Apolipoprotein E epsilon 4 (APOE‐ε4) genotype is associated with decreased 6‐month verbal memory performance after mild traumatic brain injury

Abstract Introduction The apolipoprotein E (APOE) ε4 allele associates with memory impairment in neurodegenerative diseases. Its association with memory after mild traumatic brain injury (mTBI) is unclear. Methods mTBI patients (Glasgow Coma Scale score 13–15, no neurosurgical intervention, extracranial Abbreviated Injury Scale score ≤1) aged ≥18 years with APOE genotyping results were extracted from the Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot (TRACK‐TBI Pilot) study. Cohorts determined by APOE‐ε4(+/−) were assessed for associations with 6‐month verbal memory, measured by California Verbal Learning Test, Second Edition (CVLT‐II) subscales: Immediate Recall Trials 1–5 (IRT), Short‐Delay Free Recall (SDFR), Short‐Delay Cued Recall (SDCR), Long‐Delay Free Recall (LDFR), and Long‐Delay Cued Recall (LDCR). Multivariable regression controlled for demographic factors, seizure history, loss of consciousness, posttraumatic amnesia, and acute intracranial pathology on computed tomography (CT). Results In 114 mTBI patients (APOE‐ε4(−)=79; APOE‐ε4(+)=35), ApoE‐ε4(+) was associated with long‐delay verbal memory deficits (LDFR: B = −1.17 points, 95% CI [−2.33, −0.01], p = .049; LDCR: B = −1.58 [−2.63, −0.52], p = .004), and a marginal decrease on SDCR (B = −1.02 [−2.05, 0.00], p = .050). CT pathology was the strongest predictor of decreased verbal memory (IRT: B = −8.49, SDFR: B = −2.50, SDCR: B = −1.85, LDFR: B = −2.61, LDCR: B = −2.60; p < .001). Seizure history was associated with decreased short‐term memory (SDFR: B = −1.32, p = .037; SDCR: B = −1.44, p = .038). Conclusion The APOE‐ε4 allele may confer an increased risk of impairment of 6‐month verbal memory for patients suffering mTBI, with implications for heightened surveillance and targeted therapies. Acute intracranial pathology remains the driver of decreased verbal memory performance at 6 months after mTBI.

