Association of KIBRA rs17070145 polymorphism and episodic memory in individuals with severe TBI

Abstract

Background: Studies implicate single nucleotide polymorphism (SNP) rs17070145, a common T → C polymorphism on the KIBRA gene, in mediating differences in episodic memory. In healthy adults, T-allele carriers perform better than non-carriers on episodic memory measures. However, this association is reversed in adults with subjective memory complaints and populations vulnerable to memory deficits, a problem common in traumatic brain injury (TBI).

Methods: This study assessed associations between variation in the KIBRA gene and cognitive function in 129 adults with severe TBI. In addition to other executive functioning and functional/global outcomes, the Buschke Selective Reminding Test (SRT), Rey-Osterrieth Complex Figure Test and California Verbal Learning Test-II (CVLT-II) were administered 6 and 12 months post-injury.

Results: T-allele non-carriers performed better than carriers on multiple episodic memory measures. At 6 months, T-allele non-carriers performed better for delayed recall measures on the SRT. At 12 months, T-allele non-carriers performed better on multiple SRT measures and on List-B learning with CVLT-II. No associations occurred with executive function or global outcome measures.

Conclusion: These results suggest that rs17070145 T-allele effects are specific to episodic memory and support the hypothesis that associations between rs17070145 variation and memory are disparate between healthy and impaired populations.

Introduction

Difficulty with memory is a common cognitive complaint for those with traumatic brain injury (TBI). Human memory is a complex polygenic trait, with heritability estimates of ∼50% [1], suggesting that both environmental and genetic influences can have an effect on individual differences in memory capabilities, including memory function after a significant event like TBI. A number of genes have been implicated in human memory, including the recently identified KIBRA gene. KIBRA is highly expressed in the human kidney and brain, with particularly high levels of expression in the medial temporal lobe and hippocampus [2], [3]. The KIBRA protein acts as a substrate for the atypical protein kinase C zeta (aPKC zeta), which is important for the occurrence of long-term potentiation [4], [5]. As a result, KIBRA may be involved in a number of neurological functions, including synaptic transmission and plasticity, signal transduction and memory formation.

Papassotiropoulos et al. [3] established the first association between KIBRA genetic variability and episodic memory. The results of a genome-wide association study suggested a strong association between episodic memory performance and the single nucleotide polymorphism (SNP) rs17070145, a common T → C exchange located on the ninth intron of the KIBRA gene. Individuals carrying the rs17070145 T allele (CT/TT genotypes) significantly outperformed non-carriers of the allele (CC genotype) on measures of delayed episodic recall in three independent populations including young and older adults. Functional magnetic resonance imaging (fMRI) studies conducted by Papassotiropoulos et al. [3] further support the relationship between the T-allele and better memory performance in healthy young adults. Some studies have not replicated these findings [6], [7] and the associations with KIBRA genotype have been mixed for populations like those with Alzheimer's type dementia [8–10]. Interestingly, a study involving elderly adults with subjective memory complaints revealed that carriers of the T-allele performed worse on measures of episodic memory than non-carriers, an effect opposite of that typically found in healthy populations [11].

The discovery of the association between KIBRA rs17070145 genetic variation and memory is relatively recent and, to date, only a limited number and types of populations have been studied. The present study investigated the association between KIBRA rs17070145 genetic variation and episodic memory recall in a cohort of adult patients with severe TBI. Given the inverse association between the T-allele in populations who subjectively report memory deficits [11], it was hypothesized that non-carriers of the T-allele (CC genotype) would perform better than carriers (CT and TT genotypes) on measures of episodic memory after TBI and this association would be specific to episodic memory.

Methods

Design and participants

Participants included patients aged 16–75 who sustained a severe (Glasgow Coma Scale (GCS) score ≤8) non-penetrating head injury with positive findings on a head CT and survived their injury. Participants were enrolled and consented as part of a larger study examining relationships between genetics and outcomes for TBI patients. This study was approved by the Institutional Review Board at the University of Pittsburgh. The population used for analysis consisted of 129 participants successfully genotyped for KIBRA SNP rs17070145 and with valid scores for the Buschke Selective Reminding Test (SRT) or the California Verbal Learning Test-II (CVLT-II) at 6 months post-injury. A sub-set of 121 participants from this population also completed the SRT or CVLT-II at 12 months post-injury. In addition, the Rey-Osterrieth Complex Figure Test of visuo-spatial episodic memory and other measures of functional cognition and global outcome were assessed in these subjects at both 6 and 12 months post-injury.

Although Yasuda et al. [12] demonstrated similar associations with better verbal memory performance among carriers of the T-allele in a healthy Asian population, due to population variations in ancestry-based allelic frequency [13], non-Caucasian participants with valid SRT or CVLT-II scores were excluded from analysis. Consistent with previous studies, and because the effects of KIBRA rs17070145 on memory are believed to be mainly the result of the C → T polymorphism, participants were grouped by the presence (genotypes CT and TT) or absence (genotype CC) of the KIBRA T allele for analysis. Allelic (C vs T) associations were also assessed in relation to outcome.

