the RELN-COLBOS variant, we conducted both in vitro and in vivo molecular genetic studies. These studies aimed to provide further evidence to support our initial findings and increase confidence in our attribution of the variant's genetic effects. The binding of RELN results in the clustering and activation of VLDLr and APOEr2, initiating a signaling pathway that leads to the activation of Dab1 (refs. 4,[7][8][9]. Our studies found that RELN-COLBOS significantly increased Dab1 phosphorylation compared to wild-type (WT) RELN ( Fig. 2a, P = 0.0246) in primary culture mouse cortical neurons (Supplementary Figs. 3 and 4). We also examined RELN-COLBOS binding to its receptors to better understand the potential mechanisms for the observed enhanced signaling. Our findings from cell-free binding assays suggest that RELN-COLBOS did not directly impact binding of RELN to VLDLr or APOEr2, suggesting that other mechanisms were involved ( Supplementary Fig. 5).
The C-terminal region of RELN (CTR-RELN), where the H3447R variant is located, modulates signaling indirectly via interactions with previously unidentified coreceptors on cell membranes 10 . Recently, neuropilin 1 (NRP1) was identified as a coreceptor via binding to a fragment of CTR-RELN consisting of the last six amino acids. This fragment is removed by furin protease. The uncleaved 'long' CTR-RELN interacts with NRP1 and the cleaved 'short' CTR-RELN does not 11 . CTR-RELN has many basic amino acids that are extremely well conserved across species 10 , which we hypothesized could also mediate interactions with glycosaminoglycans (GAGs) on cell membranes. Interactions with GAGs are a rate-limiting step in the interaction of the apolipoprotein E (APOE) protein with some receptors 12 , whereas the role of GAGs in RELN activity has not been fully resolved 10,13 . We used affinity chromatography to examine heparin (a type of GAG) binding of recombinant CTR-RELN peptides. WT CTR-RELN and H3447R CTR-RELN bound to heparin. H3447R CTR-RELN required higher NaCl concentrations to be released from the heparin column, suggesting increased binding affinity (Fig. 2b). Surface plasmon resonance (SPR) measures of kinetic constants showed that the affinity of CTR-RELN H3447R is about twice that observed in the WT CTR-RELN ( Fig. 2c and Supplementary Fig. 6).
We confirmed enhanced interactions of H3447R CTR-RELN with heparin through the use of isothermal titration calorimetry (ITC) ( Fig. 2d and Supplementary Table 6) and biolayer interferometry (BLI) (Fig. 2e). The association constant (K a ) was the main contributor to the and total protein staining (bottom) levels in primary mouse cortical neurons treated with full-length WT RELN or RELN-H3448R, mouse ortholog of RELN-H3447R (mock, P < 0.0029 and WT RELN, P = 0.0246). Data are presented as the mean ± s.e.m. and were analyzed using a Kruskal-Wallis test with Dunn post hoc analysis for multiple comparisons of n = 4 independent biological experiments. b, Spectroscopic analysis of heparin chromatography fractions of WT CTR-RELN (blue plot) and the CTR-RELN-H3447R mutant (green plot) eluted at increasing gradients of NaCl (0.05 M NaCl step gradient). Data are expressed as the percentage of input over 0.4-5 M NaCl gradient fractions. Data show that 0.55 M NaCl can displace WT CTR-RELN binding from a heparin column. The affinity for heparin of CTR-RELN increases in the presence of the H3447R mutation, as suggested by the shift of the peak with maximum height of the eluted fraction from 0.55 to 0.7 M NaCl. n = 3 independent chromatography experiments. The error bars represent the s.e.m. c, Representative sensorgrams of the binding analysis between chip sensors coated with heparin and 0-25 nM increasing concentrations of CTR-RELN variants. Data are expressed as response units per second. The equilibrium disassociation constant (K D ) for each SPR analysis are shown inside the graph and support the difference in affinity binding between heparin and the CTR-RELN variants: H3447R (right plot, K D = 3.75 × 10 −9 M -1 s -1 ) > H3347 (left plot, K D = 6.53 × 10 −9 M -1 s -1 ). The sensorgrams of CTR-RELN with the H3447K and H3447D control variants are reported in Supplementary Fig. 6 for comparison. d, Isothermal calorimetry measurements of short-variant WT CTR-RELN (left) and CTR-RELN-H3447R (right) titrated with 5 μM heparin. Affinity calculations are reported above each plot. e, Binding analysis via BLI between Fc-fusion WT CTR-RELN and H3447R and a heparin-coated biosensor. Association (k a ) and dissociation constants (k d ) were used to calculate the K D that is displayed in the plot. f, Docking of WT CTR-RELN (purple) with a representative heparin molecule (cyan). Amino acids in CTR-RELN that have polar contacts with heparin are highlighted in magenta.

