Immunogenomics Parameters for Patient Stratification in Alzheimer’s Disease

Abstract

Background:

Despite the fact that only modest adaptive immune system related approaches to treating Alzheimer’s disease (AD) are available, an immunogenomics approach to the study of AD has not yet substantially advanced.

Objective:

Thus, we sought to better understand adaptive immune receptor chemical features in the AD setting.

Methods:

We characterized T-cell receptor alpha (TRA) complementarity determining region-3 (CDR3) physicochemical features and identified TRA CDR3 homology groups, represented by TRA recombination reads extracted from 2,665 AD-related, blood- and brain-derived exome files.

Results:

We found that a higher isoelectric value for the brain TRA CDR3s was associated with a higher (clinically worse) Braak stage and that a number of TRA CDR3 chemical homology groups, in particular representing bloodborne TRA CDR3s, were associated with higher or lower Braak stages. Lastly, greater chemical complementarity of both blood- and brain-derived TRA CDR3s and tau, based on a recently described CDR3-candidate antigen chemical complementarity scoring process (https://adaptivematch.com), was associated with higher Braak stages.

Conclusion:

Overall, the data reported here raise the questions of (a) whether progression of AD is facilitated by the adaptive immune response to tau; and (b) whether assessment of such an anti-tau immune response could potentially serve as a basis for adaptive immune receptor related, AD risk stratification?

Keywords

Alzheimer’s Disease Sequencing Project (ADSP)Braak stages blood and brain sample exomes CDR3-tau chemical complementarity T-cell receptor recombination reads

INTRODUCTION

Alzheimer’s disease (AD) is an irreversible neurodegenerative disease where the pathophysiology is still not fully understood. However, in all patients, the processing of amyloid-β protein precursor (AβPP) is altered, leading to deposition and aggregation of amyloid-β, in turn associated with hyperphosphorylation of tau (MAPT) and neurofibrillary tangles. Keeping in mind this pathophysiology, investigators have assessed potential contributions from other patient features, including genetics, the environment, and the immune system.

In particular, recent studies have identified low levels of chronic inflammation in aging, known as “inflammaging”, as a risk factor for decreased brain function and AD [1]. This neuroinflammation is thought to exacerbate AD via the brain’s innate immune cells, including the microglia and macrophages. Furthermore, a possible role of the adaptive immune response has been partly elucidated, as a recent study has shown gamma T-cell receptor (TCR) specificities related to AD, in both the brain and blood of AD patients [2]. Another recent study has shown an increased response by antigen-experienced T-cells in the blood and cerebrospinal fluid of patients affected by age-related neurodegeneration, noting an interesting TCR specificity for Epstein Barr virus antigens [3]. Liu et al. [4] have specifically noted an association of anti-amyloid-β antibodies with cognitive decline. Interestingly, genetic variants are associated with altered regulation of macrophage specific genes, raising questions about roles for antigen presenting cells, as part of the role of the adaptive immune system in the development of AD [5].

Appreciating a potential role for the adaptive immune system in AD, investigators have considered modification of the immune response using strategies such as vaccination and antibody-therapy against known antigens, such as amyloid-β, as potential disease-modifying processes for AD. However, the previous widely-assessed vaccination trial, where amyloid-β was the intended target, was terminated early as a result of the adaptive immune receptor (IR) response leading to meningoencephalitis in some patients [6]. Other vaccination trials with both amyloid-β and tau are underway, but concerns over safety and effectiveness remain. For example, the FDA recently approved, amidst controversy, treatment with an antibody, aducanumab, targeting amyloid-β [7].

In sum, a further understanding of the potential role of the adaptive IRs in AD is needed. Thus, we characterized TCR recombination reads from blood- and brain-derived, AD patient exome (WXS) files, including an approach that employs a novel chemical complementarity assessment for CDR3s and candidate antigens designed for the big data setting [8]. Results indicated, among other things, a correlation of higher AD+ patient Braak stages with increased chemical complementarity scores for both the blood- and brain-derived TRA CDR3s and tau.

METHODS

Alzheimer’s disease sequencing project (ADSP) WXS files [9]; and biospecimen and survival data files

AD WXS files were downloaded from the ADSP Amazon Web Services (AWS) S3 bucket to this research group’s AWS bucket in binary alignment map (BAM) format with authorization via database of genotypes and phenotypes (dbGaP) approved project 23565 (ADSP project ID: 10229). The download protocol is at https://dss.niagads.org/documentation/. To download only the WXS file slices needed for recovery of the IR recombination reads, we used an original script freely available at https://github.com/thudaUSF/ADSP-dl-process. This script generates a unique URL for each WXS file using boto3, which allows SAMtools, an S3-aware alignment reader, to download and process a section of the aligned data (e.g., the 930 kbp region of TRA, on chromosome 14), without having to download the full WXS file. The majority of samples in the ADSP study were derived from blood or brain and from either AD cases or controls. These samples were taken at multiple study sites and sequencing centers, where steps were taken to reduce variance in data. Notably, each case was associated with only one blood or brain sample, i.e., there are no case-matched blood and brain samples in the dataset. This report utilizes only the blood or brain WXS samples for AD cases (AD+) that have a Braak stage, representing a staging mechanism based on histological features, including the location of brain lesions and density of neurofibrillary tangles and neuropil threads [10]. These stages progress from 0, transentorhinal (I-II), limbic (III-IV), and end with isocortical stages (V-VI) [11]. This staging process relies on a neuropathologist’s opinion and above algorithm, blinded to age and other clinical data. Braak staging was chosen as the dependent variable for this study, as multiple previous studies have shown that Braak stages have correlated well with clinical symptoms of AD. Such clinical symptoms include cognitive measures, such as Mini-Mental State Examination scores [12, 13], but also other behavioral symptoms beginning with symptoms such as agitation, anxiety, and depression in stages I-II and culminating in cognitive decline in stages V-VI [14]. Thus, the dataset used in this study includes 1,113 blood cases and 1,552 brain cases, for a total of 2,665 cases with WXS files and an associated Braak stage. The file used for downloading all of the WXS files, termed the sample file, was created by combining information in the manifest and phenotype files obtained from the ADSP AWS bucket (Supplementary Table 1).