Severe TBI status was determined using the Glasgow Coma Scale (GCS), a widely used measure of injury severity in TBI populations which categorizes severity based on level of consciousness and responsiveness and is a sensitive predictor of outcome in heterogeneous injury mechanisms and locations in populations with TBI [14–17]. The GCS was taken within 8 hours of injury to limit the influence of alcohol, sedatives or paralytics, and participant GCS scores ranged from 3–8. GCS was examined as an assessment of global injury. Additional demographic and medical information including age, gender, years of education, mechanism of injury and injury type, based on initial neuroradiology CT reports, was obtained through chart review as well as participant and family interviews after enrolment. Only one individual had a reported injury to the hippocampus, thus limiting potential analyses aimed at studying how injury location with TBI might impact genetic associations with outcome. Injury types derived from CT reports include subarachnoid haemorrhage (SAH), contusion, diffuse axonal injury (DAI), intraventricular haemorrhage (IVH), epidural haematoma (EDH), subdural haematoma (SDH), intracerebral haemorrhage and other miscellaneous injury types. A small set of participants (n = 11) was determined to be too cognitively impaired to complete cognitive testing for the evaluated outcomes. As applicable, these participants were assigned the most impaired possible score and were included in the analyses. Inclusion of cognitively impaired individuals allows generalizing the results to a severe TBI population, as many are too impaired to complete cognitive testing. Also, this approach has been taken in other large scale TBI studies [18], [19].

Assessments

A battery of neuropsychological assessments was administered to patients as a part of the larger study. Rest breaks were provided to subjects as needed in order to minimize the confounding effects of fatigue on testing performance. Tests that assessed verbal episodic memory, visuo-spatial episodic memory, executive function and psychomotor processing speed were examined in the present study. The primary measures utilized for episodic memory performance were the Buschke SRT [20] and CVLT-II [21]. Several studies have confirmed the effectiveness of the SRT as an assessment of episodic memory in people with TBI [22], [23]. No normalized scores are provided for the modified version of the SRT utilized in this study.

In completing the SRT procedure, participants were read a list of 12 unrelated words and asked to repeat as many as they could remember. On each of the five subsequent trials, participants were only read those words that they had omitted in the previous trial and asked again to repeat the 12 word list. This assessment construct allowed for the evaluation of several aspects of recall through a single test. Components captured in the immediate recall portions of the test include (1) Long-Term Storage, which represents the number of words that are recalled on at least one trial without reminding; (2) Consistent Long-Term Retrieval score, which reflects the number of words that were recalled without reminding consistently across trials; (3) Random Long-Term Retrieval, which includes words recalled inconsistently; (4) Long-Term Retrieval, which is comprised of both the consistent long-term retrieval and random long-term retrieval scores; and (5) Short-Term Retrieval, which measures recall of items dependent upon reminding. Delayed free recall were also assessed after a 20-minute break. Possible scores for the immediate recall portions of the test range from 0–72, and scores from the delayed recall portion range from 0–12, where higher scores indicate better performance.

The California Verbal Learning Test-II (CVLT-II [21]) is a brief, easily administered measure of immediate memory span, new learning, susceptibility to interference and memory recognition using word lists of related nouns. A parallel form is available to preclude practice effects from repeated administration and was administered at the 12-month time point. The CVLT paradigm has been studied extensively in TBI [24], [25], including the use of a modified version in an O-15 PET investigation of individuals with TBI [26]. This test involves immediate recall of 16 categorically related words (List A) over five trials (Total Words 1–5, maximum score is 80). After a 16 word interference trial (List B), individuals were asked to freely recall the words from List A (Short Delay Free Recall) for a total score of up to 16. Subjects were then cued to recall the words from List A based on four categories (Short Delay Cued Recall) for a total score of up to 16. The free recall and cued recall of List A was repeated after a 20 minute delay period (Long Delay Free Recall and Long Delay Cued Recall). Individuals were then presented with 48 words and asked if the word belonged to List A, which yields the number of recognition hits (maximum score of 16) and the number of recognition false positives (maximum score of 32).

The Rey-Osterrieth Complex Figure Test was used to assess visuo-spatial episodic memory. In the Rey-Osterrieth Complex Figure Test [27] participants copy an abstract line drawing. The measure also consists of an immediate and a 25 minute delayed recall condition where participants were asked to draw the object from memory as accurately as possible. Scores (out of a total of 36) are based on the number of features of the drawing that are correctly reproduced and assess visuo-spatial construction and memory.

The Trail Making Tests A and B were also administered as a part of the neuropsychological assessment battery and were analysed to rule out the impact of KIBRA rs17070145 variation on cognitive domains other than those involving episodic memory. The Trail Making Tests A and B [28] assess executive function and psychomotor processing speed. In part A, participants were asked to draw lines between 25 consecutive numbers. Part B is similar but requires participants to switch alternatively between letters and numbers, thereby testing cognitive flexibility and task-switching skills. For both tests, scores are based on the time (in seconds) needed to complete the task, with a maximum score of 300, such that lower scores indicate better performance.

Associations with several functional outcome measures were also assessed to explore if KIBRA allelic variation translates to variability in functional outcomes following brain injury. The cognitive sub-scales of the FIM™ rate impairment in the domains of comprehension, expression, social action, problem-solving and memory on an ordinal scale of 1–7, with higher scores indicating lower levels of impairment [29]. Scores for each cognitive sub-scale are routinely combined for a score of 5–35 [30], [31]. The Glasgow Outcome Scale (GOS) is a frequently used measure developed by Jennett and Bond [32] that rates general functional recovery on a scale of 1–5, where a score of 5 indicates a good level of recovery and a score of 1 indicated death. For the purpose of this paper, GOS was dichotomized as 2, 3 vs 4, 5, since the cohort included only TBI survivors. The Disability Rating Scale (DRS) was developed and established by Rappaport [33] as a valid instrument for quantitatively assessing disability of individuals with brain injury over time. The DRS is a 30-point scale that consists of eight items divided into the categories of arousal and awareness, cognitive abilities regarding self-care functions, physical dependence on others and psychosocial adaptability for productive work. Several studies provide strong evidence of the DRS's ability to measure functional change over time in individuals with TBI, and they indicate it is a convenient and versatile tool that can be administered to persons with TBI or through family members [34–36]. The 29-item Neurobehavioural Rating Scale (NRS) provides a broad measurement of the extent of neurobehavioural disturbances after neurological injury. The measure is scaled from 29–116, where higher scores suggest more severe disturbances [37], [38].