Article
https://doi.org/10.1038/s41591-023-02318-3 difference in equilibrium dissociation constant (K D ) values, and the mutant CTR-RELN had a more negative Gibbs free energy in comparison to WT, suggesting that the COLBOS mutation enables spontaneous CTR-RELN reactions with heparin (Supplementary Table 6). Our nuclear magnetic resonance (NMR) (Figs. 2f and 3a) study revealed that CTR-RELN may have an α-helix structure including a flexible region with a domain we named 'flexibility vertex' (Extended Data Fig. 3a) in the presence of trifluoroethanol, although it may be unstructured under native conditions as revealed by circular dichroism (Extended Data Fig. 3b and Supplementary Table 7). Heparin-binding analyses of a library of mutant CTR-RELN peptides uncovered two binding sites for GAGs, which we named 'α-GAG binding site' and 'β-GAG binding site' (Fig. 3a). The α-GAG binding site is located in the last six amino acids and overlaps with a previously identified binding site for NRP1, which is released by furin. The β-GAG binding site is located upstream of the furin cleavage site and spans amino acids 3,446-3,451. Our research also found that CTR-RELN-COLBOS has a tenfold higher affinity for NRP1 compared to the WT version of CTR-RELN (Supplementary Table 8 Table 9) to increase confidence in our imputation of genetic implication, a common practice to study rare variants. This mouse model is viable, fertile and lacks overt structural and phenotypic brain abnormalities of RELN loss-of-function mutants (for example, cortex lamination defects, abnormal neuronal migration and cerebellar hypotrophy; Extended Data Fig. 4a) 14,15 . Our molecular analyses of cerebellum (CB) from mice with the mRELN-COLBOS confirmed our observation of a gain-of-function (GOF) in 6-12-month old mice for RELN-H3448R as determined by enhanced phosphorylation of Dab1 in males ( Fig. 4a-c, P = 0.0284) and revealed a propensity for the formation of higher molecular protein oligomers for RELN-COLBOS (Supplementary Figs. 8 and 9), a feature that may be critical for enhanced activity 16 . Morphological analysis revealed a mild, yet statistically significant increase in the number of cerebellar neurons in male mice

