Beruflich Dokumente
Kultur Dokumente
PII: S0010-9452(18)30103-5
DOI: 10.1016/j.cortex.2018.03.022
Reference: CORTEX 2286
Please cite this article as: Audrain S, McAndrews MP, Cognitive and functional correlates of accelerated
long-term forgetting in temporal lobe epilepsy, CORTEX (2018), doi: 10.1016/j.cortex.2018.03.022.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Cognitive and functional correlates of accelerated long-term forgetting in temporal lobe epilepsy
PT
Abstract: While we know that hippocampal dysfunction is responsible for the memory deficits
RI
that patients with temporal lobe epilepsy exhibit at relatively short study-test delays, the role of
SC
this region in accelerated long-term forgetting (ALF) is not yet clear. In the present study, we
probed the role of the hippocampus in ALF by directly comparing memory for associations to
U
memory that could be supported by item recognition during a forced choice recognition task over
AN
delays ranging from 15-minutes to 72-hours. We additionally examined resting-state functional
connectivity between the hippocampus and cortical regions known to be involved in processing
M
these types of stimuli, as well as the relationship between ALF and various clinical variables
seizures across the retention period, and standardized composite memory scores. We found
TE
evidence of accelerated forgetting for item stimuli (but not associative stimuli) by 6 hours post-
EP
learning, which became statistically reliable by 72-hours. This finding suggests that unlike
controls, patients were unable to utilize novelty to reject the incorrect object-scene pair. While
C
none of the examined clinical variables were related to long-term forgetting, reduced resting-
AC
state functional connectivity between the affected anterior hippocampus and unaffected lateral
temporal cortex predicted forgetting of item stimuli over the 72-hour delay. Implications for the
role of the hippocampus in accelerated long-term forgetting, and existing theories of systems
Funding: This work was supported by the Natural Sciences and Engineering Research Council
[grant numbers RGPIN-2015-06471]
PT
RI
Highlights
SC
• Patients appeared unable to use novelty to reject items they hadn’t seen before
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
1 Introduction
While temporal lobe epilepsy (TLE) has long been associated with deficits in learning
and retrieval, research in the past decade has identified a specific problem with retention that has
PT
been termed accelerated long term forgetting (ALF) (Blake, Wroe, Breen, & Mccarthy, 2000; C
R Butler & Zeman, 2008). ALF is characterized by relatively normal learning and initial
RI
retention but rapid forgetting of long-term memories, which suggests a failure of consolidation.
SC
This phenomenon was first described in patients with epilepsy (De Renzi & Lucchelli, 1993;
Martin et al., 1991), and has since been further investigated in the context of transient epileptic
U
amnesia: a late-onset form of epilepsy characterized by discrete amnesic episodes (Asadi-Pooya,
AN
2014; Christopher R. Butler et al., 2007). ALF is of particular interest to memory researchers
given that the time-course of systems consolidation, and the time-delineated role of the
M
hippocampus in this process, has been the subject of considerable debate (Dudai, 2012; Dudai,
Karni, & Born, 2015; McGaugh, 2000; Squire & Alvarez, 1995; Winocur & Moscovitch, 2011;
D
Winocur, Moscovitch, & Bontempi, 2010). Since patients with TLE demonstrate variable
TE
structural and/or functional hippocampal abnormalities, ALF provides a unique context in which
to probe the question of hippocampal contribution to long-term memory and consolidation. Two
EP
pertinent questions have yet to be answered: at what point in time does long-term forgetting
C
The time-course of ALF in epilepsy has thus far been detected at delays spanning 1 hour
(Wilkinson et al., 2012) to 8 weeks (Blake et al., 2000), though few studies of this population
have tested for ALF with intact memory within 30 minutes of encoding, and again within the
first 24 hours after learning (in epilepsy: Bell, Fine, Dow, Seidenberg, & Hermann, 2005; Cassel,
Morris, Koutroumanidis, & Kopelman, 2016; Fitzgerald, Thayer, Mohamed, & Miller, 2013; and
1
ACCEPTED MANUSCRIPT
in transient epileptic amnesia: Muhlert, Milton, Butler, Kapur, & Zeman, 2010; Zeman et al.,
2013). Two studies have evaluated (and found) ALF within the first 12 hours of wakefulness,
suggesting that this is not necessarily a disorder of sleep-related consolidation (in TLE: Deak,
Stickgold, Pietras, Nelson, & Bubrick, 2011; in transient epileptic amnesia: Atherton, Nobre,
PT
Zeman, & Butler, 2014). Variability in estimating the onset of ALF stems from the fact that
RI
studies thus far have used different stimuli, paradigms, and delays, in what is already a
heterogeneous population. To determine the earliest time point at which ALF can be reliably
SC
detected may have important implications for the detection of ALF in the clinic given that there
is no standardized way to test for this phenomenon (though see Miller, Flanagan, Mothakunnel,
U
Mohamed, & Thayer, 2015 for normed standard clinical memory tests at longer delays). It may
AN
also speak to theories concerning the time course of consolidation, which propose relatively
“fast” molecular and synaptic changes that occur in the hippocampus and slower systems level
M
cortical brain networks (Dudai, 2012; Dudai et al., 2015; McGaugh, 2000; Winocur &
TE
While we know that hippocampal dysfunction underlies memory difficulties that TLE
EP
patients may face at shorter delays, the role of this region in ALF is unclear. Structural brain
imaging studies have revealed mixed findings. Some studies have found an association between
C
ALF and hippocampal abnormality (Narayanan et al., 2012), and more specifically with medial
AC
temporal sclerosis (MTS; Muhlert et al., 2011), though ALF was also found in individuals
without abnormality. Wilkinson and colleagues (2012) found no association between ALF and
hippocampal integrity, while Cassel et al., (2016) found that MTS was associated with forgetting
of verbal material by 10 minutes and over a week, while those without MTS only showed
2
ACCEPTED MANUSCRIPT
accelerated forgetting over longer delays. Furthermore, there is evidence that hippocampal
volume correlates with forgetting at shorter delays but not with ALF in patients with transient
epileptic amnesia (Butler et al., 2009, 2013). The functional neural correlates of ALF as
identified with neuroimaging techniques has thus far remained unexplored in both patient
PT
populations.
RI
Considering the different types of materials or testing protocols used to study ALF may
also yield insights into the engagement of medial temporal lobe and cortical regions. For
SC
instance, it has generally been found that ALF occurs for both recall and recognition (Blake et
al., 2000; Evans, Elliott, Reynders, & Isaac, 2014; Narayanan et al., 2012; Tramoni et al.,
U
2011). Recollection of specific episodic detail is known to involve the hippocampus, though
AN
over time as memories become more gist-like or schematic, recall relies to a greater extent on
neocortical areas (Moscovitch & Nadel, 1997; Winocur et al., 2010). In contrast, recognition can
M
involve recollection but also familiarity, which is the feeling of knowing you have encountered
D
something before without remembering the contextual details surrounding that event, and this
TE
has been attributed to extra-hippocampal cortices (Ranganath, 2010) . ALF has been shown for
lists of words (Deak et al., 2011; Fitzgerald et al., 2013; Narayanan et al., 2012) and visual
EP
designs (Fitzgerald et al., 2013; Narayanan et al., 2012; Wilkinson et al., 2012), and both of these
regions such as the perirhinal cortex and lateral temporal areas (Chao, Haxby, & Martin, 1999;
AC
Kravitz, Saleem, Baker, Ungerleider, & Mishkin, 2013; Ranganath & Ritchey, 2012). Another
common type of stimulus used to probe for ALF are stories (Bell, 2006; Blake et al., 2000;
Cassel et al., 2016; Evans et al., 2014; Mameniskiene, Jatuzis, Kaubrys, & Budrys, 2006;
Tramoni et al., 2011; Wilkinson et al., 2012). This type of information involves schemas,
3
ACCEPTED MANUSCRIPT
information (Tramoni et al., 2011), and depend on areas such as the anterior temporal neocortex
(Binney, Embleton, Jefferies, Parker, & Lambon Ralph, 2010; Martin & Chao, 2001). The few
studies that have used navigational tasks have found ALF (Cassel et al., 2016; Tramoni et al.,
PT
2011), while the two studies that have targeted autobiographical memory for staged events have
RI
had mixed results (Narayanan et al., 2012; Tramoni et al., 2011), and yet both navigation and
SC
(Addis, Moscovitch, Crawley, & McAndrews, 2004; O’Keefe, Burgess, Donnett, Jeffery, &
Maguire, 1998). One case study investigated memory for novel word pairs (which targets
U
hippocampal binding) and found impaired cued recall by 55 minutes in their TLE patient
AN
(McGibbon & Jansari, 2013). All of these studies suggest temporal lobe involvement, but do not
yet give a clear indication of relative hippocampal versus extrahippocampal contribution to long-
M
term forgetting.