Genotyping

For each subject, DNA was extracted from one of two sources, cerebrospinal fluid (CSF) or whole blood. Whole blood was collected in EDTA vacutainer tubes and genomic DNA was isolated by the salting out protocol of Miller et al. [39]. CSF was collected by passive drainage as a standard of care during patients’ intensive care management period, and DNA was extracted from CSF using the QIAamp DNA kit according to the manufacturer's instruction (QIAGEN, Valencia, CA). The KIBRA SNP rs17070145 was amplified using unique sequence primers, NM_015238, and genotyped by fluorescence polarization by the method of Chen et al. [40]. Control samples of known genotype were run in parallel with test samples for quality control.

Statistical analysis

Data analysis was performed with SPSS 17.0 and SAS 9.2. Descriptive analyses, including mean, standard error of the mean (SEM) and median, were calculated for continuous variables including age, GCS and years of education. Frequencies were calculated for categorical variables such as gender, mechanism of injury and injury type. Genetic information was categorized and analysed by variant, specifically the presence or the absence of the T allele (CT/TT vs CC). This grouped genotype analysis, assessing the presence vs absence of the T-allele, was selected based on previous research and because the effects of the KIBRA gene are suspected to be primarily the result of the C → T polymorphism. Allelic analyses for the C- vs the T-allele were also performed by grouping performance on each outcome scale in relation to each of the two alleles for each subject. This approach, while routine for allelic association analyses, results in a doubling of the sample size compared to grouped genotype analysis.

The Mann-Whitney U-test and chi-square analyses were performed for the sample at 6 months and also at 12 months to identify selected demographic and clinical variables having a significant relationship with either genotype or outcome. Variables within the three primary measures of episodic memory tasks were then included in multiple linear regression models if their bivariate relationship with outcome had a p-value less than 0.2. For each outcome measure, variables were then systematically removed from the regression model, starting with the variable with the highest p-value, until all remaining variables had a p-value less than 0.1. The normality assumption was assessed using the Komolgorov-Smirnov test. The literature suggests that stringent adjustments for multiple comparisons (e.g. bonferroni) are not appropriate for neuropsychological tests in which multiple scales and measures are highly inter-correlated [41]. As such, this study controlled for multiple testing by adjusting p-values for the multivariate results using the false discovery rate (FDR) approach [42]. P-values ≤0.05 were considered statistically significant for each analysis.

Results

Description of the population

This study included 129 participants (80% male) with severe TBI (median GCS = 7, mean = 6.26 ± 0.13). Within this cohort, 66 subjects were carriers of the T-allele (genotype CT/TT; 51.2%) and 63 were non-carriers (genotype CC; 48.8%). The minor allele frequency (T-allele frequency) for this population was 31.0% and similar to that reported by the International HapMap Project for a sample of individuals of Northern and Western European ancestry (31.9%) [13]. There were no significant demographic or clinical variable differences between the carriers and non-carriers of the T-allele at either the 6 or 12 month time points (Table I). However, carriers of the T-allele had a higher percentage of subarachnoid haemorrhage than non-carriers which reached significance at the 12-month time point (Table I).

Table I. .

Demographics by presence vs absence of the rs17070145 T-allele

	ALL subjects (n = 129)	6 Mo CC (n = 63)	6 Mo CT/TT (n = 66)	p-value	All subjects (n = 121)	12 Mo CC (n = 62)	12 Mo CT/TT (n = 59)	p-value
Age (years), M ± SEM	32.18 ± 1.18	31.06 ± 1.58	33.24 ± 1.74	0.372	33.13 ± 1.24	32.16 ± 1.63	34.15 ± 1.87	0.455
Education (years), M ± SEM	12.79 ± 0.25	12.61 ± 0.30	12.95 ± 0.40	0.802	12.90 ± 0.31	12.67 ± 0.33	13.23 ± 0.58	0.492
GCS, (mean ± SEM)	6.27 ± 0.13	6.26 ± 0.18	6.29 ± 0.19	0.913	6.22 ± 0.14	6.31 ± 0.19	6.13 ± 0.19	0.353
Male gender, # (%)	103 (79.84)	50 (48.54)	53 (51.46)	0.894	92 (76.03)	48 (52.17)	44 (47.83)	0.832
Motor vehicle accidents, # (%)	73 (56.6%)	35 (55.6%)	38 (57.6%)	0.751	71 (58.7%)	36 (58.1%)	35 (59.3%)	0.829
Hypothermia, # (%)	25 (19.4%)	12 (19.0%)	13 (19.7%)	0.552	26 (21.5%)	14 (53.8%)	12 (46.2%)	0.469
Injury type (CT findings)
Subdural haemorrhage	63.6%	66.7%	60.6%	0.474	63.6%	66.1%	61.0%	0.559
Subarachnoid haemorrhage	62.8%	55.6%	69.7%	0.096	67.8%	59.7%	76.3%	0.050
Diffuse axonal injury	31.8%	30.2%	33.3%	0.699	31.4%	27.4%	35.6%	0.333
Epidural haemorrhage	14.0%	17.5%	10.6%	0.260	17.4%	21.0%	13.6%	0.280
Haemorrhagic contusion	41.9%	46.0%	37.9%	0.348	40.5%	41.9%	39.0%	0.741
Intraventricular haemorrhage	23.3%	17.5%	28.8%	0.126	21.5%	17.7%	25.4%	0.303
Intracerebral haemorrhage	38.8%	36.5%	40.9%	0.608	38.0%	32.3%	44.1%	0.181
Other lesion type	5.4%	7.9%	3.0%	0.266	4.1%	6.5%	1.7%	0.365
Total # of lesion types, M ± SEM	2.81 ± 0.10	2.78 ± 0.14	2.85 ± 0.15	0.835	2.84 ± 0.10	2.73 ± 0.13	2.97 ± 0.16	0.292