Fig. 3 | Binding profiles of CTR-RELN variants with heparin. a,b,
Representative in silico models depicting the orientation of basic amino acids in the heparin-binding motif (highlighted with colors) using the WT CTR-RELN NMR structure (a). For the H3447R CTR-RELN variant (b) the model was determined by a homology-based model of WT CTR-RELN that was calculated by Swiss-Model. Position 3,447 orients in the same direction as most other arginines (magenta). Arginines in positions 3,446, 3,453 and 3,457 (cyan) may also interact with heparin as part of the heparin-binding motif but are oriented differently from most basic amino acids. c, RELN peptide variant sequences used for the HPLC analysis. d, Representative chromatographic profiles normalized to the maximum of each eluted peak of short or long CTR-RELN peptides with zero (R3446H, orange), one (WT, blue) or two (H3447R, green) arginines in positions 3,446-3,447, which are predicted to contribute to increased interaction with heparin and are indicated by later peak retention time in the isocratic 1 M KCl elution. e,f, Long (e) or short (f) CTR-RELN peptides with zero (R3446H, orange), one (WT, blue) or two (H3447R, green) arginines in positions 3,446-3,447 showing increased interaction with heparin for the long variants, as indicated by later peak retention time in the isocratic 1 M KCl elution. Conversely, short RELN variants have earlier peak retention times compared to long RELN variants (e). n = 2 independent chromatographic repeats within <0.5 min of representative peaks. Data are expressed as normalized to the maximum emission wavelength for each peak.
This mouse model allowed us to examine sexually dimorphic effects of the RELN-COLBOS variant, a feature that has been described for conditions linked to genetic variation in RELN, including schizophrenia, bipolar disease, autism and Alzheimer's disease [17][18][19][20][21][22] . Increased Dab1 phosphorylation and enhanced oligomerization of RELN were observed only in male mice Extended Data Fig. 5 and Supplementary Figs. 8 and 9). This finding was consistent with our observation of optimal association of RELN-COLBOS with protection against ADAD in a male versus a female case. Homozygosity was required to detect changes in Dab1 and in GSK3β activity showed Tau pathology to a lesser degree (soma of an affected neuron depicted with a dotted line). Bar scale, 100 μm. e, Bar graph for pTau T205 signal intensity values in hTau tg/WT (n = 3 mice) and hTau tg/RELN-H3448R mice (n = 3 mice). The latter showed significantly lower signal intensity. *P = 0.022, two-sided Student's t-test. The error bars represent the s.d. from the mean. f, Representative phenotype observed during the tail elevation test and relative score (0 = severely impaired, 1 = 50% impaired, 2 = normal). g, Tail elevation recorded on WT RELN/Tau-P301L (n = 13 male mice) and RELN-H3448R/Tau-P301L crossed male mice (n = 11 male mice) showed a significantly improved tail elevation score in the presence of the RELN-H3448R variant compared to Tau 11 and 12). Whether this homozygosity requirement is a species-specific effect or whether it depends on other factors such as age is to be examined. Altogether, these data indicate that RELN-H3447R is a GOF (hypermorph) variant.
To attempt to correlate the phenotypes of RELN-COLBOS in mice and humans, we used a crossbreeding strategy, using our knockin mouse model and a tauopathy mouse model, specifically the STOCK Tg(Prnp-MAPT*P301L)JNPL3Hlmc mouse from the Hutton's laboratory, distributed by Taconic Biosciences. This mouse model expresses a mutation in the Tau gene, leading to accumulation of Tau tangles and neuronal loss in specific brain regions, commonly used to study tauopathies 23 . The decision to use this mouse model was based on the known effects of RELN signaling on Tau phosphorylation 24 and our clinical observations of a relative reduction of tauopathy in certain brain regions from postmortem human brain samples of the protected case. Our study found that male P301L mice expressing the RELN-COLBOS allele had a significant reduction of human Tau phosphorylation (pTau205) in the hippocampus (HIC) (Fig. 4d,e) and medulla oblongata (MO) (Supplementary Fig. 13) compared to controls (Fig. 4d). We also observed that the abnormal limb-clasping response, which is a common consequence of tauopathy in mice, was significantly rescued a c d e ERC

Sporadic AD
APOECh pTau CA1  in RELN-COLBOS mice with the Tau transgene ( Fig. 4f,g). Although additional studies of this model are necessary, our findings strongly support our hypothesis that RELN-COLBOS is a GOF mutation and it is probably genetically implicated in the resilience to tauopathy. Postmortem examination of the case with RELN-COLBOS indicated neuropathological evidence of severe AD (brain weight = 745.4 g, classified as CERAD C, Braak VI stage and Thal phase 5) with extensive amyloid and Tau pathology, while showing some hippocampal formation-specific findings (Fig. 5,Extended Data Figs. 6 and 7, and extended description in the supplementary results). Recently, we reported the neuropathological profile of the PSEN1-E280A carrier homozygous for the APOECh mutation. The Christchurch case showed a unique pathological phenotype among cases with PSEN1-E280A, with remarkably low pTau pathology in most brain regions except in the visual primary cortex 25 . In contrast, side-by-side comparisons showed that the case with RELN-COLBOS had more pTau pathology relative to the case with APOECh except in specific regions ( Fig. 6a and Supplementary Fig. 17). Both cases showed extensive Aβ pathology in all evaluated areas, albeit with some individual differences ( Fig. 6b and Supplementary Fig. 17). Microglial morphological assessment of the protected cases indicated that the APOECh microglia were significantly more active in the studied areas (Extended Data Fig. 8).