D
What is lacking in the literature are direct comparisons of these different memory
TE
subsystems. Three studies have attempted to make these comparisons. Tramoni and colleagues
(2011) conducted a study with a small sample of 5 TLE patients who underwent various memory
EP
tasks designed to target context-bound and context-free memory, and found ALF for context-
bound tasks (story recall and recognition, recall and recognition of virtual and real routes, recall
C
and recognition of staged episode, and remote autobiographical memories) but not for context-
AC
free tasks (recognition of single items or recall of new facts). Evans and colleagues (2014) had
patients (n=7) recall and recognize stories and scenes, as well as freely recall items and their
locations in a subset of the scenes. They found ALF for story recall and recognition, as well as
for spatial and descriptive free recall in their visual task, with marginally significant ALF for
4
ACCEPTED MANUSCRIPT
visual recognition, but did not find ALF for free recall of items. The fact that these two small
studies found ALF for context-specific information implicates the hippocampus in ALF, though
Narayanan et al. (2012, n=14) found ALF for a list of words and a list of visual designs but only
a trend towards greater forgetting for a staged autobiographical event. It is also peculiar that
PT
these studies did not find any ALF for items given that ALF has been documented several times
RI
for individual words (Deak et al., 2011; Fitzgerald et al., 2013; Narayanan et al., 2012) and
visual designs (Fitzgerald et al., 2013; Narayanan et al., 2012; Wilkinson et al., 2012). Of course,
SC
the nature of memory tasks for stories, routes, items, and autobiographical events are all quite
different in terms of difficulty, material specificity, and task demands, rendering them difficult to
compare directly.
U
AN
The purpose of our study was to probe the role of the hippocampus in ALF by directly
comparing long-term memory for associations to memory that could be supported by item
M
recognition in a task designed to target both types of memory with similar stimuli, and to
D
processing such stimuli could predict accelerated forgetting. We also wanted to better gauge the
earliest time-point at which ALF could be reliably detected for associative and item memory, by
EP
testing memory multiple times over the first day of learning and beyond. Given the well-
that associative stimuli would be more sensitive to ALF than item stimuli, evidenced by faster
AC
proportion of forgetting by 6 hours after learning (given recent evidence that ALF can be
detected by 8 hours post-learning in TEA: Hoefeijzers, Dewar, Sala, Butler, & Zeman, 2015) and
that altered hippocampal connectivity to brain areas known to be involved in contextual memory
5
ACCEPTED MANUSCRIPT
2 Methods
2.1 Subjects
Twenty-three patients with temporal lobe epilepsy (Table 1) and 24 healthy controls were
PT
recruited for this study. Patients were recruited from the Epilepsy Monitoring Unit (EMU) at
RI
Toronto Western Hospital where they were being investigated for surgical candidacy. Patients
were classified as having TLE based on EEG recordings indicating exclusive temporal lobe
SC
onsets to seizures and interictal abnormalities. Given the difficulty of enlisting age-matched
controls who were able to be tested repeatedly at matched times to our inpatients, controls were
U
recruited through Amazon’s Mechanical Turk, an online crowdsourcing data collection platform,
AN
and had no self-reported history of psychiatric or neurological disorders. Overall, patients and
controls did not significantly differ on age (mean ±SD: controls: 35.25 ±9.28 years, TLE: 37.17
M
±12.35 years, F(1, 46)=0.35, p=0.65, r=0.09), years of post-secondary education (controls: 4.4
D
±2.33 years, TLE: 3.0 ±2.56 years, F(1, 46)=3.91, p=0.15, r=0.28), or gender (controls: 12 males,
TE
TLE: 15 males, X2(1)=0.21, p=0.65, φ=0.10). There was some degree of attrition during the
study for both patients and controls, the details of which can be located in Supplementary
EP
6
ACCEPTED MANUSCRIPT
PT
9 50 F <1 Left Left MTS* -1.13 No
10 26 M 18 Left Left amygdala -2.18 Yes
harmatoma
11 61 M 30 Bilateral Left MTS* -1.58 No
RI
(Left>Right)
12 47 M 31 Left Normal -1.51 n/a test
13 24 M 18 Right Right MTS* 1.32 Yes
SC
14 27 M 7 Left Normal -2.44 No
15 55 M 35 Right Right MTS* 1.32 No
16 27 M 25 Right Right 1.32 n/a test
parahippocampal
U
DNET
17 36 M 16 Right Mild right MTL -1.29 No
AN
hypersignal
18 23 F 21 Left Left 1.47 Yes
amygdala/uncus
DNET
M
glioma
23 37 F 32 Right Right MTL 0.96 n/a test
glioma*
Table 1. Clinical characteristics of patients with TLE. Composite memory scores represent
EP
standardized verbal memory for left TLE, and visual memory for right TLE. For patients with
C
bilateral TLE, scores for the hemisphere with the most epileptogenic activity was used. Laterality
AC
refers to seizure laterality. Patients with n/a test were not tested at the 72-hour delay. Asterisks
indicate subjects who had abnormal hippocampi. M=male, F=female, MTL= medial temporal
7
ACCEPTED MANUSCRIPT
chair adjacent to the bed, depending on where they were most comfortable. The curtain around
their bed was drawn to prevent visual distraction from the rest of the room. The testing laptop
PT
was placed on an adjustable table at a comfortable viewing angle, and they wore padded
RI
headphones to reduce auditory distraction from other patients and healthcare staff.
The task instructions and stimulus presentation were the same for patients and controls,
SC
except that patients were instructed to give their responses verbally and controls indicated their
responses via button press. Controls completed the study online through a combination of
U
Mechanical Turk and Inquisit 5 (2016), which is an experimental tool that allows the
AN
development and implementation of online experiments. Mechanical Turk was used to recruit
and to pay participants for their time, while Inquisit was used to display and record responses for
M
the actual experiment. Control participants were reminded to sit at their computer and ensure
D
We developed a series of object-scene pairs for this experiment. Five different sets of
study stimuli, consisting of 78 trials each, and 5 corresponding sets of test stimuli, consisting of
EP
52 trials each, were used to create different stimulus lists for each of the 5 different delays. All
pictures of objects were useable tools or gadgets, foods, or animals, and were standardized on a
C
white background measuring 300 x 300 pixels. All scenes were photos of places and activities
AC
measuring 1024 x 768 pixels, which spanned the computer screen. None of the stimuli were
Both patients and controls provided informed consent and completed a brief practice test
before the first study session. Controls were then given access to the first set of stimuli (both the
encoding and test block for a given list) and instructions on how to comply with the delay
8
ACCEPTED MANUSCRIPT
assigned for that set of stimuli. Organization of the timing of the encoding sessions, test sessions,
and delays were decided verbally between patients and the experimenter.
During the study sessions (Figure 1a), subjects viewed each object-scene pair one at a
time in the center of the screen. Each stimulus presentation began with a display of the scene for
PT
1 second, followed by a 3-second overlay of the object on the scene. Half of the objects were
RI
semantically related to the paired background, and the subject’s task was to indicate if each
object was related to the scene, the purpose of which was to promote deep encoding and
SC
consequently better recall at test (e.g. Henke, Weber, Kneifel, Wieser, & Buck, 1999). After each
response, a blank screen appeared for 4 seconds. During the last 3 seconds, a fixation cross
U
appeared in the center of the screen, promoting orientation of the subject’s gaze to a standardized
AN
center point before the next trial. Each object-scene pair was presented once and the order of
Each study block was followed by a different delay: 15 minutes, 90 minutes, 6 hours, 16
D
hours, or 72 hours. The 15-minute delay was chosen as a baseline measure of “long-term”
TE
memory to assess equality of learning and initial retention of the different types of stimuli
between patients and controls. The 90-minute and 6-hour delays were chosen to probe forgetting
EP
over the same day of learning. The 16-hour delay provided a measure of forgetting after a night’s
sleep, and finally, the 72-hour delay provided longer-term assessment of retention over days. All
C
subjects were free to do as they pleased until subsequent testing. All study and test sessions were
AC
time-stamped, and we were able to verify that each control subject correctly complied with the
At each delay, memory was assessed with the corresponding forced-choice recognition
test for the studied list (Figure 1c). Stimuli consisted of a scene from the encoding block with
9
ACCEPTED MANUSCRIPT
two objects superimposed, one on the left and one on the right. In half of the trials, one of the
objects had been seen at encoding (the target object) and one was new: these old-new (ON) trials
enabled the participant to choose the correct object by virtue of recollection or familiarity of old
items, or by identifying the new item as new, without necessarily remembering the association of
PT
the target item with the scene. We will refer to the memory processes targeted by these types of
RI
stimuli as item memory, though we acknowledge that accuracy during these trials could also be
due to recall or familiarity for the association between the object and scene. In the other half of
SC
the trials, both objects had been presented during the study phase, one having been paired with
the current background and the second paired with a different background. Familiarity of the
U
items was equated in these old-old (OO) trials, and memory for the object-scene association must
AN
be accessed to choose the correct match; hence we will refer to memory for these pairs as
associative in nature. Each stimulus appeared on the computer screen one at a time, wherein the
M
scene was presented for 1 second followed by a 3 second overlay of the objects. The
D
participants’ task was to indicate which object was originally paired with the scene. Whether the
TE
target object appeared on the right or left was randomized and equal across the trials. After each
response, a blank screen appeared for 4 seconds before onset of the next trial. During the last 3
EP
seconds, a fixation cross appeared in the center of the screen, promoting orientation of the
subject’s gaze to a standardized center point for the next trial. Each object from encoding was
C
only seen once during the testing phase, as either the target item or a familiar foil item. Each
AC
After each test session, patients took a break for at least 10-minutes, during which time
they conversed with the experimenter or visiting friends or family in order to prevent
interference of recently recalled items with learning of the next set of stimuli. In contrast,
10
ACCEPTED MANUSCRIPT
controls were given a code upon test completion, which they input back into Mechanical Turk
for payment. The experimenter then quality checked their work to ensure they had complied with
the delay, and then granted them access to the next set of stimuli, to be completed over a
PT
pattern was repeated until the subject’s memory for the visual stimuli had been tested at each of
RI
the 5 delays. The testing times, order of the delays, and stimulus lists at each delay were
counterbalanced across subjects to reduce the impact of effects of mental state due to different
SC
times of day, practice effects with repeated exposure to the task, and medication withdrawal of
the patients. However, controls had more command over the times that they were tested at, and
U
could schedule their sessions at night, while patient testing was generally restricted from 9am to
AN
7pm. The 16-hour delay lists were always studied in the evening and tested the next day after a
night of sleep. Each patient was tested over the period of a week in an effort to complete testing
M
before discharge from the hospital, and each control had two weeks to complete all of the delays.