rs 17070145 variant and episodic memory

Univariate differences in cognitive performance between the CC and CT/TT groups were assessed at 6 and 12 months and are reported in Table II. At 6 months post-injury, non-carriers performed significantly better than carriers of the T-allele on the measures of delayed recall of the SRT, specifically delayed free recall (p = 0.024) and delayed cued recall (p = 0.002). Non-carriers also demonstrated trends to outperform carriers on immediate recall measures of the SRT such as long-term retrieval, long-term storage and consistent long-term retrieval. Trends were also noted with delayed recall measures of recognition on the SRT and long-term delayed free recall on the CVLT-II. At 12 months post-injury, non-carriers performed significantly better than carriers on SRT measures of long-term retrieval (p = 0.013), long-term storage (p = 0.014) and consistent long-term retrieval (p = 0.023). In addition, non-carriers performed significantly better on immediate, delayed and recognition measures of the CVLT-II (Table II). Group genotype differences in measures of delayed recall on the SRT were no longer significant by 12 months post-injury, although trends were still present (Table II). Allelic analyses (see Table III) supported grouped genotype trends at both 6 and 12 months. At 6 months the C-allele associated with significantly better delayed free recall and delayed cued recall performance on the SRT (p ≤ 0.033 all comparisons), while, at 12 months, the C-allele was associated with SRT long-term retrieval, long-term storage and consistent long-term retrieval performance (p ≤ 0.36 all comparisons). The C-allele was also associated with significantly better performance on all measures of the CVLT-II at the 12-month time point (p ≤ 0.047 all comparisons) with the exception of recognition false positive, which was significant at 6 months (p = 0.048) but not at 12 months (Table III).

Table II. .

Analysis of neuropsychological and functional outcomes by grouped genotype CC vs (CT + TT)

	6 Mo CC (n = 63)	6 Mo CT/TT (n = 66)	p-value	12 Mo CC (n = 62)	12 Mo CT/TT (n = 59)	p-value
Selective Reminding List Learning	(n = 54)	(n = 59)		(n = 49)	(n = 53)
Immediate
Long-Term Retrieval, M ± SEM	25.96 ± 2.51	20.10 ± 2.50	0.073	31.24 ± 2.69	21.58 ± 2.63	0.013
Short-Term Retrieval, M ± SEM	13.26 ± 1.01	13.34 ± 1.28	0.617	11.67 ± 0.96	12.15 ± 1.18	0.846
Long-Term Storage, M ± SEM	27.81 ± 2.67	21.93 ± 2.66	0.089	34.06 ± 2.81	24.06 ± 2.76	0.014
Consistent Long-Term Retrieval, M ± SEM	20.48 ± 2.29	15.24 ± 2.16	0.057	24.00 ± 2.58	16.40 ± 2.39	0.023
Random Long-Term Retrieval, M ± SEM	5.64 ± 0.73	4.83 ± 0.71	0.307	7.24 ± 0.81	5.23 ± 0.68	0.069
Delayed
Free Recall, M ± SEM	5.47 ± 0.54	3.75 ± 0.50	0.024	5.69 ± 0.57	4.21 ± 0.54	0.057
Cued Recall, M ± SEM	6.25 ± 0.49	4.09 ± 0.45	0.002	6.61 ± 0.55	5.15 ± 0.58	0.064
Recognition, M ± SEM	10.21 ± 0.45	8.53 ± 0.61	0.079	10.57 ± 0.35	8.81 ± 0.65	0.140
Rey Osterrieth	(n = 63)	(n = 64)		(n = 62)	(n = 57)
Copy	24.81 ± 1.26	21.73 ± 1.65	0.056	25.77 ± 1.12	22.67 ± 1.65	0.388
Immediate Recall	14.16 ± 1.05	10.96 ± 1.21	0.015	15.36 ± 1.10	10.96 ± 1.13	0.007
Delayed Recall	13.26 ± 1.08	11.17 ± 1.23	0.094	14.65 ± 1.10	11.63 ± 1.15	0.040
California Verbal Learning Test-II	(n = 31)	(n = 35)		(n = 33)	(n = 35)
Total Words Trials 1–5	37.29 ± 2.94	28.80 ± 3.69	0.101	40.55 ± 2.78	29.20 ± 3.45	0.027
List B	4.16 ± 0.31	3.23 ± 0.43	0.144	4.70 ± 0.32	3.11 ± 0.44	0.012
Short Delay Free Recall	6.65 ± 0.78	4.89 ± 0.85	0.102	7.91 ± 0.88	5.43 ± 0.83	0.038
Short Delay Cued Recall	7.42 ± 0.81	5.91 ± 0.93	0.206	9.09 ± 0.81	6.60 ± 0.86	0.036
Long Delay Free Recall	6.87 ± 0.86	4.97 ± 0.85	0.089	7.94 ± 0.90	5.63 ± 0.84	0.058
Long Delay Cued Recall	7.55 ± 0.84	5.71 ± 0.87	0.117	9.12 ± 0.91	6.63 ± 0.88	0.038
Recognition Hits	11.45 ± 0.99	9.69 ± 1.05	0.175	12.97 ± 0.65	10.00 ± 1.06	0.043
Recognition False Positives	2.97 ± 0.89	4.54 ± 0.96	0.144	4.06 ± 0.90	5.19 ± 0.89	0.155
Trail Making Test	(n = 57)	(n = 63)		(n = 48)	(n = 48)
Trail Making Test A	59.56 ± 7.32	93.48 ± 12.61	0.460	65.00 ± 9.43	98.71 ± 15.01	0.273
Trail Making Test B	120.63 ± 10.26	138.00 ± 12.25	0.563	127.87 ± 12.49	143.38 ± 13.82	0.530
Functional/Global Outcomes	(n = 63)	(n = 65)		(n = 62)	(n = 59)
GOS, % better outcome	60.3	53.9	0.480	74.2	59.3	0.122
FIM™	29.26 ± 0.82	27.00 ± 1.12	0.237	30.13 ± 0.68	26.97 ± 1.21	0.104
Disability Rating Scale	3.81 ± 0.10	5.38 ± 0.71	0.159	3.03 ± 0.45	5.24 ± 0.83	0.098
Neurobehavioural Rating Scale	40.69 ± 1.06	40.80 ± 1.21	0.881	42.00 ± 1.84	40.84 ± 1.54	0.972