RELN-COLBOS
We focused our analysis on the HIC and associated cortices because these structures are known to be affected early in AD 26 . Neurons in layer II of the ERC and ERC neurons in general are particularly vulnerable to aging and AD 27 . We measured neuronal density in hippocampal and parahippocampal areas of the cases with RELN-COLBOS, the case with AD-resistant APOECh, cases with typical PSEN1-E280A and cases with typical sporadic AD (Fig. 5b, Extended Data Figs. 9 and 10,Supplementary Figs. 18 and 19 and Supplementary Table 10). We found that lower AD pathology was associated with high neuronal density in the ERC of the case with RELN-COLBOS compared to the case with APOECh or the controls with familial and sporadic AD (Fig. 5c).
This association was not apparent in other subregions such as the CA1 (Fig. 5c and Supplementary Figs. 18 and 19). The cases with RELN-COLBOS and APOECh showed distinctively lower intraneuronal APOE signal than the controls with familial and sporadic AD (Fig. 5d and Supplementary Table 10) whereas the case with RELN-COLBOS showed higher Reelin intracellular signal in the white matter ( Fig. 5e and Extended Data Fig. 10).
Neuropathological findings were consistent with our in vivo neuroimaging observations and confirmed a potential role for the ERC as a target of RELN-mediated mechanisms critical for the resilience to ADAD.

Discussion
We characterized a male heterozygous for the RELN-COLBOS variant who was resilient to the cognitive impairment associated with the PSEN1-E280A mutation. The observation of low Tau pathology and increased neuronal density in the entorhinal cortex compared to other cases with AD implicates this brain region in RELN-mediated mechanisms relevant to protection against AD (Supplementary Table 1 and Figs. 1 and 5b,c). Because the comparative neuropathology was conducted in a relatively low number of cases, the results should not be considered definitive and they are only helpful as informative to generate hypotheses.
A female sibling carrier of the RELN-COLBOS and PSEN1-E280A variants also presented with delayed age at onset of cognitive decline, although with less optimal protection compared to her brother and prolonged end-stage disease. The female sibling had a history of a severe head injury, which required reconstructive surgery, as well as a history of depression and hypothyroidism. These factors may have had an impact on her clinical profile and should be taken into consideration when evaluating her phenotype 2 . In addition, RELN-specific sexual dimorphism may have contributed to her distinct features. We cannot rule out the possibility that other factors, including additional variants, may have contributed to the AD resilience phenotype in RELN-COLBOS carriers. Notwithstanding these potential limitations, others identified RELN as a candidate gene associated with AD pathology in cognitively healthy individuals 28 and DAB1 variants were linked to AD risk in APOE4 homozygotes, further linking the RELN/DAB1 pathway to Alzheimer's 29 .
We previously reported a female case homozygous for APOECh who was resistant to ADAD-related dementia, had widespread amyloid pathology and low Tau pathology in the ERC 2 . Tauopathy was more extensive in the case with RELN-COLBOS compared to the APOECh homozygote, except for the ERC, which was largely spared in both, suggesting resilience in the case with RELN-COLBOS.
The hypermorphic effect of RELN is mild. This is the first known report of a RELN hypermorph and a stronger effect may not support proper development in this critical signaling process. The experimental evidence of a GOF mechanism for the RELN-COLBOS variant, and the fact that a patient with extreme protection against ADAD has it, establish a rationale for genetic implication in the observed phenotypes. Our hypothesis is that RELN-COLBOS is not a neutral variant and probably contributes to the phenotype of the individual. It is possible that other genetic variants may also contribute; if so, these variants and their effects will have to be interpreted and compared to the effect of RELN-COLBOS using the data we have provided.
The APOECh mutation impairs APOE binding to GAG and the APOE receptors 2,30 . Conversely, the RELN-COLBOS variant enhances RELN binding to GAG and NRP1, possibly giving it a competitive advantage for binding to its receptors 7 . Interactions with GAG-containing proteins like heparan sulfate proteoglycans may enhance local concentration of RELN and amplify its signaling effect. Our analyses of the case with RELN-COLBOS revealed a mechanism potentially linking RELN interactions via GAG or other receptors to the protection against AD. Regulation of this RELN-protective pathway, particularly in the ERC, may have a profound therapeutic impact on the resistance to Tau pathology and neurodegeneration, and resilience against cognitive decline and dementia in AD.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41591-023-02318-3.