Obtaining the adaptive IR recombination reads

The IR recombination reads were retrieved from the WXS files for four T-cell receptor genes (TRA, TRB, TRD, TRG), and three immunoglobulin genes (IGH, IGK, IGL) for AD+ blood and brain WXS files (Supplementary Tables 2 and 3). Only IR recombination reads with both a verifiable V- and J-gene segment ID on one read were used. This was done using a previously and extensively described algorithm and script [15 –17], freely available at https://github.com/bchobrut-USF/blanck_group, with a container version at https://hub.docker.com/r/bchobrut/vdj. In both cases, a readme file is available at the web links. Briefly, three processes were applied to obtain and further evaluate all IR recombination reads: (i) a low-stringency search of the reads constituting the WXS files, using 10-mers that matched germline IR, V- and J- nucleotide sequences in multiple places in the V- and J-gene segments; (ii) a higher stringency search that required a 90% match of 19 nucleotides to a V- and J-gene segment on one read; and (iii) translation of the gene segment recombination junction, i.e., the CDR3 region, and retention of only those IR recombination reads that lacked a stop codon or an out-of-frame joining sequence (Supplementary Tables 2 and 3). The following sections of the WXS files were targeted based on the hg38 version of the reference genome: TRA: chr14:21621904-22552132; TRB: chr7:142299011-142813287; TRG: chr7:38240024-38368055; TRD: chr14:22422546-22466577; IGH: chr14:105586437-106879844; IGK: chr2:88857361-90235368; IGL: chr22:22026076-22922913.

Identifying the single value physicochemical parameters of the TRA CDR3s associated with Braak stages

Utilizing TRA CDR3 amino acid (AA) sequence data from the above brain and blood samples, single value physicochemical parameters, for example, aromaticity, fraction-negative, and secondary-structure, were assessed using the algorithms of the Pappu laboratory [18] and a script available at https://github.com/bchobrut-USF/blanck_group, package B. An Excel pivot table was used to obtain the average of each physicochemical parameter for the TRA CDR3 AA sequences representing each case (Supplementary Table 4). The analyses that linked the case IDs, based on physicochemical parameters, to the Braak stages were conducted using an original script available at https://github.com/thudaUSF/ADSP-dl-process. This script utilized both Python and rpy2 to first align cases with a Braak stage, then perform a Kruskal-Wallis (KW) test by ranks to determine whether the distribution of a physicochemical parameter is the same among the Braak stage groups (Results, Supplementary Table 5). Samples with Braak stages III and IV were merged into one group for analyses due to sample size constraints. For all KW tests, the post-hoc Dunn test was performed with the Bonferroni correction.

Establishment of CDR3 homology groups

CDR3 homology groupings were based on calculations using the chemical complementarity scoring algorithms described in [8], which were designed for matching CDR3s with candidate antigens. By inverting the sign, positive or negative, for an electrostatic complementarity score, the complementarity scoring process becomes effectively a homology grouping process, referred to as an Electrostatic group. For this report, this process also included a separate homology grouping process based on Uversky hydropathy values, as in [19], to obtain homology groups based on hydrophobicity (termed, Hydro group). The process also creates homology groups based on a combination of electrostatic and hydrophobic CDR3 AAs (termed, Combo group). Then, a distance matrix of homology grouping scores for all distinct TRA CDR3 domains in the dataset was established using the DistanceMatrix function of the skbio python package (https://scikit-bio.org/docs/0.5.2/generated/generated/skbio.stats.distance.DistanceMatrix.html). The phylogenetic grouping of CDR3s was performed using the distance matrix input and the neighbor joining algorithm of the skbio.tree python package (https://scikit-bio.org/docs/0.5.2/generated/skbio.tree.nj.html) followed by the Tree algorithm of the ete3 python package (https://etetoolkit.org/). Duplicate CDR3s were removed prior to input, to minimize inclusion of any public CDR3s in the analyses, leaving 515 unique blood-derived cases and 89 unique brain-derived cases. The chemical homology groups for AD+ blood and brain TRA CDR3s were associated with the case ID’s Braak stage, and the Braak stage for the case IDs in the chemical homology group were compared to the Braak stage for the case IDs not in the homology group, referred to as the “out-group”, using the Mann-Whitney U (MW) test (Results, Supplementary Tables 6 and 7). The homology grouping was performed using an original web tool, adaptivematch.com, which is publicly accessible and has instructions for preparing input files.