Table III. .

Analysis of neuropsychological and functional outcomes by presence of C- vs T-allele

	6 Mo C (n = 177)	6 Mo T (n = 79)	p-value	12 Mo C (n = 168)	12 Mo T (n = 74)	p-value
Selective Reminding List Learning	(n = 153)	(n = 73)		(n = 138)	(n = 66)
Immediate
Long-Term Retrieval, M ± SEM	23.99 ± 1.50	20.62 ± 2.29	0.150	28.33 ± 1.64	21.82 ± 2.39	0.026
Short-Term Retrieval, M ± SEM	13.26 ± 0.64	13.38 ± 1.20	0.540	12.02 ± 0.63	11.71 ± 1.02	0.768
Long-Term Storage, M ± SEM	25.88 ± 1.60	22.36 ± 2.42	0.152	31.04 ± 1.71	24.30 ± 2.51	0.026
Consistent Long-Term Retrieval, M ± SEM	18.58 ± 1.34	15.99 ± 2.04	0.135	21.72 ± 1.54	16.56 ± 2.17	0.036
Random Long-Term Retrieval, M ± SEM	5.51 ± 0.44	4.60 ± 0.62	0.171	6.63 ± 0.47	5.29 ± 0.60	0.131
Delayed
Free Recall, mean ± SEM	4.97 ± 0.32	3.76 ± 0.45	0.033	5.23 ± 0.34	4.29 ± 0.48	0.126
Cued Recall, mean ± SEM	5.61 ± 0.30	4.10 ± 0.40	0.003	6.18 ± 0.34	5.18 ± 0.51	0.099
Recognition, mean ± SEM	9.71 ± 0.31	8.55 ± 0.54	0.099	10.07 ± 0.28	8.80 ± 0.58	0.155
Rey Osterrieth	(n = 175)	(n = 77)		(n = 167)	(n = 71)
Copy	23.74 ± 0.81	20.73 ± 1.37	0.056	25.25 ± 0.76	22.01 ± 1.47	0.115
Immediate Recall	13.20 ± 0.64	10.51 ± 0.98	0.017	14.36 ± 0.67	10.64 ± 1.01	0.003
Delayed Recall	12.59 ± 0.64	10.68 ± 1.02	0.097	14.04 ± 0.67	11.22 ± 1.03	0.013
California Verbal Learning Test-II	(n = 90)	(n = 42)		(n = 92)	(n = 44)
Total Words Trials 1-5	34.61 ± 1.90	28.88 ± 3.49	0.134	37.41 ± 1.87	29.05 ± 3.04	0.025
List B	3.92 ± 0.22	3.12 ± 0.39	0.101	4.26 ± 0.23	3.09 ± 0.39	0.012
Short-Delay Free Recall	6.03 ± 0.47	5.02 ± 0.81	0.167	7.28 ± 0.52	5.27 ± 0.75	0.027
Short-Delay Cued Recall	6.99 ± 0.50	5.83 ± 0.86	0.198	8.42 ± 0.51	6.52 ± 0.77	0.033
Long-Delay Free Recall	6.23 ± 0.51	5.07 ± 0.80	0.141	7.34 ± 0.54	5.52 ± 0.76	0.047
Long-Delay Cued Recall	6.99 ± 0.50	5.69 ± 0.82	0.132	8.48 ± 0.55	6.50 ± 0.79	0.033
Recognition Hits	11.00 ± 0.60	9.48 ± 0.95	0.130	12.16 ± 0.50	9.93 ± 0.93	0.027
Recognition False Positives	3.47 ± 0.53	4.32 ± 0.92	0.048	4.41 ± 0.54	4.97 ± 0.81	0.387
Trail Making Test	(n = 164)	(n = 76)		(n = 133)	(n = 59)
Trail Making Test A	69.82 ± 5.59	93.64 ± 11.76	0.931	74.01 ± 6.86	99.54 ± 13.47	0.285
Trail Making Test B	126.11 ± 6.44	137.07 ± 11.37	0.897	131.17 ± 7.56	146.00 ± 12.89	0.520
Functional/Global Outcomes	(n = 177)	(n = 79)		(n = 168)	(n = 74)
GOS, % better outcome	59.32%	51.90%	0.268	72.02%	55.41%	0.011
FIM™	28.72 ± 0.55	26.74 ± 1.03	0.156	29.45 ± 0.52	26.60 ± 1.08	0.030
Disability Rating Scale	4.13 ± 0.34	5.63 ± 0.65	0.040	3.50 ± 0.36	5.49 ± 0.72	0.012
Neurobehavioural Rating Scale	40.72 ± 0.65	40.80 ± 1.10	0.850	41.58 ± 1.08	41.13 ± 1.33	0.661

On the Rey-Osterrieth Complex Figure Drawing test, grouped genotype analysis showed that non-carriers of the T-allele performed significantly better than carriers in the immediate recall condition at 6 months post-injury (p = 0.015). No significant genetic effects were noted on either the copy or the delayed condition at 6 months, although there were trends in both conditions for better performance among non-carriers. At 12 months post-injury non-carriers of the T-allele performed significantly better than carriers on immediate recall (p = 0.007) and delayed recall (p = 0.040) of the Rey Figure (Table II). Allelic analysis supported grouped genotype results (Table III), where those with the C-allele had significantly better performance on the Rey Figure for immediate recall at 6 months (p = 0.017) and for immediate and delayed recall at 12 months (p ≤ 0.013 both conditions).