D
Controls were allowed more time overall, because they had busier schedules that were more
TE
11
ACCEPTED MANUSCRIPT
PT
RI
U SC
AN
M
Figure 1. Stimuli and procedure. a) Each subject studied a series of object-scene pairs and
decided whether the object was related to the background scene. b) After each study session,
D
participants experienced a variable delay. c) Subjects were asked to choose which object was
originally paired with the scene. In half of the test-stimuli, the target object was presented
TE
alongside a new object to create old-new (ON) pairs. In the other half, the target object was
presented alongside a familiar object they had viewed during the study phase with a different
EP
background (OO pairs). Subjects studied new images and completed the corresponding
recognition task across all five delays, where possible. In this representation, the object:scene
C
ratio is not to scale, and there was an inter-trial fixation cross presented in the experiment.
AC
account for the odd missing response (total correct/total responses). These proportion correct
scores were used for statistical analysis of baseline memory (memory at 15 minutes). For all
12
ACCEPTED MANUSCRIPT
‘standardize’ forgetting rates given individual differences in baseline memory. Data wasn’t
included in the analysis for both patients and controls if the test was taken more than 15 minutes
late for the 15-minute delay, more than 30 minutes late for the 90-minute delay, more than 2
PT
hour late for the 6-hour delay, or more than 3 hours late for the 16-hour and 72-hour delays. Six
RI
controls had missing or inaccurate baseline data. For the two individuals who did not complete
the baseline test, we used their accuracy scores at 90 minutes as baseline to calculate proportion
SC
forgetting at the 6-hour, 16-hour, and 72-hour delays, given that there were no significant effects
U
R2=0.01) on accuracy scores at baseline and 90 minutes. The other four individuals were late in
AN
completing the baseline test, wherein they took it 40-50 minutes after study rather than 15-30
minutes. We used their scores at these times to calculate proportion forgetting at later delays, but
M
omitted these people from any analyses of baseline memory given that they had not completed
D
the test within 15-30 minutes after study, rendering them less comparable to the TLE group at
TE
this dely.
All statistical analysis was carried out using R studio version 1.0.136 (RStudio Team,
EP
2016). We used linear mixed effects models for the majority of analyses, which is a type of
maximum likelihood estimation (Cnaan, Laird, & Slasor, 1997). All mixed effects models were
AC
computed using the lme function of the nlme package (Pinheiro, Bates, DebRoy, Sarkar, & R
Core Team, 2017) in R studio, with random intercepts for each subject as the random effect
structure, to account for inter-subject variability. Degrees of freedom were estimated with a
between-within method, and the model was estimated with an unstructured covariance matrix. A
13
ACCEPTED MANUSCRIPT
weighted least squares estimation method using a constant variance structure function was used
to correct for heteroskedasticity where applicable (Pinheiro & Bates, 2000). Multiple
comparisons were corrected for each family of tests using False Discover Rate (FDR)
(Benjamini & Hochberg, 1995) analyzed with the p.adjust function in R Studio, and all reported
PT
p-values for planned comparisons are corrected. Effect sizes were calculated with Pearson
RI
correlations for all continuous predictors and 2-level categorical predictors, while semi-partial R2
was used for categorical predictors with more than 2 levels. Pearson correlations of 0.1 to 0.29
SC
were considered small effect sizes, 0.3 to 0.49 were considered medium effect sizes, and
correlations of 0.5 or greater were considered large effects (Cohen, 1992). An alpha level of
U
0.05 was used to determine statistical significance for all analyses.
AN
M
surgical planning. Testing was either concurrent with the EMU stay or carried out up to several
weeks later. For the current study, we used composite scores reflecting verbal and visual
EP
memory, as described in previous work from our lab (St-Laurent et al., 2014). In brief, these
C
scores were obtained from principle components analyses of clinical measures of intellectual
AC
functioning (verbal and performance IQ scores), verbal memory (Rey Auditory Verbal Learning
Test total recall and percent retained over a 20-min delay, Warrington Recognition Memory
Test) and visual memory (total recall on Rey Visual Design Learning Test, Warrington Face
Recognition Memory Test, total trials on a conditional associative learning test). Composite
scores in Table 1 can be considered as pseudo-z scores, as they indicate how the individual
14
ACCEPTED MANUSCRIPT
patient compares to the normative sample of TLE patients in terms of material-specific memory.
We have previously shown the resulting components for IQ, Verbal Memory and Visuospatial
Memory to provide high classification accuracy for left and right TLE patients as well as reliable
PT
RI
2.4 Neuroimaging Methods
SC
2.4.1 Data acquisition
Functional images were acquired for 16 out of the 23 subjects, 14 of whom had also
U
participated in the 72-hour delay (7 females, 35.86 ±12.18 years of age). Images were acquired
AN
on a 3T Sigma MR System (GE Medical Systems, Milwaukee). A whole brain high-resolution
anatomical scan was acquired for each subject using a 3D fast spoiled gradient echo sequence
M
with the following parameters: flip angle: 12º, TR=7.88 ms, 120 sagittal slices, FOV: 22 x
22cm2, 256 x 256 matrix, 0.9 x 0.9 x 1 mm voxels, 1-mm thick. Each subject also received a 6-
D
minute task-free resting state T2*-weighted functional MRI scan with an echo-planar pulse
TE
imaging sequence. The parameters were as follows: TR=2000 ms, TE=25 ms, FOV: 24 x 24
cm2, 64 x 64 matrix, 3.75 x 3.75 x 5 voxels, 5-mm thick, 28-32 slices depending on head size,
EP
for 180 volumes. During the scan, subjects were instructed to lie still and “not to think about
C
2.4.2 Preprocessing
Preprocessing and resting state connectivity analysis was conducted with CONN
(Whitfield-Gabrieli & Nieto-Castanon, 2012), a toolbox that runs through MATLAB v8 12.0
(Mathworks). The first three frames of each fMRI scan was dropped to allow for signal
15
ACCEPTED MANUSCRIPT
equilibrium. Anatomical and functional images were reoriented to the anterior commissure, and
each functional scan was coregistered to each subject’s anatomical image, which was then
spatially realigned and unwarped. Each subject’s anatomical image was segmented and
normalized to the T1-weighted NBI152, and the normalization parameters were written to the
PT
functional data to align the anatomical and functional scans. fMRI data were then smoothed
RI
using a Gaussian kernel of 8-mm full-width half-maximum. A MNI152-T1-weighted image was
segmented into gray matter, white matter, and cerebrospinal fluid. The data were temporally
SC
preprocessed to restrict the analysis to frequencies of interest (<0.1 Hz). We used linear
regression to remove potential sources of noise such as estimated subject motion parameters (3
U
translation components and 3 rotation components) and removal of BOLD covariates in white
AN
matter and CSF areas with the anatomical component based noise correction method
(aCompCor; Behzadi, Restom, Liau, & Liu, 2007). Scans were flipped across the midline for
M
individuals with RTLE, in order to align the affected medial temporal lobes (MTLs) for the
D
functional connectivity analysis given that we did not have the statistical power to examine left
TE
In addition to examining the rate of forgetting for item and associative stimuli, we sought
AC
OO pairs. Given the episodic nature of our task, we hypothesized that the hippocampus would be
involved in both types of memory, but that resting-state connectivity associated with memory
would differ depending on the type of stimulus involved. In recent years, the literature has
started to emphasize that the hippocampus is not functionally homogeneous along its anterior-
16
ACCEPTED MANUSCRIPT
posterior axis (Adnan et al., 2016; Poppenk, Evensmoen, Moscovitch, & Nadel, 2013; Strange,
Witter, Lein, & Moser, 2014). The anterior hippocampus demonstrates greater connectivity to
anterior regions in the brain, and of particular relevance here, it is connected to the perirhinal
cortex (which has been shown to be atrophic in a sample of patients with TEA; Butler et al.,
PT
2013; Kahn, Andrews-Hanna, Vincent, Snyder, & Buckner, 2008; Libby, Ekstrom, Ragland, &
RI
Ranganath, 2012) as well as anterior and lateral temporal cortex (Adnan et al., 2016; Kahn et al.,
2008; Poppenk & Moscovitch, 2011), which are regions involved in object processing and item
SC
memory (Poppenk et al., 2013; Ranganath & Ritchey, 2012). We therefore chose to target
connectivity between the anterior hippocampus and these lateral temporal regions, with the
U
hypothesis that connectivity here would covary with forgetting of stimuli that targets item
AN
memory (ON stimuli). In contrast to the anterior hippocampus, the posterior hippocampus is
more connected to posterior regions in the brain, including parahippocampal cortex (Kahn et al.,
M
2008; Libby et al., 2012) and posterior cingulate cortex (PCC, Adnan et al., 2016; Kahn et al.,
D
2008; Libby et al., 2012; Poppenk & Moscovitch, 2011). This posterior memory system is
TE
McCormick, St-Laurent, Ty, Valiante, & McAndrews, 2015; Poppenk et al., 2013; Ranganath &
EP
Ritchey, 2012). Given this literature, as well as findings from our own lab that suggest
connectivity between the hippocampus and PCC is a good index of memory integrity in TLE
C
(McCormick et al., 2014; McCormick, Quraan, Cohn, Valiante, & McAndrews, 2013) and
AC
between these regions would covary with forgetting of information where the object-scene
association must be evoked (OO pairs), given that this condition required remembering the
context that the items were viewed in. We used anterior and posterior hippocampal masks that
17
ACCEPTED MANUSCRIPT
were created as part of another project in our lab. These masks were created by segmenting the
anterior and posterior hippocampus, using a k-means clustering algorithm. This produced an
anterior and posterior ROI for each hippocampus in a data driven way. Detailed methods of this
PT
procedure are described in Supplementary material (Supplementary Figure 1). We arbitrarily
RI
used the left anterior and posterior hippocampal masks for analysis of the affected hippocampus,
but the left and right hippocampal masks derived from the aforementioned analysis were very
SC
similar.