Chemical complementarity scoring

The chemical complementarity scoring involved a sliding window approach to assess multiple alignments of a CDR3 and a candidate antigen with regard to chemical features, as described in detail in [8]. The supplementary material of [8] includes an mp4 file that provides a user-friendly video of the alignment and chemical complementarity scoring process. For any given CDR3-candidate antigen pair, the highest calculated, chemical complementarity score was used in subsequent analyses. This process was facilitated by the use of the web tool, adaptivematch.com (Supplementary Table 8; for example adaptivematch.com input files, see Supplementary Files 1 and 2). For this study, tau was first matched, using the adaptivematch.com web tool, as a whole protein sequence, with blood and brain TRA CDR3s, for the chemical complementarity scoring. Then, tau was split into approximately seven equal segments, and the chemical complementarity of the CDR3s with each segment of tau was determined using the same electrostatic, hydrophobic, or combo calculations referenced above.

RESULTS

Single value physicochemical assessments for the brain, AD+ TRA CDR3s

Adaptive IR reads were mined (Methods) from the WXS files representing 1,113 blood cases and 1,552 brain cases, and the resulting IR recombination read counts are tallied in Table 1. Three thousand, nine hundred seventy-seven TRA CDR3s from blood were recovered, with at least one CDR3 from 46.36% of the 1,113 blood cases. One hundred sixty-three TRA CDR3s from brain were recovered, with at least one CDR3 from 6.06% of the 1,552 brain cases. TRA recombination reads represented the greatest number of recoveries for both blood and brain datasets, compared to recoveries of recombination reads representing the other IRs (TRB, TRD, TRD, TRG, IGH, IGK, IGL). The Braak stage co-variates for the cases representing the TRA recombination read recoveries from brain and blood are in Tables 2 and 3, respectively. The subsequent analyses largely focused on the TRA CDR3 reads.

Table 1

Numbers of adaptive immune receptor recombination reads, representing productive recombinations, recovered from the WXS files of the Alzheimer’s disease dataset

Immune receptor	Blood recovery count	Unique cases out of 1,113 total blood cases	Brain recovery count	Unique cases out of 1,552 total brain cases
TRA	3,977	46.36%	163	6.06%
TRB	1,447	31.72%	51	1.74%
TRD	63	2.88%	3	0.13%
TRG	2240	38.36%	88	2.96%
IGH	716	25.97%	6	0.26%
IGK	633	28.30%	3	0.13%
IGL	679	18.87%	11	0.19%

Table 2

Characteristics of AD+ cases with brain-sample WXS samples with a TRA recovery, classified by Braak stage

Braak stage	N	Age at Onset Mean (SD)	Sex % Female	APOE4 carrier %
III-IV	20	75.32 (7.72)	45	35
V	39	70.47 (6.06)	46.1	69.2
VI	34	69.35 (5.66)	47.1	79.4
Total	93	72.13 (6.56)	46.2	65.6
p	0.01 (KW)	0.885 (MW)	0.002 (MW)

Table 3

Characteristics of AD+ cases with blood sample WXS files with a TRA recovery, classified by Braak stage

Braak stage	N	Age at Onset Mean (SD)	Sex % Female	APOE4 carrier %
0-II	29	75.32 (7.72)	41.4	10.3
III-IV	85	75.7 (7.66)	47.1	31.8
V	157	73.7 (7.60)	39.5	49.7
VI	245	70.63 (7.31)	51.8	55.1
Total	516	72.61 (7.72)	46.7	47.1
p	< 0.001 (KW)	0.078 (MW)	< 0.001 (MW)

We next tested the hypothesis that single value physicochemical parameters for the TRA CDR3s could be associated with the histological Braak stages, which positively correlate with the severity of the disease. Results indicated that higher isoelectric values, for the TRA CDR3 AA sequences recovered from the brain WXS files, were associated with the higher Braak stages (Table 4, Fig. 1). Additional single value physicochemical parameters were similarly assessed (Supplementary Table 4), and notably, physicochemical parameters for the brain TRA CDR3 AAs related to isoelectric point were consistent with higher isoelectric point values being associated with higher Braak stages (Supplementary Table 5).

Table 4

Isoelectric point (physiochemical) values for TRA CDR3s from brain-derived AD+ samples correlated with Braak stage distinctions (Supplementary Table 5).

AD brain or blood as TRA CDR3 source	n (number of cases) Braak group 0–II	n Braak group III, IV	n Braak group V	n Braak group VI	KW p	Associated with a higher (worse) Braak stage group	Braak groups compared in a Dunn post-hoc test (comparison across the hyphen)	Post-hoc p
Brain tissue	1 (dropped)	20	39	34	0.044^*	Increased isoelectric	III, IV –V	0.13
						point	III, IV –VI	0.045^*
							V –VI	1
Blood samples	29	85	157	245	0.53	–	–	–

Fig. 1

Violin plots of isoelectric point values for TRA CDR3s from brain-derived AD+ samples correlated with Braak stage distinctions (Table 4).

TRA CDR3 AA sequence homology grouping

Given the above conclusion of TRA CDR3 AA sequences having similar single value physicochemical features associated with higher Braak stages, we employed a more extensive AA sequence, chemical homology grouping process for the TRA CDR3 AA sequences recovered from the AD+ WXS files (Methods; adaptivematch.com). This approach led to the identification of groups of TRA CDR3 AA sequences that had a specific level of homology based on electrostatic charge, based on hydrophobicity, or based on the integration of both electrostatic charge and hydrophobicity in forming the homology group, i.e., in this latter case, there was an algorithmic incorporation of both chemical features in establishing the homology group [8] (Methods). Results indicated that multiple homology groups representing both blood- and brain-derived TRA CDR3 AA sequences were associated with the distinct Braak stages (Table 5). These homology groups ranged, in the percentage of cases contributing to the homology group, from 1–26% of the collection of cases representing the TRA CDR3 AA sequence recoveries. While some TRA CDR3 AA sequence homology groups were associated with a higher Braak stage, and therefore a worse clinical condition, such as the Hydro homology group 547 (Table 5), other homology groups were associated with a lower Braak stage. (See Supplementary Table 7 for a complete list of TRA CDR3 AA sequence homology groups and their relationships to Braak stages). The TRA CDR3 AA sequences for two examples of the homology groups, associated with lower Braak stages and representing blood- and brain-derived WXS files, are listed in Table 6.