Multivariate analyses also indicated that age and GCS have a significant influence on several measures with SRT and Rey performance; however, other demographic variables, including education, mechanisms of injury and injury type were not related to episodic memory scores. Multiple linear regression was performed for each outcome measure to determine the relationships between KIBRA variation and episodic memory performance scale after adjusting for covariates associated with specific sub-scales as appropriate. After adjusting for covariates and multiple testing, non-carriers of the T-allele performed significantly better than carriers on the SRT for measures of delayed cued recall (p = 0.020) and delayed free recall (p = 0.046) at 6-months post-injury. There was a similar trend on the SRT delayed recognition (p = 0.082) and Rey immediate recall (p = 0.082) at 6 months post-injury. At 12 months post-injury, there were differences with the SRT on long-term retrieval (p = 0.046), long-term storage (p = 0.046) and delayed recognition (p = 0.073) scales where the non-carriers of the T-allele performed better than carriers. List B learning of the CVLT-II was also significantly (p = 0.045) influenced by grouped genotype, after controlling for covariates, where the non-carriers of the T-allele performed better than carriers. Age was significantly associated with all episodic memory measures in multivariate analyses listed in Tables IV and V, except List B on the CVLT at 12 months. In contrast, GCS score was significantly associated with two performance measures and no other covariates remained significant in multivariate analysis.

Table IV. .

6M multivariate analysis results

Independent variable	SRT delayed cued recall (n = 109)		SRT delayed recognition (n = 108)		SRT delayed free recall (n = 109)		Rey immediate recall (n = 119)
	Beta	p-value	Beta	p-value	Beta	p-value	Beta	p-value
Intercept	6	<0.001	6.2	<0.001	6.8	<0.001	10.7	0.007
Age	−1.2	<0.001	−1	0.012	−1	<0.001	−1.8	0.007
Genotype	2	0.02	1.3	0.082	1.6	0.046	2.8	0.082
GCS	0.3	0.154	0.9	<0.001			0.9	0.101

Table V. .

12M multivariate analysis results

Independent variable	SRT long-term retrieval (n = 102)		SRT long-term storage (n = 102)		SRT consistent long-term retrieval (n = 102)		SRT random long-term retrieval (n = 102)		SRT delayed recognition (n = 98)		CVLT-II list B (n = 63)
	Beta	p-value	Beta	p-value	Beta	p-value	Beta	p-value	Beta	p-value	Beta	p-value
Intercept	38.4	<0.001	41.6	<0.001	30.5	<0.001	7.9	<0.001	6.7	<0.001	4.2	<0.001
Age	−5.1	<0.001	−5.4	<0.001	−4.3	0.002	−0.8	0.049	−0.6	0.038	−0.3	0.108
Genotype	8.6	0.046	8.8	0.046	6.7	0.073	1.8	0.091	1.5	0.073	1.5	0.045
GCS									0.7	0.012

rs17070145 variant and other neuropsychological and global outcome measures

Univariate associations between KIBRA variation and performance on non-memory-specific neuropsychological measures were also assessed. There were no significant associations between KIBRA and performance on Trail Making Tests A and B based on grouped genotype (Table II) or allelic (Table III) analysis at either 6 or 12 months post-injury. On measures of functional outcome there were no statistically significant differences between carriers and non-carriers of the T-allele for GOS, the cognitive sub-scale of the FIM™, DRS or NRS (Table II). Conversely, allelic analyses (Table III) suggests that the C-allele is associated with better outcome on the DRS at 6 months (p = 0.040), while there were significant differences on the GOS, FIM™ and DRS at 12 months (p ≤ 0.030, all comparisons). Grouped genotype analysis (Table II) and allelic analysis (Table III) did not demonstrate significant differences in performance on the NRS, a measure of global neurobehavioural outcome, in association with KIBRA genetic variants.

Discussion

Memory deficits are a common problem for many patients with TBI that are often associated with functional deficits and impaired quality-of-life [43], [44]. Determining which patients have the highest risk for memory impairment presents a challenge, in part, because the factors that influence individual differences in memory are complex and not completely understood. Further, TBI is a heterogeneous disease, with the location and type of injury (e.g. based on CT findings) being variable and resulting in different patterns of cognitive deficit and associated outcomes. Recent genetic advances, however, have allowed researchers to identify genes that may shape memory capacity. One gene is KIBRA, which contains the SNP rs17070145 and has been implicated as influencing episodic memory. To date, no studies have investigated the relationship between KIBRA variation and episodic memory performance in individuals with TBI, despite the prevalence of the disorder and the high incidence of memory difficulties following head injuries. Based on Nacmias et al. [11], we hypothesized that non-carriers of the T-allele would perform better than carriers on measures of episodic memory.