were conferred up to 25 additional years of aging after suffering TBI (Teasdale, Murray, & Nicoll, 2005). In moderate and severe TBI patients assessed at least 6 months postinjury, Ariza et al. (2006) found that ε4-carriers performed notably worse on tests of frontal lobe function and long-term memory. In a large Australian cohort stratified by age, the APOE-ε4 allele was associated with poorer baseline episodic memory and slower reaction time in younger adults with a history of childhood TBI (Eramudugolla et al., 2014). Similarly, ε4 carriers may be at more risk for cognitive impairment and chronic traumatic encephalopathy (CTE) (Liberman, Stewart, Wesnes, & Troncoso, 2002;Müller et al., 2009;Lenihan & Jordan, 2015;). Conversely in children, Moran et al. report that the APOE-ε4 allele has little impact on neurocognitive measures following TBI, although ε4-carriers were associated with a more negative early response to injury (Moran et al., 2009), which mirrors the adult literature (Friedman et al., 1999;Jiang et al., 2006). Interestingly, pediatric ε4 carriers sustaining moderate to severe TBI showed poorer adaptive functioning in the context of positive parenting, whereas non-ε4-carriers displayed worse detriments in the context of less optimal parenting (Treble- Barna et al., 2016). These studies further highlight the pleiotropic association between APOE-ε4 and neurobehavioral outcomes after TBI, and underscore the need to elucidate its relationship with outcome subdomains after different types of TBI.
To date, mild-to-moderate TBI studies in the literature have reported mixed associations for verbal memory outcome by APOE-ε4 status (Chamelian, Reis, & Feinstein, 2004;Han et al., 2007;Padgett, Summers, Vickers, McCormack, & Skilbeck, 2016). Interestingly, Han and colleagues reported improved neuropsychological outcome at 1 month was associated with APOE-ε4 in the military population (Han et al., 2007), whereas Crawford et al. (2002) reported memory and verbal impairment in military personnel with APOE-ε4 within 6 months of injury. Using an extensive neuropsychological battery at 3 and 6 weeks postinjury, Liberman et al. (2002) reported in adults with mild and moderate TBI that patients with the APOE-ε4 allele had lower F I G U R E 1 Effects of APOE on Aβ metabolism and postinjury repair. Neuronal injury upregulates astrocyte secretion of ApoE, which clears lipid cell debris and assists in cholesterol delivery for synaptogenesis. APOE is lipidated to form ApoE lipoprotein, and in the extracellular space binds in an isoform-dependent pattern (APOE-ε2, -ε3 > APOE-ε4) to soluble beta-amyloid protein (Aβ), a peptidic neurotoxin. APOE genotype determines the capacity for Aβ clearance and parenchymal amyloid plaque accumulation. Evidence suggests that relative to APOE-ε2 and -ε3, APOE-ε4 is preferentially susceptible to proteolytic degradation, thus reducing the capacity to fulfill postinjury needs for membrane repair and synaptogenesis mean scores for nearly all neuropsychological tests and needed more time to complete tasks, although none of the differences were significant. The authors suggested that APOE genotype may specifically influence the severity of the acute injury along with delaying recovery.
In concussed athletes tested within 6 months of injury, carriers of the ε4 allele demonstrated greater neurocognitive variability than athletes without an ε4 allele (Merritt, Rabinowitz, & Arnett, 2017). However, an Indian cohort of mild-to-moderate TBI patients at 6 months showed no clear differences in neuropsychological test performance by APOE-ε4 status (Pruthi et al., 2010). Reports of association between ε4 carriers and postconcussional symptoms months to years postinjury also exist (Sundström et al., 2007;Merritt & Arnett, 2016;Pan et al., 2016).
Population characteristics and TBI severity may contribute to the discrepancies in the current literature. Hence, the association of APOE-ε4 status and subdomains of verbal memory performance following isolated mTBI warrant further study.
For the current analysis, we utilized data from the prospective multicenter Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot (TRACK-TBI Pilot) study (Yue et al., 2013) to explore associations between APOE and components of 6-month verbal memory postinjury. We demonstrate that the APOE-ε4 carrier status is specifically associated with decreased performance on long-delay recall trials, without effects on immediate recall or short-delay recall trials.
Inclusion criteria for TRACK-TBI Pilot were external force trauma to the head presenting to Level I trauma center, and clinically indicated head CT scan within 24 hr of injury. Exclusion criteria were pregnancy, comorbid life-threatening disease, incarceration, psychiatric hold, and non-English speaking due to limitations in participation with outcome assessments.
As the goal of the current analysis was to investigate the association between the APOE-ε4 allele and 6-month verbal memory outcome after mTBI, uncomplicated by massive intracranial trauma [e.g., CT evidence of mixed density lesions >25 mm, midline shift >5 mm, or evidence of cisternal compression (Marshall et al., 1992)] or polytrauma, patients were included if ≥18 years of age, with GCS score of 13-15 at ED admission per current definition of mTBI (Teasdale & Jennett, 1974;), Marshall CT Score 1-2 (Marshall et al., 1992), no acute neurosurgical intervention, and no extracranial injuries greater than "mild" (Abbreviated Injury Scale [AIS] score >1 in any extracranial body region). To preserve the integrity of cognitive outcomes analysis, patients with a baseline history or diagnosis of central nervous system (CNS) tumor, cerebrovascular accident (CVA), schizophrenia, bipolar disorder, learning disability, and/or developmental delay were excluded from analysis.
Eligible patients were enrolled through convenience sampling at all three sites. Procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Institutional Review Board (IRB) approval was obtained at all participating sites. Informed consent was obtained for all patients prior to enrollment in the study. For patients unable to provide consent due to their injury, consent was obtained from their legally authorized representative (LAR). Patients were then re-consented, if cognitively able, at later inpatient and/or outpatient follow-up assessments for continued participation in the study.