Table 5

Chemical homology groups representing the AA sequences of the blood and brain-derived TRA CDR3s were associated with distinct Braak stages (Supplementary Tables 6 and 7)

Homology grouping parameter and TRA CDR3 source	Homology group (Supplementary Table 7)	Designation of outside-homology-group or in-homology-group for the parameter indicated two columns to the left	Number of cases either out or in the chemical homology group	Mann-Whitney p	Associated with the higher (worse) Braak stage group
Hydro, blood	Group 547	All remaining TRA CDR3s	491	0.040	In-group
		Homologous TRA CDR3s	24
Hydro, brain	Group 39	All remaining TRA CDR3s	84	0.021	Out-group
		Homologous TRA CDR3s	5
Electrostatic, blood	Group 416	All remaining TRA CDR3s	379	0.025	Out-group
		Homologous TRA CDR3s	136
Electrostatic, brain	Group 43	All remaining TRA CDR3s	81	0.035	Out-group
		Homologous TRA CDR3s	8
Combo, blood	Group 61	All remaining TRA CDR3s	474	0.011	Out-group
		Homologous TRA CDR3s	41
Combo, brain	Group 33	All remaining TRA CDR3s	83	0.024	Out-group
		Homologous TRA CDR3s	6

Table 6

TRA CDR3s in the indicated homology in-groups from Table 5

Blood Group 61 CDR3s, representing the combo grouping parameter	Brain Group 33 CDR3s, representing the combo grouping parameter
CAETTLDFNKFYF	CAFMTQWGGSQGNLIF
CALSEETGANSKLTF	CCGGGGGGSQGNLIF
CSEWWAAGNKLTF	CAGLNYGGSQGNLIF
CALSDIGGATNKLIF	CAASAFSSGSARQLTF
CAVSDGGGSNYKLTF	CAVNPHGSAGNMLTF
CATDASAAGNKLTF	CAVGNYGGSQGNLIF
CAEIPINDYKLSF
CGADGTSGSARQLTF
CALSEAGGATNKLIF
CATDATGGGNKLTF
CAVSEGSGGYQKVTF
CLVGESSSGTYKYIF
CGEPLSGYNKLIF
CGTEGITSGSRLTF
CAEAGGGYQKVTF
CAVEGTGGYQKVTF
CAVSDHSSGSARQLTF
CGEQGTGGFKTIF
CALGGENAGNNRKLIW
CAVSEIYNQGGKLIF
CAFADSTSGTYKYIF
CAYEAGGANSKLTF
CATDPSNQGGKLIF
CAVSENSGGYQKVTF
CALIEFSGGYNKLIF
CLVGDLNTGFQKLVF
CATDLIQGAQKLVF
CAMSEYGGATNKLIF
CAESHSGSARQLTF
CGTELTSGTYKYIF
CLVGDGSGNTGKLIF
CATDYGGATNKLIF
CVVSDGSGGYNKLIF
CAELLSGTYKYIF
CAGQDSGGSNYKLTF
CAVSDVTGGGNKLTF
CAVEGGTSYGKLTF
CSDALQGAQKLVF
CAEGSSNTGKLIF
CCAPDGSSNTGKLIF
CALSDGTGGGNKLTF

Chemical complementarity scores for TRA CDR3s with tau, as a whole protein, and for seven segments of the tau protein

We next applied a recently developed, CDR3-candidate antigen, chemical complementarity scoring algorithm [8], using the blood- and brain-derived TRA CDR3s and the entire tau protein sequence. We identified a correlation between the Combo complementarity score and the higher Braak stages using brain-derived TRA CDR3s (Methods, Table 7). Next, the tau protein sequence was divided into seven segments (Supplementary Table 8) to assess potentially more specific TRA CDR3-tau chemical complementarity. Results indicated that the upper 50th percentile, Combo complementarity scores for brain, TRA CDR3s and tau protein segment 7 was associated with the higher Braak stages, while the lower 50th percentile, Combo complementarity scores for Segment 7 was associated with the lower Braak stages (p = 0.044) (Table 8). For the blood-derived, TRA CDR3s, the Hydro complementarity score calculation (Methods) indicated an association of the TRA CDR3-tau segment, chemical complementarity with the Braak stage. Thus, for tau Segments 2 through 6, there was an association of a higher Hydro complementarity score with the higher Braak stages (Table 9).

Table 7

Higher Combo complementarity scores (CSs) for AD+ brain TRA CDR3s and the tau protein (MAPT) were associated with higher Braak stages (Supplementary Table 8).