Consistent with this hypothesis, non-carriers of the T-allele (CC genotype) performed better than carriers (CT/TT genotypes) on episodic memory measures (see Tables II and III). Although there were trends with immediate recall measures at 6 months post-injury for the SRT, differences were more pronounced on delayed recall measures. After adjusting for age, non-carriers of the T-allele still performed better than carriers on these delayed recall measures. At 12 months post-injury, there were still trends with genetic differences in delayed recall measures on the SRT; however, non-carriers also outperformed carriers on several components of immediate episodic recall, with significant associations observed for long-term storage, long-term retrieval and consistent long-term retrieval SRT measures that held up in multivariate analyses. The association of the KIBRA gene with verbal episodic memory performance on the SRT was supported with a second measure of verbal memory, using the CVLT-II, in a sub-set of participants. Although there were no strong associations at 6 months, there were significant genetic differences on both immediate and delayed memory measures at 12 months post-injury, with List B performance holding up in multivariate analysis. Allelic analysis supports these associations by showing the T-allele to be detrimental within the same domains at 6 and 12 months and multivariate analysis further supported these associations at 12 months for these measures by demonstrating independent genetic associations even after consideration and adjustments for a broad range of covariates. Grouped genotype associations were present for immediate recall measure with the Rey Figure Drawing test at 6 months and for the immediate and delayed recall measures at 12 months. Also, allelic analysis supports the association with immediate recall at both 6 and 12 months.

Taken together, these results support the hypothesis that non-carriers of the T-allele demonstrate better performance on episodic memory measures than T-carriers following TBI—a relationship opposite to that previously suggested in the literature for healthy populations, but consistent with reports on clinical populations known to have cognitive performance complaints [11]. Although some studies suggest that delayed recall scales are more affected by genotype, trends vs significant findings regarding performance on specific sub-scales relating to delayed vs immediate recall in this study may be more a function of sample size, scale metrics and level of test difficulty across recovery time than specific genetic-mediated effects on the nuanced differences present with the cognitive elements required for completing each of these related sub-scales. Multivariate analysis of SRT scales capturing delayed recall show the largest genetic associations at 6 months post-TBI. The SRT data for long-term retrieval, long-term storage and consistent long-term retrieval suggest that this KIBRA gene variant tended to be more associated with scores that require organized encoding and retrieval at 12 months after injury. Multivariate analysis demonstrates that only performance on the interference list learning, List B, remained significantly associated with the KIBRA gene at 12 months post-injury. However, the large number of associations in the univariate analysis at 12 months and the relatively low sample suggests that these task components may be significantly influenced by genetic variability within the KIBRA gene in larger samples. It is also possible that the lack of associations observed between the KIBRA gene and performance on the CVLT-II at 6 months post-injury may be influenced by overall difficulty in test performance at this earlier time-point. Non-carriers of the T-allele performed significantly better than carriers in the immediate recall condition of the Rey Figure Drawing Test at 6 and 12 months post-injury, suggesting that KIBRA influences on episodic memory are not exclusive to verbal memory. For this task, it may be the case that the delayed recall conditions at 6 months post-injury were too difficult to discriminate genotype mediated differences in delayed recall abilities for visuo-spatial information in this population. Although not all sub-scales met significance at each time point, the trends and positive findings in this exploratory analysis, spanning several scales of three measures of episodic memory, when taken together, provide overall support for a potentially useful and informative association with memory abilities after TBI.

Previous research has indicated that the effect of KIBRA variation appears to be specific to episodic memory in both healthy and impaired populations. In the present study, the Trail Making Test parts A and B were evaluated to further assess the specificity of KIBRA to episodic memory in individuals with TBI. Consistent with previous research, there were no significant differences between groups for the Trail Making Test, indicating that KIBRA variation does not affect psychomotor processing speed and does not affect executive functioning performance in a task where planning capabilities and general cognitive flexibility are required to complete the task. The lack of grouped genotype effects on a global and functional outcome measures is not surprising since these measures incorporate elements of cognition not related specifically to memory and the relationship between memory ability and behavioural/cognitive performance can be influenced by many factors like the use of compensatory strategies, fatigue and other co-morbid conditions. However, in combination with other informative genetic variants influencing memory and other elements influencing cognition, it may be that the KIBRA genotype significantly impacts these broader measures of outcome and recovery. The KIBRA allelic analyses suggest that the genetic associations with memory may translate to differences in global and functional outcome in larger replication studies adequately powered to more definitively explore these relationships.

Some individuals (n = 11) who had been successfully genotyped for KIBRA did not have scores for neuropsychological tests because they were unable to complete the assessments due to severe cognitive deficits. Similar to other studies in TBI [18], [19], these participants were included in the cohort and assigned the most impaired possible scores. Of these 11 individuals, nine were T-allele carriers and two were non-carriers. The unequal distribution of cognitively unable participants in each genotype group is notable and further supports the hypothesis that non-carriers are less susceptible to cognitive damage. Removal of these cognitively unable individuals, however, did not substantially alter the overall associations between the KIBRA gene and performance on measures of episodic memory (data not shown).

A number of genes play a role in shaping human memory capabilities. For instance, a single nucleotide polymorphism on the CLSTN2 gene may interact with KIBRA to modulate episodic memory. In one study, carriers of the KIBRA T-allele and the CLSTN2 C-allele—both previously associated with increased memory performance in healthy adult populations—performed better on episodic memory tasks than subjects with other allelic combinations [45]. Identification of additional genes that impact memory after TBI and their interaction with KIBRA gene effects may shed further light on how non-T-carriers are relatively protected after TBI and why the genetic association with cognition is different from healthy controls. Although a study by Almeida et al. [46] supported the association between better episodic memory performance in non-carriers of the T-allele in healthy older adults, they did not find an association between phenotype expression in mild cognitive impairment (MCI).