| Biospecimen acquisition and genotyping
Specimen acquisition was performed as previously described (Yue et al., 2013). In brief, blood samples were collected via peripheral venipuncture or existing peripheral venous or arterial indwelling catheters within 24 hr of injury for DNA genotyping. Samples were collected in BD Vacutainer K 2 -EDTA vacutainer tubes, then aliquoted and frozen in cryotubes at −80°C within 1 hr of collection in accordance with

| Outcome measures
The California Verbal Learning Test, Second Edition (CVLT-II) is a verbal learning and memory task with five learning trials, an interference trial, an immediate recall trial performed with and without cues, and a post-20 min recall trial performed with and without cues. The CVLT-II was substituted for the Rey Auditory Verbal Learning Test (RAVLT) listed in the NIH NINDS outcome CDEs due to relevant revisions of the Second Edition and higher consistency on between-norm sets (Stallings, Boake, & Sherer, 1995). The CVLT-II has been used in over 200 published clinical and experimental research studies as a measure of episodic verbal learning and memory (Woods, Delis, Scott, Kramer, & Holdnack, 2006). It has also been found to be one of the most sensitive tests for detecting residual brain damage in patients with brain trauma, able to differentiate between outcomes following mild versus moderate-severe TBI (Davis, 2016;Delis, Kramer, Kaplan, & Ober, 2000). Researchers have found that the CVLT can also accurately predict functional outcomes at the time of testing, including level of current disability, type of current supervision, and return to work status (Hanks, Jackson, & Crisanti, 2016).
The CVLT-II subscales allow for differentiation between encoding, consolidation, and retrieval deficits (Vanderploeg, Crowell, & Curtiss, 2001). Lower scores on the CVLT-II Short-Delay Free Recall (SDFR) indicate retroactive interference, while lower scores on the CVLT-II Long-Delay Free Recall (LDFR) score indicate the occurrence of rapid forgetting. Free recall trials reflect the degree to which words on the first list are recalled without assistance. Cued recall trials assist the examinee in two ways: (1) by informing them of the categorical structure of the list; and (2) requiring them to use semantic clustering in recalling the target words. Patients with mild cognitive impairment (MCI) will often demonstrate aberrant responses on cued recall trials. Patients with more severe deficits will typically respond with deficits on both free and cued recall trials.

| Statistical analysis
Descriptive statistics are presented as means and standard deviations (SD) for continuous variables and as proportions for categorical variables. Group differences in demographics and injury characteristics across APOE-ε4 genotypes were assessed by Pearson's chi-squared test (X 2 ) for categorical variables, and analysis of variance (ANOVA) for continuous variables. Fisher's Exact Test was used to assess for differences in categorical variables with individual cell counts ≤5.
Multivariable linear regression was performed to assess the association between APOE-ε4 genotype and each of the five CVLT-II outcome measures. Demographic and injury variables were selected based on recommendations and cited predictive value in large studies and systematic reviews in mTBI to include age in years (scalar), education in years (scalar), sex (male vs. female), race (dichotomized to Caucasian vs. other due to small samples of other races), loss of consciousness (LOC; no vs. yes/unknown), posttraumatic amnesia (PTA; no vs. yes/ unknown), and presence of intracranial pathology on initial head computed tomography (CT) scan (Jacobs et al., 2010;Kashluba, Hanks, Casey, & Millis, 2008;Van der Naalt, 2001;Carroll et al., 2004). A history of seizures has been correlated with decreased verbal memory performance in the neurology literature and hence prior medical history (PMH) of seizures was included to serve as a surrogate for baseline cognitive burden (Blake, Wroe, Breen, & McCarthy, 2000;Schefft et al., 2008;Berl et al., 2005). To prevent overfitting, we adhered to the rule of 10 for multiple regression analysis (predictors=9; n = 114) (Howell, 2013;Green, 1991;). To account for attributable variance (R 2 ), a two-tiered approach was employed for each multivariable regression model. First, demographic and injury history predictors were entered onto the model for each CVLT-II outcome measure and R 2 was recorded (Table S1); then, APOE-ε4 status was entered onto the model, and the new R 2 as well as the ΔR 2 from the model without APOE-ε4 status was calculated and reported. The multivariable regression mean differences (B) and their associated 95% confidence intervals (CI) are reported for each predictor in the regression analyses. When appropriate, the regression means and standard errors (SE) are reported for APOE-ε4(+) and APOE-ε4(−) cohorts. Significance was assessed at α = 0.05. All analyses were performed using Statistical