Clinical assessment
This research complies with all relevant ethical regulations. Written informed consent was obtained from all participants and the local institutional review boards for human research approved the study. This includes the institutional human research ethics committee of the University of Antioquia in Colombia and the Partners Human Research Committee in Boston.

In vivo neuroimaging
We used PiB and FTP PET to image cerebral Aβ and Tau burden, respectively. Structural MRI and both PET scans were conducted at MGH. The FDG PET was performed at the University of Antioquia and the procedures and data analyses were performed as described previously 2 . Briefly, MRI images were processed with FreeSurfer (v.6.0) to identify surface boundaries and standard ROIs. PET data were acquired and processed according to previously published protocols 31,32 , whereby PiB data were expressed as DVR (Logan, 0-60 min) and FTP as SUVR (80-100 min), both using cerebellar gray matter as the reference region. PET images were affine coregistered to each individual's T1 images and visualized using FreeSurfer surface projections (sampled at the gray matter midpoint, surface-smoothed 8 mm). No partial volume correction was applied to PET images for the purposes of this study.

Genetic and molecular studies
We conducted whole-exome (WES), WGS and Genomizer (v.10.10) analysis of the individual to obtain a ranking of AD-related potential risk factors as shown previously 2 and as described in detail in the additional genome sequencing section of the Supplementary Information.

Cell culture
Plasmid encoding for the full-length murine recombinant RELN was a gift from T. Curran via Addgene (plasmid no. 122444 (ref. 14)). The plasmid was subsequently mutagenized to obtain the H3448R mutation homologous of human RELN-H3447R as fee-for-service by Custom DNA Constructs. Primary CD1 brain cortex mouse neurons (catalog no. M-CX-400, Lonza) were cultured in neurobasal medium (Gibco) supplemented with B-27 (Thermo Fisher Scientific), glutaMAX (Gibco) and normocin (Invivogen). Cells were plated on poly-L-lysine-coated (Sigma-Aldrich) wells and treated on day 6 after liquid nitrogen recovery. Treatments with recombinant RELN (WT RELN or RELN-H3448R, 4 μg ml −1 , produced by Innovagen) were incubated for either 5 min or 1 h at 37 °C, 5% CO 2 in the presence of 10 μM MG-132 proteasome inhibitor (catalog no. ab141003, Abcam). Cells were washed in ice-cold PBS (Gibco) and lysed in radioimmunoprecipitation assay ( Fig. 3) and mass spectrometry ( Supplementary Fig. 4). All other western blot images were acquired using the Odyssey Infrared Imaging System and visualized with Image Studio v.

Heparin-sepharose affinity chromatography
We tested changes in binding to heparin of RELN variants chromatographically using an optimized version of a protocol previously published by our laboratory 2 . Briefly, after equilibration of the heparin column (catalog no. 6554-1, BioVision) at room temperature, columns were washed with five volumes of degassed 20 mM Tris-HCl buffer (pH 7.5). Recombinant C-terminal RELN peptides were produced and purified as fee-for-service by Innovagen: WT RKQNYMMNFSRQHGL-RHFYNRRRRSLRRYP and H3447R RKQNYMMNFSRQHGLRRFYNR-RRRSLRRYP. All synthetic peptides listed in Fig. 3 were also produced by Innovagen. One milliliter of 50 μg ml −1 peptide (H3447 or WT and H3447R) was recycled through the column five times; the last flow through was collected for further analysis. The column was washed five times with the same buffer and the protein was eluted using a 0.05 M step gradient of NaCl in 20 mM Tris-HCl (0-1 M, 1 ml per fraction).
To ensure the complete release of the protein, the column was washed with 5 M NaCl 20 mM Tris-HCl. Three independent experiments were conducted for C-terminal WT RELN and H3447R. All eluted fractions were analyzed spectroscopically by reading the absorbance at 280 nm with a NanoDrop 2000 Spectrophotometer. Blank-corrected fractions were subsequently analyzed with Prism 8.