A study by Papassotiropoulos et al. [3] utilized functional magnetic resonance imaging (fMRI) to determine the relationship between KIBRA variation and neuronal activity in areas of the brain involved in memory. They found that non-carriers of the T-allele displayed greater activation in memory-related regions during a memory task than did carriers, even though performance on the task was not significantly different between the two groups. These results indicated that in healthy populations, non-carriers of the T-allele may be less efficient than carriers to reach the same level of memory retrieval. A recent fMRI study by Russell et al. [47] administered measures of episodic memory to individuals with TBI and to matched controls. Interestingly, TBI participants required more activation than controls to achieve the same level of retrieval. Future research on cognition in TBI populations could utilize fMRI to assess how injury status and genotype interplay with tissue activation and recruitment to help elucidate mechanisms by which KIBRA variation impacts episodic memory performance in clinically impaired populations. Given that KIBRA genetic associations are linked to episodic memory, this study specifically showed that radiographic evidence of hippocampal injury was not a significant potential covariate in this analysis. Also, CT-derived evidence of different injury types were not largely revealing as confounders for genotype distribution or as indicators of outcome, suggesting that, while KIBRA associations are ‘memory’ specific, they can impact memory performance across a heterogeneous TBI population. While potentially interesting, this population size was too small to explore specific gene * injury location interactions on outcome.

KIBRA is a large gene, spanning over 189 kb, with 23 exons covering 6720 bp. There is little known about the gene. rs17070145 is located inside a large intron, and the linkage disequilibrium pattern around it does not suggest linkage to any functional regions of the protein (www.hapmap.org). However, there is no data currently on the function of this intron and, one cannot rule out the presence of regulatory regions in the intron that may impact the protein expressed. Despite this lack of specificity regarding the neurobiological impact of this SNP on KIBRA function, some studies suggest that KIBRA gene function is regionally altered in brain regions susceptible to AD neuropathology [9].

The concept that a genetic variant that enhances memory capabilities in healthy populations may simultaneously act as a risk factor for greater impairment under certain circumstances is not unique within the KIBRA literature. In addition to the study on subjects with cognitive complaints by Nacmias et al. [11], this effect has been suggested for KIBRA and Alzheimer's disease risk. Research on healthy populations indicates that carriers of the rs17070145 T-allele demonstrate better episodic memory performance than non-carriers. Rodriguez-Rodriguez et al. [10] genotyped a large population with and without Alzheimer's disease and determined that carriers of the T-allele exhibited a higher risk of very late-onset Alzheimer's disease than did non-carriers. In contrast, Corneveaux et al. [9] utilized brain imaging techniques in addition to genetic association tests to find that non-carriers of the T-allele had an increased risk of late-onset Alzheimer's disease when compared to carriers. This inconsistency further highlights the need for replication studies in Alzheimer's disease as well as additional research on the genetics of memory in various clinical populations such as TBI. Both the TBI and Alzheimer's disease populations have cognitive deficits and share some elements of overlapping disease pathology [48], [49], suggesting that more work is needed with understanding the nature and potential genetic implications of these elements of common pathology. Evidence suggests that TBI can be linked with the later development of neurodegenerative diseases and that genes like APOE may have shared risk variants across both the Alzheimer's and the TBI populations [50], further supporting the notion that additional genetic susceptibility studies to primary and post-traumatic neurodegenerative syndromes are warranted.

Future work needs to be done to verify the significance and direction of this association in other TBI and acquired injury populations. The estimated effect size of this gene variant (based on SRT delayed free recall) in this study population is (0.43) compared to (0.48) in healthy population KIBRA studies [51], suggesting that similarly sized genetic effects are observed in each of these populations, despite the fact that opposite alleles are referenced as the risk allele in each of these studies. Further in vitro work may be helpful to further understand the role of this genetic variant on local hippocampal circuitry and function in order to speculate what the functional implications may be in the population with TBI.

The frequency of the KIBRA alleles can vary in different populations with different ethnicity. However, the population was restricted to Caucasians, limiting the potential confounding effects of this issue on these findings. Age was significantly related to episodic memory measures across nearly all of the multivariate models created. Scores for episodic memory performance were not normalized for age. As such, it is difficult to determine if age effects on outcomes are TBI-specific or generally applicable to age-related differences in memory performance for these scales. However, age does remain a significant predictor of outcome for measures like GOS and DRS in other studies [16], [17]. Depression was not assessed in this population. However, some studies suggest that processing speed decreases associated with depression may impact neuropsychological testing performance [52] and future studies could consider this point given known effects of major depression on cognition as well as post-traumatic depression associations with cognition [43]. Interestingly, injury severity, as measured by GCS scores, was not associated with episodic memory performance, with the exception of two scales. The limited sensitivity of GCS in relation to episodic memory could be multi-factorial; however, one possibility may be that the population was comprised only of subjects with severe TBI, thus precluding use of the full scale metrics associated with the GCS score. When considering future work to generalize findings to other populations, including those with concussion, incorporating other measures of injury severity, like post-traumatic amnesia duration, may help minimize ceiling effects with regard to injury severity associations with neuropsychological performance.

In conclusion, this study demonstrates that genetic variability in the KIBRA gene appears to influence episodic memory performance in a population with severe TBI and the direction of the association is opposite of other studies in healthy populations and some populations with dementia. Further work will be required to corroborate risks for worse outcome for those with the T-allele, to delineate potential mechanisms underlying this association and to determine if/how variability in other genes governing memory may interact with variability within the KIBRA gene to influence cognition and recovery.

Acknowledgements

We would like to thank the University of Pittsburgh Brain Trauma Research Center for their assistance in some elements of data collection. We also thank Sandra Deslouches for her technical assistance with the project.

Declaration of Interest: This work was supported in part by R01 HD048162, DODW81XWH-071-0701, R01NR008424 and P01NS030318.