Package for the Social Sciences (SPSS) version 22 (IBM Corporation,
Chicago, IL, USA).

| Demographic and injury characteristics
In total, 114 patients were included in the current analysis ( Table 1).
Demographic and injury characteristics by APOE-ε4 carrier status are shown in Table 1. Notably, ApoE-ε4(+) patients were older (49.6 vs. 39.7 years, p = .002). APOE-ε4 carrier status differed across races, with African-American/African patients showing the highest incidence, followed by Caucasians (p = .012); the higher prevalence of the ε4 allele in African-American/African patients, and of the ε3 allele in Caucasians, is consistent with prior epidemiological reports (Farrer et al., 1997;Kern et al., 2015). No other demographic or injury history differences were observed by APOE-ε4 carrier status.

| APOE-ε4(+) is not associated with immediate recall or short-delay recall at 6 months postinjury
For IRT, no significant differences were observed by APOE-ε4 status (p = .190   Race distributions are reported as row percentages. All other distributions reported as column percentages. The race subgroup "Other races" was combined due to individual small sample sizes of Asian with APOE-ε4(−) (Figure 2). Age was associated with decreased recall

| DISCUSSION
In a prospective study examining the association between APOE-ε4 and verbal memory deficits after isolated mTBI, we find that APOE-ε4(+) is independently associated with impaired long-delay free and cued recall at 6 months postinjury when compared with APOE-ε4(−) patients. We also demonstrate that intracranial CT pathology remains a strong independent driver of decreased 6-month verbal memory performance after mTBI.

Nathoo et al. noted that APOE-ε4 carriers show a greater propensity
for developing age-related cognitive impairment, a decrease in the synapse-neuron ratio, and an increased susceptibility to exogenous neurotoxins and hippocampal atrophy independent of head trauma (Nathoo, Chetry, van Dellen, Connolly, & Naidoo, 2003).
Of note, ApoE-ε4 has been associated with verbal memory deficits in mTBI but not with the other aspects of executive function. Significant hippocampal atrophy and temporal horn enlargement have a tendency to occur after TBI, and between 2-7 months postinjury, temporal horn volumes correlated with measured intelligence quotients, while hippocampal volume correlated with verbal memory function (Bigler et al., 2002). Findings in APOE-knockout mice after TBI demonstrated marked learning deficits associated with neuronal death specifically in the hippocampus. Thus, ApoE is proposed to play a key role in the neuronal repair of the hippocampus and the surrounding regions through isoform-dependent clearance of beta-amyloid depositions after trauma (Fagan & Holtzman, 2000).