AD brain Combo CS 50th percentile groups	Number of cases used for correlation	MW p	Associated with the higher (worse) Braak stage group
Lower 50th percentile, brain TRA CDR3-full length tau CSs	44	0.037	Increased CS
Higher 50th percentile, brain TRA CDR3-full length tau CSs	45

Table 8

Higher Combo complementarity scores (CSs) for AD+ brain TRA CDR3s and the tau protein (MAPT) Segment 7 were associated with higher Braak stages

AD brain Combo CS 50th percentile groups	Number of cases used for correlation	MW p	Associated with the higher (worse) Braak stage group
Lower 50th percentile, brain TRA CDR3-tau Segment 7 Combo CSs	44	0.044	Increased CS
Higher 50th percentile, brain TRA CDR3-tau Segment 7 Combo CSs	45

Table 9

Higher Hydro complementarity scores (CS) for AD+ blood TRA CDR3s and the tau Segments 2 through 6 were associated with higher Braak stages

Tau protein segment	AD blood Hydro CS 50th percentile groups	Number of cases	MW p	Associated with the higher (worse) Braak stage group
2	Lower 50th percentile blood TRA CDR3-Segment 2	257	0.026	Increased CS
	Upper 50th percentile blood TRA CDR3-Segment 2	258
3	Lower 50th percentile blood TRA CDR3-Segment 3	257	0.028	Increased CS
	Upper 50th percentile blood TRA CDR3-Segment 3	258
4	Lower 50th percentile blood TRA CDR3-Segment 4	257	0.032	Increased CS
	Upper 50th percentile blood TRA CDR3-Segment 4	258
5	Lower 50th percentile blood TRA CDR3-Segment 5	257	0.017	Increased CS
	Upper 50th percentile blood TRA CDR3-Segment 5	258
6	Lower 50th percentile blood TRA CDR3-Segment 6	257	0.020	Increased CS
	Upper 50th percentile blood TRA CDR3-Segment 6	258

Table 10

Physicochemical parameters of AD+ bloodborne TRG and IGH CDR3s were associated with Braak stage distinctions (Supplementary Table 5).)

Receptor and or physicochemical parameter used	n (number of cases) Braak group 0–II	n Braak group III, IV	n Braak group V	n Braak group VI	KW p	Associated with a higher (worse) Braak stage group	Braak groups compared in Dunn post-hoc test	Post-hoc p
Cases with TRG recovery Isoelectric point	18	68	132	209	0.021^*	Increased property	0–II –III, IV	1
							0–II –V	0.11
							0–II –VI	0.10
							III, IV –V	0.27
							III, IV –VI	0.35
							V –VI	1
Cases with IGH recovery Molecular weight	61	64	88	74	0.018^*	Decreased property	0–II –III, IV	1
							0-II –V	1
							0-II –VI	1
							III, IV –V	0.47
							III, IV –VI	0.034^*
							V –VI	0.037^*

Analysis of single value physicochemical parameters for the blood, AD+ TRG CDR3s

As noted in Introduction, previous studies have noted a potential role of γδ T cells in neuroinflammation and noted a particular TRG clonotype repertoire in AD patients. To further characterize AD+ TRG CDR3 AAs from this project, the single value physicochemical parameters of the TRG CDR3 AA sequences from both blood and brain samples were analyzed. Results indicated that the higher isoelectric point values for the TRG CDR3 AA sequences recovered from the blood WXS files were associated with the higher Braak stages (Table 10). Moreover, a lower molecular weight for bloodborne IGH CDR3s was associated with the higher Braak stages (Table 10).

DISCUSSION

This is apparently the first report of IR recombination reads obtained via mining AD+ blood or brain genomics files, a technical approach that has led to many advances in the cancer setting. Taken together, the above results indicate a potential future opportunity to establish immunogenomics parameters for risk stratification in AD. Specifically, the most important part of TRA for antigen binding, the CDR3, varies with regard to the chemical features of its AAs, and subcategories of this variation can be linked to different Braak stage categories. Furthermore, a recently established CDR3-candidate antigen, chemical complementarity scoring algorithm [8 , 19–22], that has been used to identify differences in immune marker gene expression and survival probabilities in the cancer setting, and that is specifically applicable in a big data setting [8], indicated that TRA CDR3-tau chemical complementarity was associated with a higher Braak stage and thereby possibly a worse clinical condition. No such chemical complementarity associations were detected for the amyloid-β peptide.

Of course, the above results represent many limitations with regard to drawing conclusions. Most importantly, this is a purely correlative study with no opportunity to draw cause and effect conclusions. In addition, the identification of recombination reads in genomics files, although well benchmarked in many settings [23], would have to be complemented if not superseded by analogous studies using a PCR-based, immune repertoire approach to the identification of blood and solid tissue resident immune receptor recombination reads. However, it is important to note that PCR-based immune repertoire approaches remain without a “gold standard” and have inherent biases in the amplification of the IR recombination reads, depending on the methodological details of the amplification process [24]. Thus, with regard to limitations, in general, the recovery of IR recombination reads from genomics files could justify a more expensive investment in a PCR-based immune repertoire approach, but such genomics file, IR recombination read recoveries may still independently address some knowledge gaps.

The recovery of adaptive IR reads from the brain was quantitatively modest, in comparison to certain cancer samples, such as melanoma [8 , 25–28], and it remains to be seen how such IR read recoveries from brain genomics files track lymphocyte counts in the brain, an assessment that has been specifically addressed in the cancer setting [29]. However, more robust IR recoveries for blood samples do indicate the potential application of immunogenomics parameters using blood derived IR reads, e.g., the association of high TRA CDR3-tau chemical complementarity and higher Braak stages, when using TRA recombination reads obtained from blood (Table 9). Again, this is consistent with a number of other studies indicating blood-based immune parameters associated with AD features or occurrence [30, 31]. And most importantly, the linkage of blood-based immune parameters with cognitive decline is consistent with immunological dynamics representing distant organs and systemic features, e.g., microbiome immunology features [32, 33], being associated with cognitive decline.