BLI
We used Fc-fusion Reelin CTR peptides because peptides alone would be smaller than the limit of detection in this experimental design. The fusion proteins were obtained by cloning DNA sequences encoding amino acids 3,429-3,460 of RELN-CTR into the pFUSEN-hG2Fc plasmid from InvivoGen. To assess the heparin-Fc-Reelin interaction, the Octet system was used to assess heparin-protein kinetics as a fee-for-service by Precision Antibody. A total of 50 μg ml −1 biotinylated heparin (catalog no. 375054, Merck Millipore) was immobilized on the biosensor tip surface for 300 s on preconditioned biosensors. This was followed by quenching with biocytin at 50 μg ml −1 , baseline buffer dilution for 120 s, 200 nM of analyte (Fc-fusion proteins) for 120 s and disassociation in assay buffer for 120 s. BLI was additionally used to assess NRP1 protein kinetics at 30 °C and 1,000 rpm agitation; 1 mg ml −1 NRP1 (catalog no. 3870-N1-025, R&D Systems) was biotinylated at a 1:2 molar ratio, desalted and immobilized on the streptavidin biosensor tip (Pall Forte-Bio) surface. This was followed by (1)

ITC
Heparin and peptides were dissolved in Dulbecco's PBS buffer (pH 7.22) and spun for 5 min at 10,000 rpm. Then, 2 μl of the 500 μM peptide was successively injected into the cell containing 5 μM heparin at 25 °C with 180 s between injections and at 1,000 rpm stirring speed (MicroCal iTC200). Data were evaluated using the MicroCal iTC200 Evaluation software (Malvern).

Mouse model and in vivo analyses
We generated the RELN-H3448R-Tg knockin mouse model carrying the RELN-COLBOS variant via homologous recombination as fee-for-service (Taconic-Cyagen) by introducing the H3448R (CAC>CGT) mutation into exon 64 in the 3′ homology arm of the RELN gene. Mouse and human RELN have very high homology with 95.2% identity. Human RELN is missing a valine residue at position 15 resulting in it having 3,460 instead of 3,461 amino acids like its mouse counterpart. Structurally, the two proteins have similar domain structure consisting of a signal peptide, an F-spondin-like domain, eight Reelin repeats (RR1-8) and a positively charged sequence at the C terminus. The last 105 amino acids including the region impacted by RELN-COLBOS are identical between mice and human. Gene targeting was obtained using C57BL/6 embryonic stem (ES) cells. Knockin mice were generated by injecting targeted ES cells into blastocysts that were introduced into the foster mothers used to generate the mouse crossings ( Supplementary Fig. 7). Mice were killed by placing them in chambers saturated with CO 2 . HICs, midbrains, frontal cortices, parieto-occipital regions and CBs were dissected and stored at −80 °C upon cervical dislocation, ensuring a postmortem interval of less than 3 min. Brain homogenates were obtained in modified RIPA buffer (Cell Signaling Technology) supplemented with protease (Roche) and phosphatase (Sigma-Aldrich) inhibitors, using a tissue homogenizer (two times, 15 s pulses). Homogenized tissue was then vortexed for 20 s every 10 min for 1 h and centrifuged for 10 min at 10,000 rpm and 4 °C. The soluble protein fraction was then analyzed using a BCA assay (Pierce). Using western blotting, we measured the levels of RELN (clone CR-50), Dab1 (clone G-5, catalog no. sc-271136, Santa Cruz Biotechnology), pDab1 (Tyr232) in the CB of adult male and female mice (6-12 months, n = 3 mice for 6-month-old (m.o.) and 12-m.o. homozygous, n = 4 mice for 12-m.o. WT and heterozygous) either WT, heterozygous or homozygous for the RELN-H3448R mutation. This knockin model was crossed with the Tau P301L model from Taconic Biosciences. Littermates were included for analyses throughout. Additional details about mouse model design and clone selection are reported in Supplementary Table 9 and Supplementary Fig. 7. Eighteen-month-old WT and RELN-COLBOS knockin mice, human Tau P301L transgenic mice and crossbred Tau P301L/RELN-COLBOS mice were tested for the behavioral analyses and their brains were collected and prepared for morphological and immunohistochemical studies (Cresyl violet, Klüver-Barrera and pTau T205 IHC, 1:10,000; catalog no. EPR23505-13, Abcam). Animal study protocols were approved by the Schepens Eye Research Institute and the Institutional Animal Care and Use Committee.

Behavioral studies
We used limb-clasping scoring to assess motor deficits in mice according to a previously published protocol 23 . Briefly, we habituated the animals to the user for 3 days and assessed the escape response when we elevated each mouse by the tail to promote the limb-clasping reflex while standing on a metal grid. We scored from 0 (observed legs in crossed position) to 2 (observed complete opening of the hind limbs).