Masks for the lateral temporal cortex (LTC) and PCC were created with 8-mm spheres at
U
the following MNI coordinates: left middle temporal gyrus [-60 -24 -18], right middle temporal
AN
gyrus [60 -24 -18], left PCC [-8 -56 26], and right PCC [8 -56 26]. These coordinates are based
Reidler, Sepulcre, Poulin, & Buckner, 2010), which the authors showed were involved in self-
D
memory. Indeed, we have shown that connectivity using ROIs from this network is a useful
The time series of voxels within each ROI was averaged, and those from the anterior and
posterior hippocampal ROIs were correlated with the time series of every other ROI for each
subject. These correlation values were then converted to z scores with a Fisher transformation.
Each subject’s score of proportion forgotten of OO and ON stimuli from baseline to 72-hours
was converted to a z-score relative to control subject’s group mean and standard deviation, to
18
ACCEPTED MANUSCRIPT
quantify how abnormal long-term forgetting was for each subject. We chose to focus on the 72-
hour delay because we were the most confident that abnormal forgetting at this delay would be
reflective of ALF in the context of delays used in the existing literature. Each subject’s z-scores
of ON and OO forgetting at 72-hours relative to controls were added to the analysis as second-
PT
level covariates, to determine if hippocampal connectivity to the a priori defined regions
RI
covaried with long-term forgetting. Statistical maps of connectivity were thresholded at p<0.05,
one-tailed (we were only interested in instances of reduced connectivity), with FDR correction.
U SC
3 Results
AN
3.1 Task Results
We first wanted to determine if there were differences in baseline retention (memory at
M
15-minutes) within and between the two groups. There were significant differences in memory
for OO and ON stimuli at baseline within groups (controls: F(1, 20)=19.89, p<0.01, r=0.71, TLE:
D
F(1, 22)=11.34, p<0.01, r=0.58), with higher accuracy for ON pairs than for OO pairs in both
TE
controls (ON: 97.18 ±3.07%, OO: 89.15 ±6.61% ) and patients (ON: 95.39 ±5.71%, OO: 88.18
EP
±8.53%). There were no significant differences in accuracy scores between groups for either ON
(F(1, 42)<0.01, p=0.98, r<0.01) or OO pairs (F(1, 42)=0.06, p=0.98, r=0.04), suggesting that
C
both groups were able to initially learn and retain both types of stimuli to a comparable degree.
AC
Next, we investigated how group, delay, and stimulus-type affected forgetting over time.
We ran a linear mixed effects model with proportion forgotten as the dependent variable, and
group, delay, stimulus-type, and their interactions as fixed effects. We found significant main
effects of group (F(1, 42)=4.68, p=0.04, r=0.32), delay (F(3, 238)=9.27, p<0.01, R2=0.10), and
stimulus type (F(1, 238)=15.26, p<0.01, r=0.25), where subjects generally forgot more over time,
19
ACCEPTED MANUSCRIPT
patients forgot more than controls, and more OO pairs were forgotten than ON pairs. There was
also a significant interaction between group and stimulus-type (F(1, 238)=4.29, p=0.04, r=0.13),
indicating that patients and controls forgot OO or ON pairs at different rates. Although delay did
not significantly interact with group (F(3, 238)=0.12, p=0.95, R2<0.01) or stimulus-type (F(3,
PT
238)=0.34, p=0.79, R2<0.01), we probed the group by stimulus-type interaction across the delays
RI
given that one of our research questions concerned identifying the point in time where
accelerated long-term forgetting becomes detectable in TLE for these different stimulus types.
SC
We next examined forgetting for OO and ON stimuli over time within each group by
running linear mixed effects models at each delay, with proportion forgotten as the dependent
U
variable and type of stimulus as the predictor. While controls forgot little on average over 90
AN
minutes for both types of stimuli (ON: 0 ±8.49%, OO: 3.14 ±14.38% , F(1, 15)=0.65, p=0.43,
r=0.20), they did forget a greater proportion of OO pairs than ON pairs by 6 hours (ON: 1.83
M
±6.66%, OO: 11.13 ±9.48%, F(1, 17)=11.60, p<0.01, r=0.64), and 16 hours (ON: 2.97 ±6.61%,
D
OO: 9.72 ±10.69%, F(1, 17)=6.42, p=0.04, r=0.52). This difference was not significant at 72-
TE
hours after FDR correction (ON: 7.00 ±9.34%, OO: 14.33 ±14.15%, F(1, 16)=3.34, p=0.12)
though the difference still produced a medium effect size (r=0.42) suggesting that OO pairs were
EP
indeed forgotten more than ON pairs but this difference was not quite reliable enough to denote
significance. In contrast, patients forgot ON and OO pairs at the same rate over time, with no
C
significant differences evident between forgetting of the two types of stimuli at 90 minutes (ON:
AC
4.49 ±11.36%, OO: 7.35 ±9.31%, F(1, 20)=1.06, p=0.68), 6 hours (ON: 8.11 ±10.95%, OO:
11.18 ±13.22%, F(1, 21)=0.97, p=0.68), 16 hours (ON: 9.57 ±12.34%, OO: 10.49 ±13.85%, F(1,
19)=0.06, p=0.89) or 72 hours (ON: 16.01 ±10.46%, OO: 16.52 ±13.79, F(1,15)=0.02, p=0.89).
20
ACCEPTED MANUSCRIPT
While there were small effects of stimulus-type at 90 minutes (r=0.22) and 6 hours (r=0.21),
these effects become minuscule over time (16 hours: r=0.06, 72 hours: r=0.04).
Next, we compared proportion forgotten for OO and ON pairs between groups over time
(Figure 2), in order to determine if either type of stimulus was forgotten at an accelerated rate in
PT
patients compared to controls, and at what time point that occurred by. To do this, we ran linear
RI
mixed effects models at each delay with proportion forgotten as the dependent variable, and
group as the predictor for each stimulus type. We found no significant differences in forgetting
SC
for OO stimuli between patients and controls at any delay (90mins: controls: 3.14 ±14.38%,
TLE: 7.35 ±9.31%, F(1, 35)=1.04, p=0.99, r=0.17, 6 hours: controls: 11.13 ±9.48%, TLE: 11.18
U
±13.22%, F(1, 38)<0.01, p=0.99, r<0.01, 16 hours: controls: 9.72 ±10.69%, TLE: 10.49
AN
±13.85%, F(1, 36)=0.04, p=0.99, r=0.03, 72 hours: controls: 14.33 ±14.15%, TLE: 16.52 ±13.79,
F(1, 31)=0.20, p=0.99, r=0.08). In contrast, while patients and controls forgot ON stimuli at a
M
comparable rate by 90 minutes (controls: 0 ±8.49%, TLE: 4.49 ±11.36%, F (1, 35)=2.09, p=0.16
D
r=0.24), patients with TLE started to forget ON stimuli at a faster rate by 6 hours (controls: 1.83
TE
±6.66%, TLE: 8.11 ±10.95%, F(1, 38)=4.98, p=0.05, r=0.34) and 16 hours (controls: 2.97
±6.61%, TLE: 9.57 ±12.34%, F(1, 36)=4.33, p=0.05, r=0.33), and this difference survived FDR
EP
correction for statistical significance by 72 hours (controls: 7.00 ±9.34%, TLE: 16.01 ±10.46%,
F(1,31)=6.77, p=0.04, r=0.42). These results indicate that patients were forgetting OO and ON
C
pairs at the same rate due to an increase in forgetting of ON pairs beyond that which is normal,
AC
21
ACCEPTED MANUSCRIPT
PT
RI
SC
Figure 2. Group comparisons of forgetting over time (relative to baseline memory). Left panel shows forgetting
of OO pairs for patients and controls, while the right panel shows forgetting of ON pairs for the two groups.