Over the last several years, the potential role of inflammation, including numerous systemic immune markers [34, 35], has been indicated for the occurrence and progression of AD, although a complete mechanism of effect is lacking. Results above raise the question of whether the binding of tau by adaptive IRs could be either a cause or effect in such a mechanism? With regard to cause, can IR-tau binding lead to the killing of tau expressing cells by cytotoxic T-cells? And a follow up question could be, does IR-tau binding result from the exposure of persons to a pathogen that effectively leaves behind autoreactive (anti-tau) T-cells [36]?

While the data reported here identify a novel source and collection of IR recombinations in the AD patient tissues, it does not allow for hypotheses to directly test cause and effect, as noted above. However, further validation for these correlative results could be attempted with other measures of AD staging that were not available in this ADSP study. Such AD parameters could include other neuropathologic variables, such as the CERAD criteria based on amyloid plaques [37], or cognitive measurements. While Braak staging has been extensively described and relies on specific pathologic findings, pathologists’ determinations could potentially lead to differences from pathologist to pathologist.

Despite current limitations, the data reported here do indicate an opportunity for biomarker development regarding disease severity, using a precise, quantitative representation of chemical interaction potential. A biomarker such as the TRA CDR3-tau chemical complementarity score based on blood TRA recombinations would be particularly useful, given the noninvasive access to a blood sample. Furthermore, with such a noninvasive measurement, there is the potential to identify biomarkers in a relatively large cohort, i.e., a cohort large enough to assess persons at risk for AD. These biomarkers could additionally determine whether a high TRA CDR3-tau complementarity score can predict transition to AD, as opposed to establishing disease severity as implied by the above-described associations of the blood-based, high TRA CDR3-tau complementarity scores (Table 9).

Furthermore, while this study has identified characteristics regarding the TCR repertoire for neuroinflammation in AD, other neurodegenerative diseases including Parkinson’s disease have yet to be fully explored. In Parkinson’s disease, previous studies have shown increased numbers of γδ T-cells [38] and have demonstrated the presence of α-synuclein-specific T cells [39]. These results, along with the results presented in this report, thus raise the question, are there TCR V- or J-gene segment usage, HLA allele combinations that could be at issue in the development of neurodegenerative diseases? In many cases where T-cell responses are at issue, this has indeed been the case, including infectious diseases [40], autoimmunity [41], and cancer [29 , 42–46].

Footnotes

ACKNOWLEDGMENTS

Authors gratefully acknowledge the contributions of USF research computing and the taxpayers of the State of Florida; Ms. Corinne Walters for extensive administrative support related to the NIH dbGaP approvals; and Morsani College of Medicine RISE research fellowships (TIH, BIC). This project was supported in part by a proposal enhancement grant from USF. For Mom and Dad.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Onyango

, Jauregui

, Carna

, Bennett

Jr , Stokin

(2021) Neuroinflammation in Alzheimer’s disease. Biomedicines 9, 524.

Aliseychik

, Patrikeev

, Gusev

, Grigorenko

, Andreeva

, Biragyn

, Rogaev

(2020) Dissection of the human T-cell receptor gamma gene repertoire in the brain and peripheral blood identifies age- and Alzheimer’s disease-associated clonotype profiles. Front Immunol 11, 12.

Gate

, Saligrama

, Leventhal

, Yang

, Unger

, Middeldorp

, Chen

, Lehallier

, Channappa

, De Los Santos

, McBride

, Pluvinage

, Elahi

, Tam

, Kim

, Greicius

, Wagner

, Aigner

, Galasko

, Davis

, Wyss-Coray

(2020) Clonally expanded CD8 T cells patrol the cerebrospinal fluid in Alzheimer’s disease. Nature 577, 399–404.

Liu

, Wang

, Li

, Fowler

, Zeng

, Deng

, Xu

, Zhou

, Doecke

, Villemagne

, Lim

, Masters

, Wang

(2021) Association of naturally occurring antibodies to beta-amyloid with cognitive decline and cerebral amyloidosis in Alzheimer’s disease. Sci Adv 7, eabb0457.

Soskic

, Cano-Gamez

, Smyth

, Rowan

, Nakic

, Esparza-Gordillo

, Bossini-Castillo

, Tough

, Larminie

CGC

, Bronson

, Wille

, Trynka

(2019) Chromatin activity at GWAS loci identifies T cell states driving complex immune diseases. Nat Genet 51, 1486–1493.

Gallardo

, Holtzman

(2017) Antibody therapeutics targeting Abeta and tau. Cold Spring Harb Perspect Med 7, a024331.

Mukhopadhyay

, Banerjee

(2021) A primer on the evolution of aducanumab: The first antibody approved for treatment of Alzheimer’s disease. J Alzheimers Dis 83, 1537–1552.

Chobrutskiy

, Chobrutskiy

, Zaman

, Yeagley

, Huda

, Blanck

(2021) High-throughput, sliding-window algorithm for assessing chemical complementarity between immune receptor CDR3 domains and cancer mutant peptides: TRG-PIK3CA interactions and breast cancer. Mol Immunol 135, 247–253.

Beecham

, Bis

, Martin

, Choi

, DeStefano

, van Duijn

, Fornage

, Gabriel

, Koboldt

, Larson

, Naj

, Psaty

, Salerno

, Bush

, Foroud

, Wijsman

, Farrer

, Goate

, Haines

, Pericak-Vance

, Boerwinkle

, Mayeux

, Seshadri

, Schellenberg

(2017) The Alzheimer’s Disease Sequencing Project: Study design and sample selection. Neurol Genet 3, e194.