Neuropathology characterization
The postmortem interval was 210 min after death. The brain presented frontal lobe-predominant atrophy; the weight of the brain and associated structures was 745.4 g and the interuncal distance was 2.3 cm. and specific secondary antibodies: anti-mouse and anti-rabbit (P0260 and P0447, respectively, DAKO); and anti-neuronal nuclei antibody (clone A60; ms host, 1:100; catalog no. MAB377, Merck Millipore). Visualization was achieved with 3,3′-diaminobenzidine (DAB) (Ventana, Roche) and the Ultraview Universal Detection Kit (Roche) according to the manufacturer's instructions. Automatic immunostaining was performed with a Ventana BenchMark XT system (Roche) according to the manufacturer's instructions. Selected brain areas were also stained with Luxol Fast Blue (for myelin staining) and Klüver-Barrera staining. Cresyl violet staining was used for neuronal perikarya. The neuropathological workup was performed by experienced morphologists blinded to the origin of the sample (M.G. and D.S-F.). Sections were scanned using a Hamamatsu NanoZoomer automatic digital slide scanner (Hamamatsu Photonics) and obtained images and ROIs (cortex for cortical areas, whole stained sections for non-cortical areas) at a resolution of at least 1 px μm −1 . Signal intensity, together with particles and total area, were assessed after performing color deconvolution and thresholding in the brown (DAB) color channel using ImageJ (v.1.52p, Article https://doi.org/10.1038/s41591-023-02318-3 NIH) 33 . Neuronal counting was performed manually and normalized by area in selected ROIs in the hippocampal and parahippocampal structures. Information on the statistical analysis is reported in the dedicated section.

Statistical analysis
All data presented are expressed as averaged values and errors are expressed either as s.e.m. or s.d. Statistical analyses were performed using Prism 8 or 9. P < 0.05 and an α of 0.05 were considered statistically significant. We used a Kruskal-Wallis test with Dunn post hoc analysis for multiple comparisons of four independent experiments to compare changes between primary cortical neurons treated with either mock, WT RELN or RELN-H3448R and presented data as the mean ± s.e.m. For the SPR data ( Fig. 3c and Supplementary Fig. 4), we verified the accuracy of the results with a chi-squared test and compared the sensorgrams obtained experimentally with the sensorgrams generated mathematically by the BIAnalysis software (black line). Values ranging from 1 to 2 were interpreted as significant (accurate), and those below 1 as highly significant (highly accurate). Western blot analyses or neuronal counting of Cresyl violet-stained slides were done using a one-way ANOVA followed by Fisher's least significant difference test for multiple comparisons using Prism 9 (v.9.4.1) as reported in the figure legends. The neuropathological data (Fig. 2) were analyzed and graphs were generated with Prism 8 (v.8.1.1) and the R (v.3.6.3) statistical software (R Foundation for Statistical Computing). Distribution and correlation analyses were performed using a Spearman correlation test. The brain color maps were created with the cerebroViz package for R. The statistical significance of all analyses was determined with *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001. Statistical comparisons of two groups were performed using a two-sided Student's t-test.

Inclusion and ethics in global research
The study received institutional review board approval from the MGH, the Mass Eye and Ear and the local ethics committee at the School of Medicine at the Universidad the Antioquia, Medellín, Colombia. This work involved a collaboration between scientists in multiple countries including Colombia, the United States of America and Germany. Contributors from all sites are included as coauthors or in acknowledgements according to their contributions. Researchers residing in Colombia have been involved in study design, study implementation, data ownership and intellectual property as appropriate. The research is locally relevant due to the high prevalence of ADAD. Roles and responsibilities were agreed among collaborators ahead of the research. Local ethics committees approved all research involving human participants.
To prevent any stigmatization, any and all identifying information has been removed to preserve the privacy of individuals. The Colombian team has retained ownership of any and all human biological material shared for research purposes.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
Anonymized clinical, genetic and imaging data are available upon request during working hours, subject to an internal review by F.L., J.  For all quantification, we used one-way ANOVA, followed by Fisher's Least Significant difference (LSD) test for multiple comparisons. All data is presented as average ± s. e. m. A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Reporting on sex and gender
We reported data for men and women according to their reported sex. We reported potential sexual dimorphism for the protective phenotype.