Error bars represent standard error, tildes (~) represent marginal statistical significance (p=0.05 after FDR
U
correction) and asterisks (*) represent statistical significance (p<0.05 after FDR correction).
AN
M
Another goal of this study was to examine whether resting state functional connectivity
TE
between the hippocampus and areas known to be involved in associative and item memory could
predict long-term forgetting of such stimuli at 72 hours. We first examined connectivity with the
EP
posterior hippocampus (Figure 3, left panel). We found that while functional connectivity
between the affected posterior hippocampus and the affected PCC correlated with forgetting of
C
OO stimuli, such that less connectivity between these regions predicted greater forgetting
AC
(r=0.45, t(12)=-1.75, p=0.05), this relationship did not survive correction for multiple
connectivity with any of the other ROIs and forgetting of OO or ON pairs. Turning to the
anterior hippocampus (Figure 3, right panel), connectivity between the affected anterior
hippocampus and unaffected LTC predicted forgetting of ON pairs, where worse connectivity
22
ACCEPTED MANUSCRIPT
between these regions predicted greater forgetting at 72 hours (r=0.62, t(12)=-2.77, p=0.03 FDR
corrected). Connectivity between the anterior hippocampus and the rest of the ROIs did not
PT
RI
U SC
AN
M
D
Figure 3: Correlations between connectivity and forgetting over the 72-hour delay. The panel on the left
shows the relationship between forgetting of OO stimuli and resting state functional connectivity between
TE
the affected posterior hippocampus and PCC. The panel on the right represents the relationship between
forgetting of ON stimuli and connectivity between the affected anterior hippocampus and unaffected
EP
lateral temporal cortex. Blue circles represent patients who had seizures during the delay. Asterisks (*)
denote significance after FDR correction.
C
AC
We ran a linear mixed effects model where z-scores of forgetting of ON stimuli relative to
controls at 72 hours was the dependent variable, with laterality of seizure focus, clinical
composite memory score (representing verbal memory for patients with left TLE, and visual
23
ACCEPTED MANUSCRIPT
memory for patients with right TLE), presence of hippocampal abnormality (including MTS, and
tumours or lesions directly affecting the hippocampus), and presence of seizures over 72 hours as
predictors. None of these variables predicted forgetting (laterality of focus: F(2, 6)=0.37, p=0.70,
r=0.24, composite memory score: F(1, 6)=0.83, p=0.40, r=0.35), presence of MTL abnormality:
PT
F(1, 6)=0.31, p=0.60, r=0.22), presence of seizures: F(1, 6)=0.03, p=0.86, r=0.07). Given that
RI
connectivity between the anterior hippocampus and LTC predicted accelerated forgetting of ON
pairs, we also wanted to determine if connectivity between these two regions was related to any
SC
clinical variables. We re-ran the same analysis with connectivity between the affected anterior
hippocampus and unaffected LTC as the dependent variable and found no reliable relationship
U
between connectivity and laterality of seizure focus (F(2, 5)=0.08, p=0.92, r=0.13), composite
AN
memory score (F(1, 5)=0.09, p=0.78, r=0.13), or presence of MTL abnormality (F(1, 5)=0.20,
p=0.67, r=0.20). Patients who had seizures over the 72-hour delay tended to have worse
M
connectivity, but this relationship was not reliable despite its medium effect size (F(1, 5)=1.31,
D
p=0.30, r=0.46).
TE
EP
4 Discussion
We found that accelerated long-term forgetting can be reliably identified at the group
C
level in patients with TLE by 72-hours after learning (r=0.42), though consideration of effect
AC
sizes indicate that greater forgetting is already present at 90-minutes (r=0.24, a small effect), and
is marginally significant by 6 hours (r=0.34, a medium effect, p=0.05 after FDR correction).
Contrary to expectation, we found that Old-New pairs of stimuli (which enabled the use of item
memory or familiarity/novelty detection to aid in the selection of the correct target) were
sensitive to ALF, while Old-Old pairs (which required associative memory) were not. While
24
ACCEPTED MANUSCRIPT
both controls and patients with TLE had better item memory than associative memory at
baseline, controls forgot OO pairs faster than ON pairs while patients forgot both types of
information at the same rate, due to accelerated forgetting of the ON pairs. Forgetting of ON
stimuli over the 72-hour delay was related to reduced resting state functional connectivity
PT
between the affected anterior hippocampus and unaffected lateral temporal cortex in patients,
RI
compatible with a disruption of systems consolidation. With respect to clinical variables of
interest, there was some evidence that reduced connectivity between these regions was related to
SC
presence of seizures over the delay, but ALF itself, as defined by more rapid loss of accuracy
than seen in controls, was not related presence of seizures, side of epileptic focus, clinically-
U
relevant memory test performance, or presence of structural MTL abnormality.
AN
4.1 No evidence of ALF for old-old pairs
M
The fact that at baseline (15 minutes post-learning), patients retained more pairs in the
item memory than associative memory condition agrees with literature concerning the relative
D
integrity of item memory and familiarity in this population (e.g. Bowles et al., 2010) . Indeed,
TE
recognition accuracy was better when decisions could be based on item familiarity/novelty rather
than associative information at all delays in both controls and TLE (see Supplementary Figure
EP
2). Surprisingly, we found baseline (15-minute) recognition and forgetting rates for associative
C
stimuli were comparable for patients and controls, despite studies including from our own lab
AC
showing that TLE patients have more pronounced deficits for associative relative to item
binding operations (Cohn, McAndrews, & Moscovitch, 2009; Saling, 2009). Furthermore, the
functional neuroimaging literature indicates that relational binding and associative retrieval
typically engages the hippocampus to a greater extent than encoding or retrieval of individual
25
ACCEPTED MANUSCRIPT
items (Davachi, 2006; Ranganath, 2010), and imaging and lesion data indicate that such binding
operations can be evident even at very short study-test delays (Hannula, Tranel, & Cohen, 2006;
McAndrews, Girard, Wilkins, & McCormick, 2016; Olson, Page, Sledge Moore, Chatterjee, &
Verfaellie, 2006). The lack of a behavioral impairment associated with MTL damage or
PT
dysfunction in the current study may be a function of specific methodological choices, including
RI
the richness of encoding afforded by the use of visually complex scenes or the retrieval support
SC
least a main effect of group on associative memory and therefore the possibility that a different,
sub-optimal strategy was used by our online controls cannot be ruled out. Indeed, pilot data with
U
younger controls tested in person appeared to show overall better performance specifically in this
AN
condition (see Supplementary Figure 2).
In the context of ALF, no one to our knowledge has directly compared memory for items
M
and associations using the same type of stimuli. While most studies of ALF have used stimuli
D
amenable to item memory such as lists of words (Deak, Stickgold, Pietras, Nelson, & Bubrick,
TE
2011; Fitzgerald, Thayer, et al., 2013; Mameniskiene et al., 2006; Narayanan et al., 2012) or
visual designs (Fitzgerald et al., 2013; Giovagnoli, Casazza, & Avanzini, 1995; Mameniskiene et
EP
al., 2006; Narayanan et al., 2012; Wilkinson et al., 2012), only one TLE case study (McGibbon
& Jansari, 2013) has targeted associative memory and they did find ALF for pairs of words,
C
though they did not test item memory. Nevertheless, other types of stimuli that are known to be
AC
hippocampus-sensitive have failed to elicit ALF. Tramoni and colleagues (2011) found ALF for
a real world navigational task and for a staged autobiographical event, though their findings were
hindered by a small sample size (n=5). Cassel and colleagues (2016) found that while patients
with TLE and controls had comparable memory for a virtual navigation task at 30 seconds,
26
ACCEPTED MANUSCRIPT
patients showed greater forgetting by 10-minutes after learning, which was maintained a week
later. However, 30 seconds may not be the best marker of healthy initial “long-term” memory,
which is a defining feature of ALF. Narayanan and colleagues (2012) used similar delays of
many weeks but failed to replicate Tramoni et al.’s (2011) finding of ALF for a staged
PT
autobiographical event, though they did note a medium to large effect size, suggesting that a
RI
difference might become significant with more statistical power. Needless to say, detecting ALF
SC
4.2 ALF for old-new pairs
U
In concert with much of the research on ALF that examines memory for single items, or
AN
stimuli that do not emphasize specific contextual details, we did find evidence of ALF for ON
pairs. In fact, TLE patients showed equivalent forgetting for OO and ON pairs, whereas controls
M
demonstrated a shallower forgetting curve for ON stimuli. Although formally the recognition
decision participants made was the same as in OO pairs (i.e., which object was studied in
D
conjunction with the background scene), ON stimuli afforded the use of item familiarity or,
TE
conversely, detection of item novelty to aid recognition. For those stimuli, one could simply
reject the less familiar/novel object as an appropriate response without needing to specifically
EP
evaluate the item-context relationship. There is considerable evidence for impaired novelty
C
detection in conjunction with MTL damage in both rodents who explore novel and familiar
AC
environments or objects equally (Burke, Wallace, Nematollahi, Uprety, & Barnes, 2010;
McTighe, Cowell, Winters, Bussey, & Saksida, 2010; Romberg et al., 2012) and patients with
mild cognitive impairment and Alzheimer’s disease who show abnormally high false alarm rates
to novel stimuli in recognition (Abe et al., 2011; Hildebrandt, Haldenwanger, & Eling, 2009).