10.

Braak

, Braak

(1991) Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 82, 239–259.

11.

Braak

, Braak

, Bohl

(1993) Staging of Alzheimer-related cortical destruction. Eur Neurol 33, 403–408.

12.

Mufson

, Malek-Ahmadi

, Snyder

, Ausdemore

, Chen

, Perez

(2016) Braak stage and trajectory of cognitive decline in noncognitively impaired elders. Neurobiol Aging 43, 101–110.

13.

Biel

, Brendel

, Rubinski

, Buerger

, Janowitz

, Dichgans

, Franzmeier

, Alzheimer’s

Disease Neuroimaging Initiative

(2021) Tau-PET and in vivo Braak-staging as prognostic markers of future cognitive decline in cognitively normal to demented individuals. Alzheimers Res Ther 13, 137.

14.

Ehrenberg

, Suemoto

, Franca Resende

, Petersen

, Leite

REP

, Rodriguez

, Ferretti-Rebustini

REL

, You

, Oh

, Nitrini

, Pasqualucci

, Jacob-Filho

, Kramer

, Gatchel

, Grinberg

(2018) Neuropathologic correlates of psychiatric symptoms in Alzheimer’s disease. J Alzheimers Dis 66, 115–126.

15.

Gill

, Samy

, Butler

, Mauro

, Sexton

, Blanck

(2016) Detection of productively rearranged TcR-alpha V-J sequences in TCGA exome files: Implications for tumor immunoscoring and recovery of antitumor T-cells. Cancer Inform 15, 23–28.

16.

Chobrutskiy

, Zaman

, Tong

, Diviney

, Blanck

(2018) Recovery of T-cell receptor V(D)J recombination reads from lower grade glioma exome files correlates with reduced survival and advanced cancer grade. J Neurooncol 140, 697–704.

17.

Tong

, Tu

, Samy

, Sexton

, Blanck

(2017) Identification of immunoglobulin V(D)J recombinations in solid tumor specimen exome files: Evidence for high level B-cell infiltrates in breast cancer. Hum Vaccin Immunother 13, 501–506.

18.

Holehouse

, Das

, Ahad

, Richardson

, Pappu

(2017) CIDER: Resources to analyze sequence-ensemble relationships of intrinsically disordered proteins. Biophys J 112, 16–21.

19.

Chobrutskiy

, Yeagley

, Tipping

, Zaman

, Diviney

, Patel

, Falasiri

, Uversky

, Blanck

(2020) Chemical complementarity between immune receptor CDR3s and IDH1 mutants correlates with increased survival for lower grade glioma. Oncogene 39, 1773–1783.

20.

Yeagley

, Chobrutskiy

, Gozlan

, Medikonda

, Patel

, Falasiri

, Callahan

, Huda

, Blanck

(2021) Electrostatic complementarity of T-cell receptor-alpha CDR3 domains and mutant amino acids is associated with better survival rates for smas. Pediatr Hematol Oncol 38, 251–264 arco.

21.

Arturo

, Chobrutskiy

, Yeagley

, Patel

, Falasiri

, Patel

, Blanck

(2020) Electrostatic complementarity of B-cell receptor CDR3s and TP53-mutant amino acids in breast cancer is associated with increased disease-free survival rates. Cell Mol Immunol 17, 776–778.

22.

Chobrutskiy

, Yeagley

, Diviney

, Zaman

, Gozlan

, Tipping

, Koohestani

, Roca

, Blanck

(2020) A scoring system for the electrostatic complementarities of T-cell receptors and cancer-mutant amino acids: Multi-cancer analyses of associated survival rates. Immunology 159, 373–383.

23.

Patel

, Yeagley

, Arturo

, Falasiri

, Chobrutskiy

, Gozlan

, Blanck

(2021) A comparison of immune receptor recombination databases sourced from tumour exome or RNAseq files: Verifications of immunological distinctions between primary and metastatic melanoma. Int J Immunogenet 48, 409–418.

24.

Rosati

, Dowds

, Liaskou

, Henriksen

EKK

, Karlsen

, Franke

(2017) Overview of methodologies for T-cell receptor repertoire analysis. BMC Biotechnol 17, 61.

25.

Zaman

, Chobrutskiy

, Patel

, Callahan

, Mihyu

, Diviney

, Blanck

(2019) Potential MMP2-mediated availability of HLA binding, mutant ECM peptides reflects better melanoma survival rates and greater T-cell infiltrates. Lab Invest 99, 1287–1295.

26.

Callahan

, Patel

, Fawcett

, Blanck

(2018) Cytoskeleton and ECM tumor mutant peptides: Increased protease sensitivities and potential consequences for the HLA class I mutant epitope reservoir. Int J Cancer 142, 988–998.

27.

, Tong

, Samy

, Yavorski

, Kim

, Blanck

(2017) Assessing microenvironment immunogenicity using tumor specimen exomes: Co-detection of TcR-alpha/beta V(D)J recombinations correlates with PD-1 expression. Int J Cancer 140, 2568–2576.

28.