| APOE-ε4(+) associates with impaired long-term verbal memory following mTBI
The CVLT-II tests encoding, consolidation, and retrieval. The longdelay recall subtests assess forgetting rates over longer intervals and provide a score of retention in the absence of retroactive interference (Delis et al., 2000). Encoding measures the relative rate of learning.
Consolidation is a postencoding process that involves maintenance, elaboration, and storage of new information in long-term memory (LTM). Evidence in the moderate and severe TBI literature has yielded mixed results- Vanderploeg et al. (2001) report no difference in encoding did not differ between TBI and controls on SDFR and SDCR, while Noe, Ferri, Colomer, Moliner, and Chirivella (2010) found APOE genotype to be associated with the trajectory of cognitive recovery but not in memory rehabilitation. In concussed athletes,  patients demonstrated greater neurocognitive variability compared to noncarriers (Merritt et al., 2017), and in mild-to-moderate TBI patients, Lieberman reported nonsignificant decreases in performance across a range of neuropsychological tests for APOE-ε4(+) patients (Liberman et al., 2002). Similarly, in a small study of 34 patients following mTBI, APOE-ε4(+) was associated with decreased withinperson postinjury performance on three of nine neuropsychological tests (Sundstrom et al., 2004). Interestingly and importantly we find that, in mTBI patients, subtests of encoding (IRT, SDFR, SDCR) are unaffected or marginally affected by APOE-ε4(+) status, while subtests of consolidation are worsened by APOE-ε4(+) status. Hence, we extend the findings that information not properly consolidated will rapidly decay from LTM, thus becoming poorly recognized on LDFR and LDCR (Vanderploeg et al., 2001). While the R 2 values for the regression models are relatively low, they are reasonable for human studies; the 2.2%-5.0% increase in R 2 from the addition of APOE-ε4 status for LDFR and LDCR, while small, is in line with expectations in a multifactorial model consisting of other validated predictors for mTBI outcome, and is deserving of attention considering it is from a single gene, but should not be overinterpreted. The value, and equipoise, of evaluating APOE as a locus of susceptibility and a target for intervention following mTBI remains to be determined in future prospective studies of larger scale and rigorous study design.
Padgett and colleagues describe in a literature review that the relationship between APOE and cognitive function following TBI is complex and requires a nuance-based approach-which can be partially elucidated through analysis of specific deficits in appropriate TBI subpopulations (e.g., mTBI vs. moderate/severe-TBI; isolated TBI vs. polytrauma) (Padgett et al., 2016). Following mTBI, episodic verbal memory deficiencies are greatest between complicated mTBI and healthy controls (Tayim, Flashman, Wright, Roth, & McAllister, 2016). Our analysis replicates this phenomenon, as CT pathology was the driver of decreased cognitive performance following mTBI, with greater deleterious associations with consolidation and retrieval rather than encoding.
This supports the idea that following ischemia and neuronal damage, the reduced antioxidant and biological activity of the ε4 allele may exacerbate vascular endothelial injury (Bell et al., 2012;Halliday et al., 2016) and lead to cognitive deficits (Friedman et al., 1999). Similar

| APOE-ε4(+), demographics, and outcome
Age and education effects on the CVLT-II subscales have been previously described (Delis et al., 2000). There is a known association between increasing age and decreasing performance (Delis et al., 2000;Wiens, Tindall, & Crossen, 1994;Paolo, Troster, & Ryan, 1997) and hence the CVLT-II manual recommends controlling for age when analyzing CVLT-II performance (Delis et al., 2000). In our sample, age was a statistically significant independent predictor of poorer verbal memory performance across all CVLT-II subscales, which may be due to a combination of normal cerebral aging processes leading to slower verbal memory processing-a finding consistent with prior reports in mTBI (Jacobs et al., 2010). It should be noted that in our sample, APOE-ε4(+) patients were on average 10 years older than ε4(−) patients, which may further compound their decreased verbal memory performance of the ε4-carriers. This also constitutes a limitation of the current analysis, and future studies should incorporate ε4(+) and ε4(−) cohorts matched by age (e.g., APOE-ε4(+) individuals vs. APOE-ε4(−) "controls") in the study design. Education years was also an independent predictor of verbal memory outcomes following mTBI in the current analysis. Previous reports regarding education effects on the CVLT-II are more unclear; Paolo et al. (1997) report that education affects only 14% of CVLT-II indices, and Wiens et al. (1994) did not find a consistent contribution of education to CVLT-II. However, education is a known predictor of protracted return to work following mTBI (Stulemeijer, van der Werf, Borm, & Vos, 2007), in part as a proxy for baseline reading and literacy capabilities (Dotson, Kitner-Triolo, Evans, & Zonderman, 2009;Albert & Teresi, 1999;), as well as postinjury coping/resilience, socioeconomic status, and social support.
The relationship between race and CVLT-II has been previously described (Delis et al., 2000;Nathoo et al., 2003). In our study, race was associated with CVLT performance after injury. The reason for this is unclear, and may be related to a variety of factors. The ε4 allele is more common in individuals of African-American/African descent compared with Caucasians (Farrer et al.,1997;Kern et al., 2015), and evidence suggests that ε4 allele effects are weaker in the former (Kaup et al., 2015). Caucasian patients are more likely to have better access to healthcare and rehabilitation following TBI (Gasquoine, 2009;).
Further, recent reports show that African-American and Hispanic patients are less likely to be placed in acute rehabilitation (Shafi et al., 2007). Hence, racial differences in access to care, and the quality of care received, may additionally influence CVLT performance following mTBI. APOE-ε4 has also been implicated in differential predilections for developing cerebrovascular disease across ethnicities (Gavett, John, Gurnani, Bussell, & Saurman, 2015). These represent important future directions for the study of APOE in mTBI.