27
ACCEPTED MANUSCRIPT
global amnesia during a recognition task to stimuli the patient thought were novel, even though
many of these stimuli had been seen a few minutes before (Westmacott, Silver, & McAndrews,
2008). Thus, a failure of this mechanism in TLE patients could explain why this condition is
PT
In the context of hippocampal damage, interference may play a significant role in failure
RI
to discriminate novel and familiar items (Sadeh, Ozubko, Winocur, & Moscovitch, 2014).
Yeung, Ryan, Cowell, and Barense (2013) found that individuals with mild cognitive impairment
SC
viewed novel objects as though they had been seen before under conditions of high visual
feature-level interference, which suggests that the novelty signal in question fails under
U
situations where pattern separation is required. The present experiment was not designed to
AN
target pattern separation, and it was very rare that the two objects at any given test trial would be
visually similar. We did however use some visually similar objects across the experiment (e.g.
M
there are about 5 different lightbulbs scattered throughout the 5 stimulus lists), and each
D
recognition trial was always in the context of a familiar background (all of the backgrounds at
TE
test were previously seen during the encoding period). It is possible that some of the objects look
visually similar enough to cause some interference, or otherwise the familiar background
EP
different cortical regions help illuminate the neurobiological basis of ALF in TLE. There was a
memory network nodes (posterior hippocampus and ispilateral PCC) and OO forgetting rate.
28
ACCEPTED MANUSCRIPT
This pattern of connectivity is known to be involved in associative retrieval (Adnan et al., 2016;
Poppenk et al., 2013), which was not selectively affected in our sample. In contrast, forgetting
for ON pairs was significantly correlated with decreased resting state connectivity between the
affected anterior hippocampus and unaffected lateral temporal cortex, regions that have been
PT
considered to be related to object processing and item memory (Poppenk et al., 2013; Ranganath
RI
& Ritchey, 2012). It is unclear as to why it is connectivity to the unaffected, rather than the
epileptogenic, anterolateral temporal cortex that predicts ALF. We know that there is decreased
SC
connectivity to the affected anterior temporal cortex in TLE patients (McCormick et al., 2013),
that seizure activity is more likely to arise from anterior than posterior hippocampus (King,
U
Bronen, Spencer, & Spencer, 1997) and that interictal discharges more frequently propagate in
AN
that direction (Emerson, Turner, Pedley, Walczak, & Forgione, 1995) through the uncinate
fasciculus. Here, we postulate that when reduced connectivity extends bilaterally it may
M
compromise functional reserve and impact memory more dramatically, but note that we do not
D
have resting state scans for our controls to be able to make this comparison.
TE
interference or decay. The literature draws a distinction between synaptic consolidation, which
C
is thought to involve strengthening of local connections over minutes or hours, and systems
AC
consolidation, which engages a broader network of regions with a time-course of days to months.
As a function of systems consolidation, some memories, particularly those which represent gist
of an experience rather than perceptual detail, become reliant on neocortical structures and
resilient to hippocampal damage (Moscovitch & Nadel, 1997; Squire & Alvarez, 1995; Winocur
& Moscovitch, 2011). If we hypothesize ALF is in fact a systems consolidation deficit in the face
29
ACCEPTED MANUSCRIPT
of adequate encoding and immediate retrieval, the faster loss of ON pairs in TLE could be due to
establish those memory traces (Kapur et al., 1997; Squire & Alvarez, 2005). Thus, it is possible
PT
that the issue is not so much one of gross hippocampal dysfunction, which would prevent
RI
memories from being adequately encoded but rather a disruption of memory transfer or cortical
instantiation. The observed reduction in resting cross-talk between the hippocampus and lateral
SC
temporal cortex may reflect this, and the fact that we saw a medium effect of seizures as
predictive of worse connectivity between these regions suggests this transfer may be disrupted
U
by epileptic activity. Future studies with larger samples will be needed to evaluate this putative
AN
mechanism as well as to address the contributions of other neocortical regions in this process
over time. Given that existing clinical measures (such as structural damage to the medial
M
temporal lobe, laterality of seizure focus, performance on clinical memory tests) do not seem to
D
5 Limitations
C
There are several limitations to our findings that should be noted. There was considerable
AC
variability in the clinical characteristics of our sample of patients in terms of side of epileptic
focus, exact site of lesion, number of seizures across delays, etc., and these differences likely
contributed to the relatively large individual differences in patient forgetting rates. As with many
clinical studies, exploration of these multiple sources of variability systematically requires fairly
30
ACCEPTED MANUSCRIPT
large cohorts but that also provides the opportunity to discover biological mechanisms that can
Furthermore, for both patients and controls, baseline accuracy and forgetting on the item task
PT
over the 90 minute-delay was at ceiling for some of the participants as evidenced by the standard
deviation including the 100% accuracy and 0% forgetting respectively, which may indicate over-
RI
learning. Overlearning has been associated with reduced forgetting over the first 24-hours,
though this effect is less prevalent by two days post-learning (Driskell et al., 1992). It is therefore
SC
possible that early forgetting of ON stimuli was masked by overlearning in some individuals in
U
both groups. Given that our associative memory condition was more difficult than the item
AN
memory condition and ran the risk of falling to floor over 72-hours, and given that we wanted to
make direct comparisons between these two conditions, we did not want to increase the difficulty
M
of the current task. Nonetheless, future studies will be needed to probe recognition in the face of
novel lures over time, where ceiling effects are completely avoided.
D
Of additional importance, the patients in this study were recruited from the EMU, and were
TE
being evaluated for surgical candidacy. As part of this process, many of them were undergoing
medication withdrawal in order to increase the likelihood of epileptic activity to be captured for
EP
clinical considerations. Thus far, studies have generally shown no relationship between
C
medication and ALF (Christopher R. Butler et al., 2007; Cassel et al., 2016; Fitzgerald et al.,
AC
2013; Muhlert et al., 2011), though there is mixed evidence concerning if ALF is related to
epileptic activity (Cassel et al., 2016; Fitzgerald et al., 2013; Mameniskiene et al., 2006; Muhlert
et al., 2011). We ensured that testing for each patient began as soon as possible after admission,
and counterbalanced the order of the delays across participants to mitigate these effects as much
as possible. Patients were tested at bedside, and while we controlled the environment as much as
31
ACCEPTED MANUSCRIPT
possible by drawing the curtain around their beds and giving them padded headphones, this
environment is not the quiet, standardized test room used in most cognitive studies. Likewise,
while we asked controls to complete the study and test sessions in a quiet environment with no
distractions, we cannot verify that they complied. In addition to this, controls completed study
PT
and test sessions around their own schedules, and while patients tended to be tested during
RI
working hours, many controls completed the tests early in the morning or later at night to prevent
interference with their own work schedules. Finally, since patients spent the duration of their stay
SC
in bed, they likely napped more than controls. Since sleep benefits associative memory (e.g.
Ellenbogen, Hulbert, Stickgold, Dinges, & Thompson-Schill, 2006) this may be a reason why we
U
did not see the expected difference between patients and controls for our associative stimuli.
AN
Likewise, we do not know how differences in sleep quality between the two groups may have
impacted the results, which may vary due to epileptic activity in the patient group, potential
M
noise at night due to other patients in the same room experiencing seizures, or simply being less
D
Despite the limitations of these testing environments, we feel that this was the most feasible
way to conduct an experiment of this nature. One of the challenges of probing forgetting over
EP
multiple delays concerns convincing participants to visit the lab for multiple sessions. While
piloting this task with undergraduate students (see Supplementary Figure 2), we found that
C
AC
subjects typically were only willing to participate in 2-3 test sessions. It is particularly difficult to
probe memory retention over inconvenient delays on the order of hours since subjects must stay
within proximity of the testing location. We felt that it was important to probe forgetting over
hours and days given that this time-frame has remained relatively unexplored not only in patients
with TLE, but also in healthy populations, and given that the brain is likely undergoing important
32
ACCEPTED MANUSCRIPT
processes of reorganization over these periods of time (Dudai et al., 2015). We also felt that
testing the same participants over multiple delays as much as possible was important given the
degree of individual variability in forgetting we observed in our pilot study, and in order to
enhance statistical power. Finally, undergraduate students are much younger than many of our
PT
patients, and we felt that it was important to match the ages of our samples given that memory
RI
changes with age. Future studies will certainly be needed to replicate our findings with closer
matching of test environments for patients and controls. In fact, to the extent that we decide to
SC
forgo collecting EEG data during the study, our paradigm is clearly adaptable to being used by
both patients and controls over many different delays with online administration.