Thorsson

, Gibbs

, Brown

, Wolf

, Bortone

, Ou Yang

, Porta-Pardo

, Gao

, Plaisier

, Eddy

, Ziv

, Culhane

, Paull

, Sivakumar

IKA

, Gentles

, Malhotra

, Farshidfar

, Colaprico

, Parker

, Mose

, Vo

, Liu

, Rader

, Dhankani

, Reynolds

, Bowlby

, Califano

, Cherniack

, Anastassiou

, Bedognetti

, Mokrab

, Newman

, Rao

, Chen

, Krasnitz

, Hu

, Malta

, Noushmehr

, Pedamallu

, Bullman

, Ojesina

, Lamb

, Zhou

, Shen

, Choueiri

, Weinstein

, Guinney

, Saltz

, Holt

, Rabkin

, Cancer Genome Atlas Research Network Lazar

, Serody

, Demicco

, Disis

, Vincent

, Shmulevich

(2018) The immune landscape of cancer. Immunity 48, 812–830e814.

29.

Clark

, Tong

, Callahan

, Yavorski

, Tu

, Blanck

(2019) TRB-J1 usage, in combination with the HLA-A^*01:01 allele, represents an apparent survival advantage for uterine corpus endometrial carcinoma: Comparisons with microscopic assessments of lymphocyte infiltrates. Int J Immunogenet 46, 31–37.

30.

, Jia

(2021) Single-Cell RNA sequencing of peripheral blood reveals immune cell signatures in Alzheimer’s disease. Front Immunol 12, 645666.

31.

Lutshumba

, Nikolajczyk

, Bachstetter

(2021) Dysregulation of systemic immunity in aging and dementia. Front Cell Neurosci 15, 652111.

32.

, Li

, Xian

, Li

, Zhang

, Xu

, Pei

, Zheng

, Hu

(2021) Chronic colitis exacerbates NLRP3-dependent neuroinflammation and cognitive impairment in middle-aged brain. J Neuroinflammation 18, 153.

33.

Zou

, Zhou

, Huang

, Luo

, Sun

(2022) Phenotypes, roles, and modulation of regulatory lymphocytes in periodontitis and its associated systemic diseases. J Leukoc Biol 111, 451–467.

34.

Fani

, Georgakis

, Ikram

, Malik

, Dichgans

(2021) Circulating biomarkers of immunity and inflammation, risk of Alzheimer’s disease, and hippocampal volume: A Mendelian randomization study. Transl Psychiatry 11, 291.

35.

Dressman

, Elyaman

(2021) T cells: A growing universe of roles in neurodegenerative diseases. Neuroscientist. doi: 10.1177/10738584211024907.

36.

Yin

, Ibrahim

, Petersen

, Yu

(2021) Autoimmunomic signatures of aging and age-related neurodegenerative diseases are associated with brain function and ribosomal proteins. Front Aging Neurosci 13, 679688.

37.

Boluda

, Toledo

, Irwin

, Raible

, Byrne

, Lee

, Trojanowski

(2014) A comparison of Abeta amyloid pathology staging systems and correlation with clinical diagnosis. Acta Neuropathol 128, 543–550.

38.

Fiszer

, Mix

, Fredrikson

, Kostulas

, Olsson

, Link

(1994) gamma delta+ T cells are increased in patients with Parkinson’s disease. J Neurol Sci 121, 39–45.

39.

Singhania

, Pham

, Dhanwani

, Frazier

, Rezende Dutra

, Marder

, Phillips

, Mallal

, Amara

, Standaert

, Sulzer

, Peters

, Sette

, Lindestam Arlehamn

(2021) The TCR repertoire of alpha-synuclein-specific T cells in Parkinson’s disease is surprisingly diverse. Sci Rep 11, 302.

40.

Attaf

, Malik

, Severinsen

, Roider

, Ogongo

, Buus

, Ndung’u

, Leslie

, Kloverpris

, Matthews

, Sewell

, Goulder

(2018) Major TCR repertoire perturbation by immunodominant HLA-B(^*)44:03-restricted CMV-specific T cells. Front Immunol 9, 2539.

41.

Faham

, Carlton

, Moorhead

, Zheng

, Klinger

, Pepin

, Asbury

, Vignali

, Emerson

, Robins

, Ireland

, Baechler-Gillespie

, Inman

(2017) Discovery of T cell receptor beta motifs specific to HLA-B27-positive ankylosing spondylitis by deep repertoire sequence analysis. Arthritis Rheumatol 69, 774–784.

42.

Huda

, Mihyu

, Gozlan

, Arndt

, Diaz

, Zaman

, Chobrutskiy

, Blanck

(2021) Specific HLA alleles, paired with TCR V- and J-gene segment usage, link to distinct multiple myeloma survival rates. Leuk Lymphoma 62, 1711–1720.

43.

Roca

, Chobrutskiy

, Callahan

, Blanck

(2019) T-cell receptor V and J usage paired with specific HLA alleles associates with distinct cervical cancer survival rates. Hum Immunol 80, 237–242.

44.

Tong

, Callahan

, Tu

, Zaman

, Chobrutskiy

, Blanck

(2019) Immune receptor recombinations from breast cancer exome files, independently and in combination with specific HLA alleles, correlate with better survival rates. Breast Cancer Res Treat 173, 167–177.

45.

Callahan

, Tong

, Blanck

(2018) T cell receptor-beta J usage, in combination with particular HLA class II alleles, correlates with better cancer survival rates. Immunol Res 66, 219–223.

46.

Callahan

, Yavorski

, Tu

, Tong

, Kinskey

, Clark

, Fawcett

, Blanck

(2018) T-cell receptor-beta V and J usage, in combination with particular HLA class I and class II alleles, correlates with cancer survival patterns. Cancer Immunol Immunother 67, 885–892.