| Limitations
This study has a relatively small population (n = 114), which may have impacted the true strength of association between APOE-ε4(+) and short-delay recall trials, as well as limiting the applicability of our study to all mTBI patients. Due to our small sample size, we were unable to control for additional confounders without risk of overfitting the regression model, which further limits the generalizability of our results.
The deleterious impacts on verbal memory attributable to APOE were also relatively small, especially in context of positive CT pathology.
While there is some evidence that APOE-ε2 may confer a protective effect on cognitive health (Suri, Heise, Trachtenberg, & Mackaey, 2013), we were not able to investigate this aside from APOE-ε4 effects due to small sample size. While we controlled for race using multivariable analysis, we were unable to assess differences attributable to APOE-ε4(+) within individual races. APOE effects in the context of disparate patterns of cognitive decline by race (Gu et al., 2015) present an additional confounder that cannot be addressed in detail by this study. Also, as the CVLT-II was administered in English, patients whose first language is not English may perform with subtle differences. While out of scope of the current analysis, the long-term associations between APOE-ε4 and outcomes after mTBI also warrant further investigation; another study in 396 severe TBI patients found no clear differences between ε4-carriers and noncarriers on neuropsychological test performance at a follow-up interval of 15-26 years postinjury (Millar, Nicoll, Thornhill, Murray, & Teasdale, 2003), while a separate study reports poorer cognitive functioning in APOE-ε4(+) individuals three decades after TBI (Isoniemi, Tenovuo, Portin, Himanen, & Kairisto, 2006;). Finally, as stated previously, the attributable R2 change to APOE-ε4 status is relatively small in the context of other validated predictors for mTBI outcome, and the value of APOE as a target for therapeutic evaluation following mTBI remains to be determined pending future work toward elucidating molecular mechanisms, and defining the criteria for heightened surveillance and/or intervention.
The TRACK-TBI Pilot study was also limited by the variables of the NIH NINDS TBI CDEs version 1, which did not include the highest level of granularity for certain periinjury measures, for example, scalar duration in minutes and/or hours of LOC and PTA which may have been useful for estimating relative mTBI severity. Additionally, the NIH NINDS TBI CDEs version 1 recommended follow-up using eu) study in the European Union (Maas et al., 2015).

| CONCLUSIONS
Presence of the APOE-ε4 allele is associated with decreased long-term verbal memory on two subscales of the CVLT-II independent of demographics, injury history, and intracranial CT pathology, suggesting disruption to consolidation and retrieval processes following mTBI.
These results support and extend previous findings between APOE-ε4 and impaired cognitive function, and preliminarily strengthen the link between long-term verbal memory deficits after mTBI and the pathophysiology of neurodegenerative disorders. Future prospective studies targeting APOE-ε4 are needed to validate these findings.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

STATEMENT
The data used in this study are stored in the Federal Interagency Traumatic Brain Injury Research (FITBIR) informatics system (https:// fitbir.nih.gov/) and are available for access to qualified researchers.