U
AN
6 Conclusions
M
To conclude, we found evidence of accelerated forgetting for item stimuli, wherein patients
D
with TLE appeared unable to utilize novelty as a cue to reject incorrect object-scene pairs. There
TE
was evidence of greater forgetting for these stimuli in patients by 6 hours after learning, which
laterality of seizure focus, presence of MTL abnormality, and presence of seizures over the 72-
C
hour delay did not predict ALF, forgetting for item stimuli correlated with reduced resting state
AC
functional connectivity between the affected anterior hippocampus and unaffected lateral
temporal cortex. The observed decrease in resting communication between the hippocampus and
neocortical regions known to support object processing and item memory is in line with theories
interaction.
33
ACCEPTED MANUSCRIPT
Acknowledgements: The authors would like to thank Doug McQuiggan for his help in setting
up the experiment, and Dr. David Gold for his role in patient recruitment. We would also like to
thank Alexander Barnett and Vincent Man for their creation of the anterior and posterior
PT
hippocampal masks used in the functional connectivity analysis.
RI
U SC
AN
M
D
TE
C EP
AC
34
ACCEPTED MANUSCRIPT
References
Abe, N., Fujii, T., Nishio, Y., Iizuka, O., Kanno, S., Kikuchi, H., … Mori, E. (2011). False item
https://doi.org/10.1016/j.neuropsychologia.2011.03.015
PT
Addis, D. R., Moscovitch, M., Crawley, A. P., & McAndrews, M. P. (2004). Recollective
RI
qualities modulate hippocampal activation during autobiographical memory retrieval.
SC
Adnan, A., Barnett, A., Moayedi, M., McCormick, C., Cohn, M., & McAndrews, M. P. (2016).
U
Brain Structure and Function, 221(6), 2999–3012. https://doi.org/10.1007/s00429-015-
AN
1084-x
Andrews-Hanna, J. R., Reidler, J. S., Sepulcre, J., Poulin, R., & Buckner, R. L. (2010).
M
562. https://doi.org/10.1016/j.neuron.2010.02.005.Functional-Anatomic
TE
Asadi-Pooya, A. a. (2014). Transient epileptic amnesia: a concise review. Epilepsy & Behavior :
Atherton, K. E., Nobre, A. C., Zeman, A. Z., & Butler, C. R. (2014). Sleep-dependent memory
consolidation and accelerated forgetting. Cortex; a Journal Devoted to the Study of the
C
Bell. (2006). WMS-III Logical Memory Performance after a Two-Week Delay in Temporal
https://doi.org/10.1016/j.jbiomech.2011.04.025.MOMENT
Bell, B., Fine, J., Dow, C., Seidenberg, M., & Hermann, B. (2005). Temporal Lobe Epilepsy and
35
ACCEPTED MANUSCRIPT
the Selective Reminding Test: The Conventional 30-Minute Delay Suffices. Psychological
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and
PT
https://doi.org/10.2307/2346101
RI
Binney, R. J., Embleton, K. V., Jefferies, E., Parker, G. J. M., & Lambon Ralph, M. A. (2010).
The ventral and inferolateral aspects of the anterior temporal lobe are crucial in semantic
SC
memory: Evidence from a novel direct comparison of distortion-corrected fMRI, rTMS, and
https://doi.org/10.1093/cercor/bhq019
U
AN
Blake, R. V, Wroe, S. J., Breen, E. K., & Mccarthy, R. A. (2000). Accelerated forgetting in
patients with epilepsy Evidence for an impairment in memory consolidation. Brain, 123,
M
472–483.
D
Bowles, B., Crupi, C., Pigott, S., Parrent, A., Wiebe, S., Janzen, L., & Köhler, S. (2010). Double
TE
https://doi.org/10.1016/j.neuropsychologia.2010.05.010
Burke, S. N., Wallace, J. ., Nematollahi, S., Uprety, A. R., & Barnes, C. A. (2010). Pattern
C
Butler, C. R., Graham, K. S., Hodges, J. R., Kapur, N., Wardlaw, J. M., & Zeman, A. Z. J.
(2007). The syndrome of transient epileptic amnesia. Annals of Neurology, 61(6), 587–598.
https://doi.org/10.1002/ana.21111
36
ACCEPTED MANUSCRIPT
Butler, C. R., & Zeman, A. Z. (2008). Recent insights into the impairment of memory in
epilepsy : transient epileptic amnesia , accelerated long-term forgetting and remote memory
impairment. https://doi.org/10.1093/brain/awn127
Butler, Bhaduri, A., Acosta-Cabronero, J., Nestor, P. J., Kapur, N., Graham, K. S., … Zeman, A.
PT
Z. (2009). Transient epileptic amnesia: regional brain atrophy and its relationship to
RI
memory deficits. Brain : A Journal of Neurology, 132(Pt 2), 357–68.
https://doi.org/10.1093/brain/awn336
SC
Butler, van Erp, W., Bhaduri, A., Hammers, A., Heckemann, R., & Zeman, A. (2013). Magnetic
resonance volumetry reveals focal brain atrophy in transient epileptic amnesia. Epilepsy &
U
Behavior : E&B, 28(3), 363–9. https://doi.org/10.1016/j.yebeh.2013.05.018
AN
Cassel, A., Morris, R., Koutroumanidis, M., & Kopelman, M. (2016). Forgetting in temporal
M
https://doi.org/10.1016/j.cortex.2016.02.005
D
Chao, L. L., Haxby, J. V, & Martin, A. (1999). Attribute-based neural substrates in temporal
TE
cortex for perceiving and knowing about objects. Nature Neuroscience, 2(10), 913–919.
https://doi.org/10.1038/13217
EP
Cnaan, A., Laird, N. M., & Slasor, P. (1997). Tutorial in Biostatistics: Using the General Linear
https://doi.org/10.1037/0033-2909.112.1.155
Cohn, M., McAndrews, M. P., & Moscovitch, M. (2009). Associative reinstatement: A novel
37
ACCEPTED MANUSCRIPT
https://doi.org/10.1016/j.neuropsychologia.2009.06.029
Davachi, L. (2006). Item, context and relational episodic encoding in humans. Current Opinion
PT
Deak, M. C., Stickgold, R., Pietras, A. C., Nelson, A. P., & Bubrick, E. J. (2011). The role of
RI
sleep in forgetting in temporal lobe epilepsy: a pilot study. Epilepsy & Behavior : E&B,
SC
De Renzi, E., & Lucchelli, F. (1993). Dense retrograde amnesia, intact learning capability and
U
https://doi.org/10.1016/S0010-9452(13)80253-5
AN
Driskell, J. E., Park, W., Willis, R. P., Copper, C., Corporation, F. M., & Ave, E. L. (1992).
M
Dudai, Y. (2012). The Restless Engram: Consolidations Never End. Annual Review of
TE
Dudai, Y., Karni, A., & Born, J. (2015). The Consolidation and Transformation of Memory.
EP
Ellenbogen, J. M., Hulbert, J. C., Stickgold, R., Dinges, D. F., & Thompson-Schill, S. L. (2006).
C
Interfering with Theories of Sleep and Memory: Sleep, Declarative Memory, and
AC
https://doi.org/10.1016/j.cub.2006.05.024
Emerson, R. G., Turner, C. A., Pedley, T. A., Walczak, T. S., & Forgione, M. (1995).
38
ACCEPTED MANUSCRIPT
Evans, S. J., Elliott, G., Reynders, H., & Isaac, C. L. (2014). Can temporal lobe epilepsy surgery
https://doi.org/10.1016/j.neuropsychologia.2013.11.007
PT
Evensmoen, H. R., Ladstein, J., Hansen, T. I., Møller, J. A., Witter, M. P., Nadel, L., & Håberg,
RI
A. K. (2015). From details to large scale: The representation of environmental positions
follows a granularity gradient along the human hippocampal and entorhinal anterior-
SC
posterior axis. Hippocampus, 25(1), 119–135. https://doi.org/10.1002/hipo.22357
Fitzgerald, Z., Thayer, Z., Mohamed, A., & Miller, L. a. (2013). Examining factors related to
U
accelerated long-term forgetting in epilepsy using ambulatory EEG monitoring. Epilepsia,
AN
54(5), 819–27. https://doi.org/10.1111/epi.12090
Giovagnoli, A. R., Casazza, M., & Avanzini, G. (1995). Visual Learning on a Selective
M
Reminding Procedure and Delayed Recall in Patients with Temporal Lobe Epilepsy.