Population characteristics
All research participants belonged to an extended family with autosomal dominant Alzheimer's disease from Colombia. Only individuals older than 28 years old were invited to participate, given previous findings of evidence of amyloid pathology in their brains starting at that age. Both women and men were invited to participate.

Recruitment
All participants were recruited from the Colombia Alzheimer's Prevention Registry, which currently has more than 6,000 members. This registry is maintained by the Group of Neurosciences of the University of Antioquia (PI: Francisco Lopera). Recruited participants were compensated.
Informed consent from subjects include their agreement to not know their genetic status while asymptomatic. Local researchers minimize the possibility of self-selection bias by facilitating access to participation by traveling to the locations where the subjects reside.

Ethics oversight Inclusion & ethics in global research
The study has IRB approval from Massachusetts general hospital, the Mass Eye and Ear of Boston, MA, and the local Ethics Committee at the Universidad the Antioquia. This work involves a collaboration between scientist in multiple countries including Colombia, USA, and Germany. Contributors from all sites are included as co-authors or in acknowledgments according to their contributions. Researchers residing in Colombia have been involved in study design, study implementation, data ownership and intellectual property as appropriate. The research is locally relevant due to high prevalence of ADAD. Roles and responsibilities were agreed amongst collaborators ahead of the research. Local ethics committees approved all research involving human subjects. To prevent any stigmatization any and all identifying information has been removed to preserve subjects privacy. The Colombian team has retained ownership of any and all human biological materials shared for research purposes.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
For clinical data: Sample size of subjects for comparison studies was defined solely by the availability of subjects with relevant information studied under similar conditions. For data presented in Figure 1: According to IRB regulation, PET imaging could not involve repeated measurements. For all other data: sample size was stated in both methods and figure legends for each biological experiments. Sample size calculations are not applicable for the functional assays and structural determination we used. For western blotting analyses we conducted a power analysis to determine the minimum statistically significant sample size. For cell-free experiments we arbitrarily determine sample size.
Data exclusions No data was excluded.

Replication
For Clinical data, 11 impaired and 18 unimpaired carriers were analyzed as reported in Supplementary Table 2.
Brain imaging for the male subject was only conducted once because of feasibility and availability of the patient. Data on amyloid and tau pathology was validated postmortem via immunohistology. Accordingly, all attempts for replication were confirmed. We thoroughly explored independent experimental designs to determine RELN biology and functional interactions with heparin. Data presented in Fig. 2A is expressed as mean ± s. e. m of n = 4 independent biological experiments tested on individual blots. For SPR and BLI data, a range of untagged (SPR) or Fc-fusion peptides (BLI) were used to determine the KD. The determined rate constants are independent of sample concentration. Data presented in Fig. 4A, Supplementary Fig. 8 , 9, 11, 12, and  Gravity heparin-affinity chromatography data (Fig. 2 B) was expressed as average -/+ SEM of n = 3 independent sample runs from different nature portfolio | reporting summary The Fc-fusion peptides were produced in HEK293 cells as a fee-for-service by Innovagen AB (Lund, Sweden). HEK293 cells were purchased by Innovagen from Invitrogen (cat. 51-0029). Primary CD1 brain cortex mouse neurons obtained from day 14/15 embryos of mixed sex (M-CX-400, Lonza) and Flp-In T Rex 293 mammalian cells (R78007, Thermo Fisher Scientific) cell line.

Authentication
All cell lines were authenticated by resistance/susceptibility to the antibodies listed by the providers Innovagen AB (HEK293) and Lonza (M-CX-400, cell type further confirmed via IHC and ELISA using GFAP, MAP2, and Tuj, IHC only) and Thermo Fischer

Wild animals
No wild animals were used in the study.

Reporting on sex
When possible we reported data for both, male and female sex. Findings were statistically significant in males when indicated by the statistical analysis.
Field-collected samples No field collected samples were used in the study.

Ethics oversight
Animal studies were approved by The Schepens Eye Research Institute and the Institutional Animal Care and Use Committee (IACUC).
Note that full information on the approval of the study protocol must also be provided in the manuscript.