D
Hannula, D. E., Tranel, D., & Cohen, N. J. (2006). The Long and the Short of It: Relational
8352–8359. https://doi.org/10.1523/JNEUROSCI.5222-05.2006
Henke, K., Weber, B., Kneifel, S., Wieser, H. G., & Buck, A. (1999). Human hippocampus
C
Hildebrandt, H., Haldenwanger, A., & Eling, P. (2009). False recognition helps to distinguish
patients with Alzheimer’s disease and amnestic MCI from patients with other kinds of
39
ACCEPTED MANUSCRIPT
https://doi.org/10.1159/000235643
Hoefeijzers, S., Dewar, M., Sala, S. Della, Butler, C., & Zeman, A. (2015). Accelerated Long-
PT
Kahn, I., Andrews-Hanna, J. R., Vincent, J. L., Snyder, A. Z., & Buckner, R. L. (2008). Distinct
RI
Cortical Anatomy Linked to Subregions of the Medial Temporal Lobe Revealed by Intrinsic
SC
https://doi.org/10.1152/jn.00077.2008
King, D., Bronen, R. A., Spencer, D. D., & Spencer, S. S. (1997). Topographic distribution of
U
seizure onset and hippocampal atrophy: Relationship between MRI and depth EEG.
AN
Electroencephalography and Clinical Neurophysiology, 103(6), 692–697.
https://doi.org/10.1016/S0013-4694(97)00090-4
M
Kravitz, D. J., Saleem, K. S., Baker, C. I., Ungerleider, L. G., & Mishkin, M. (2013). The ventral
D
visual pathway: An expanded neural framework for the processing of object quality. Trends
TE
Libby, L. A., Ekstrom, A. D., Ragland, J. D., & Ranganath, C. (2012). Differential Connectivity
EP
6560. https://doi.org/10.1523/JNEUROSCI.3711-11.2012
AC
Mameniskiene, R., Jatuzis, D., Kaubrys, G., & Budrys, V. (2006). The decay of memory
between delayed and long-term recall in patients with temporal lobe epilepsy. Epilepsy and
Martin, A., & Chao, L. L. (2001). Semantic memory and the brain: Structure and processes.
40
ACCEPTED MANUSCRIPT
4388(00)00196-3
Martin, R. C., Loring, D. W., Meador, K. J., Lee, G. P., Thrash, N., & Arena, J. G. (1991).
Impaired long-term retention despite normal verbal learning in patients with temporal
PT
lobe dysfunction. Neuropsychology, 5(1), 3-12. http://dx.doi.org/10.1037/0894-
RI
4105.5.1.3
McAndrews, M. P., Girard, T. A., Wilkins, L. K., & McCormick, C. (2016). Semantic
SC
congruence affects hippocampal response to repetition of visual associations.
U
McCormick, C., Protzner, A. B., Barnett, A. J., Cohn, M., Valiante, T. A., & McAndrews, M. P.
AN
(2014). Linking DMN connectivity to episodic memory capacity: What can we learn from
https://doi.org/10.1016/j.nicl.2014.05.008
D
McCormick, C., Quraan, M., Cohn, M., Valiante, T. A., & McAndrews, M. P. (2013). Default
TE
mode network connectivity indicates episodic memory capacity in mesial temporal lobe
McCormick, C., St-Laurent, M., Ty, A., Valiante, T. A., & McAndrews, M. P. (2015).
https://doi.org/10.1093/cercor/bht324
https://doi.org/10.1126/science.287.5451.248
McGibbon, T., & Jansari, A. S. (2013). Detecting the onset of accelerated long-term forgetting:
41
ACCEPTED MANUSCRIPT
https://doi.org/10.1016/j.neuropsychologia.2012.11.004
McTighe, S. M., Cowell, R. A., Winters, B. D., Bussey, T. J., & Saksida, L. M. (2010).
Paradoxical False Memory for Objects After Brain Damage. Science, 330(6009), 1408–
PT
1410. https://doi.org/10.1126/science.1194780
RI
Miller, L. a., Flanagan, E., Mothakunnel, A., Mohamed, A., & Thayer, Z. (2015). Old dogs with
SC
Epilepsy & Behavior, 45, 205–211. https://doi.org/10.1016/j.yebeh.2015.01.024
Moscovitch, & Nadel. (1997). Memory Consolidation , Retrograde Amnesia , and the Temporal
U
Lobe Memory consolidation , retrograde amnesia and the hippocampal complex. Current
AN
Opinion in Neurobiology, (October 2012), 217–227.
Muhlert, N., Grünewald, R. a, Hunkin, N. M., Reuber, M., Howell, S., Reynders, H., & Isaac, C.
M
L. (2011). Accelerated long-term forgetting in temporal lobe but not idiopathic generalised
D
https://doi.org/10.1016/j.neuropsychologia.2011.04.018
Muhlert, N., Milton, F., Butler, C. R., Kapur, N., & Zeman, a Z. (2010). Accelerated forgetting
EP
https://doi.org/10.1016/j.neuropsychologia.2010.07.001
C
Narayanan, J., Duncan, R., Greene, J., Leach, J.-P., Razvi, S., McLean, J., & Evans, J. J. (2012).
AC
https://doi.org/10.1016/j.yebeh.2012.06.038
O’Keefe, J., Burgess, N., Donnett, J. G., Jeffery, K. J., & Maguire, E. A. (1998). Place cells,
42
ACCEPTED MANUSCRIPT
https://doi.org/10.1098/rstb.1998.0287
Olson, I. R., Page, K., Sledge Moore, K., Chatterjee, A., & Verfaellie, M. (2006). Working
PT
Memory for Conjunctions Relies on the Medial Temporal Lobe. J Neurosci, 26(17), 4596–
RI
4601. https://doi.org/10.1523/JNEUROSCI.1923-05.2006
Poppenk, J., Evensmoen, H. R., Moscovitch, M., & Nadel, L. (2013). Long-axis specialization of
SC
the human hippocampus. Trends in Cognitive Sciences, 17(5), 230–240.
https://doi.org/10.1016/j.tics.2013.03.005
U
Poppenk, J., & Moscovitch, M. (2011). A hippocampal marker of recollection memory ability
AN
among healthy young adults: Contributions of posterior and anterior segments. Neuron,
Ranganath, C. (2010). A unified framework for the functional organization of the medial
D
temporal lobes and the phenomenology of episodic memory. Hippocampus, 20(11), 1263–
TE
1290. https://doi.org/10.1002/hipo.20852
Ranganath, C., & Ritchey, M. (2012). Two cortical systems for memory-guided behaviour.
EP
Romberg, C., McTighe, S. M., Heath, C. J., Whitcomb, D. J., Cho, K., Bussey, T. J., & Saksida,
C
https://doi.org/10.1093/brain/aws074
Sadeh, T., Ozubko, J. D., Winocur, G., & Moscovitch, M. (2014). How we forget may depend on
43
ACCEPTED MANUSCRIPT
https://doi.org/10.1016/j.tics.2013.10.008
Saling, M. M. (2009). Verbal memory in mesial temporal lobe epilepsy: Beyond material
Squire, R. L., & Alvarez, P. (1995). Reterograde amnesia and memory consolidation: a
PT
neurobiological persepctive. Current Opinion in Neurobiology, (5), 169–177.
RI
St-Laurent, M., McCormick, C., Cohn, M., Mišić, B., Giannoylis, I., & McAndrews, M. P.
(2014). Using multivariate data reduction to predict postsurgery memory decline in patients
SC
with mesial temporal lobe epilepsy. Epilepsy & Behavior, 31, 220–227.
https://doi.org/10.1016/j.yebeh.2013.09.043
U
Strange, B. A., Witter, M. P., Lein, E. S., & Moser, E. I. (2014). Functional organization of the
AN
hippocampal longitudinal axis. Nature Reviews Neuroscience, 15(10), 655–669.
https://doi.org/10.1038/nrn3785
M
Tramoni, E., Felician, O., Barbeau, E. J., Guedj, E., Guye, M., Bartolomei, F., & Ceccaldi, M.
D
https://doi.org/10.1093/brain/awr002
EP
Westmacott, R., Silver, F. L., & McAndrews, M. P. (2008). Understanding medial temporal
activation in memory tasks: Evidence from fMRI of encoding and recognition in a case of
C
Wilkinson, H., Holdstock, J. S., Baker, G., Herbert, A., Clague, F., & Downes, J. J. (2012).
epilepsy. Cortex; a Journal Devoted to the Study of the Nervous System and Behavior,
44
ACCEPTED MANUSCRIPT
Winocur, G., & Moscovitch, M. (2011). Memory Transformation and Systems Consolidation.
https://doi.org/10.1017/S1355617711000683
Winocur, G., Moscovitch, M., & Bontempi, B. (2010). Memory formation and long-term
PT
retention in humans and animals: Convergence towards a transformation account of
RI
hippocampal-neocortical interactions. Neuropsychologia, 48(8), 2339–2356.
https://doi.org/10.1016/j.neuropsychologia.2010.04.016
SC
Zeman, A. Z. J., Butler, C. R., Muhlert, N., Milton, F., Wilkinson, H., Holdstock, J. S., … Asadi-
Pooya, A. a. (2013). Can temporal lobe epilepsy surgery ameliorate accelerated long-term
U
forgetting? Epilepsy & Behavior : E&B, 54(3), 622–30.
AN
https://doi.org/10.1016/j.neuropsychologia.2013.11.007
M
D
TE
C EP
AC
45