آشنایی با مورد و بازیابی کنترل مرتبط در بیماری آلزایمر: مطالعه fMRI

Typical Alzheimer's disease (AD) is characterized by an impaired form of associative memory, recollection, that includes the controlled retrieval of associations. In contrast, familiarity-based memory for individual items may sometimes be preserved in the early stages of the disease. This is the first study that directly examines whole-brain regional activity during one core aspect of the recollection function: associative controlled episodic retrieval (CER), contrasted to item familiarity in AD patients. Cerebral activity related to associative CER and item familiarity in AD patients and healthy controls (HCs) was measured with functional magnetic resonance imaging during a word-pair recognition task to which the process dissociation procedure was applied. Some patients had null CER estimates (AD−), whereas others did show some CER abilities (AD+), although significantly less than HC. In contrast, familiarity estimates were equivalent in the three groups. In AD+, as in controls, associative CER activated the inferior precuneus/posterior cingulate cortex (PCC). When performing group comparisons, no region was found to be significantly more activated during CER in HC than AD+ and vice versa. However, during associative CER, functional connectivity between this region and the hippocampus, the inferior parietal and dorsolateral prefrontal cortex (DLPFC) was significantly higher in HC than in AD+. In all three groups, item familiarity was related to activation along the intraparietal sulcus (IPS). In conclusion, whereas the preserved automatic detection of an old item (without retrieval of accurate word association) is related to parietal activation centred on the IPS, the inferior precuneus/PCC supports associative CER ability in AD patients, as in HC. However, AD patients have deficient functional connectivity during associative CER, suggesting that the residual recollection function in these patients might be impoverished by the lack of some recollection-related aspects such as autonoetic quality, episodic details and verification.

Long-term memory impairment, as evidenced by impaired recall and recognition memory performance, is one hallmark of Alzheimer's disease (AD). According to the dual-process models of memory, two independent functions support recognition memory performance: recollection and familiarity. Recollection reflects the controlled, conscious retrieval of information, including the recovery of details from the encoding context. Familiarity reflects a relatively automatic process of global assessment of memory strength or stimulus recency without controlled access to the associated contextual information (for reviews, see Yonelinas, 2002; Yonelinas et al., 2010). In the same vein, within recognition memory assessment, item recognition can be distinguished from associative recognition. Associative recognition usually engages the ability to retrieve the relationship between individual items or between items and their context. Since recollection involves the retrieval of specific qualitative information about the event, it is generally assumed that associative memory performance depends mainly on the recollection function (although there are some exceptions, see below). On the contrary, it is often assumed that performance on item recognition tasks relies mostly on familiarity in the absence of recollection (e.g., Montaldi and Mayes, 2010; Yonelinas, 2002; Yonelinas et al., 2010). A few studies on long-term memory in AD have assessed memory for items and associations as well as the familiarity and recollection functions. Studies focussing on the distinction between recollection and familiarity have mainly used process-estimation methods, such as the process dissociation procedure (PDP) and the Remember/Know (R/K) procedure. The PDP allows one to meticulously assess controlled and automatic processes within a single memory task (Jacoby, 1991; Jacoby et al., 1993). This procedure focuses on the controlled episodic retrieval (CER) aspect of the recollection function. The term “controlled” in this framework should not be considered in an absolute sense. It refers to a goal-driven process involving conscious and analytic access to memories that counteracts a response driven by automatic memory influences. With this method, some studies have found that CER is severely impaired while automatic processes are relatively preserved or, at least, less impaired than CER (Adam et al., 2005; Knight, 1998; Smith and Knight, 2002). Analogous results have also been found in patients likely to be at a pre-dementia stage of AD such as patients with amnesic Mild Cognitive Impairment (aMCI) and in patients with questionable AD (Anderson et al., 2008; Bastin et al., 2010; Tse et al., 2010). Another process-estimation method, the Receiver Operating Characteristics curves analysis (ROCs; for a review, see Yonelinas and Parks, 2007), has been used in patients with MCI to assess recollection and familiarity. Like studies that used the PDP, Ally et al.’s (2009a) and Embree et al.’s (2012) experiments with the ROCs procedure showed that recollection was severely impaired. Interestingly, Ally et al. found that familiarity estimates were reduced for studied words in patients with MCI, while Embree et al. showed that familiarity estimates were reduced for studied words but not for studied pictures in MCI. Similar findings have been obtained by using event-related potential (ERP) measurements during a recognition task in patients with MCI by Ally et al. (2009b). These authors showed that ERP components typically associated with familiarity were diminished for studied words but not for studied pictures in patients with MCI. The recollection and familiarity functions can also be evaluated with the R/K procedure (Gardiner, 1988; Tulving, 1985). This procedure distinguishes recollection from familiarity on the basis of phenomenal experience or level of consciousness. More concretely, participants are asked to report whether they recognise items through recollection of episodic details of the encoding context (Remember) or through a feeling of familiarity without any recollective experience (Know). Most of the studies that assessed the subjective aspects of recollection and familiarity in AD patients by using the Remember-Know procedure found that the experience of remembering, but not familiarity, was significantly impaired in AD patients (Dalla Barba, 1997; Piolino et al., 2003; Rauchs et al., 2007). However, a more recent study has found that both Remember and Know responses were reduced in AD patients (Hudon et al., 2009). In contrast, to date, all the studies that have assessed phenomenal aspects of recollection and familiarity in patients with MCI have found that the recollective experience was reduced but not the feeling of familiarity (Belleville et al., 2011; Hudon et al., 2009; Serra et al., 2010). Finally, recollection and familiarity have been more indirectly assessed by means of experimental manipulations (e.g., test format: Westerberg et al., 2006; salience of fluency: Algarabel et al., 2009). Whereas Westerberg et al. found that recollection was severely impaired and familiarity relatively preserved in MCI patients, Algarabel et al. (2009) suggested that familiarity could be impaired in MCI patients. In summary, there is no consensus regarding the preservation of familiarity in patients with MCI and AD, but all studies agree that recollection is severely impaired in both patient populations. In the same vein, Irish et al. have recently examined the quality of the recollective experience for retrograde memories in AD patients (Irish et al., 2011a, 2011b). They reported that AD patients were impaired across a range of behavioural markers inherent in the recollective experience such as self-referential imagery, vividness and retrieval of contextual details. Self-reference and vividness are two main aspects of autonoesis, that characterized episodic retrieval. According to Tulving (2002), the term autonoetic has been used to refer to this special kind of consciousness that allows us to be aware of subjective time in which events happened. Autonoetic awareness (or autonoesis) is required for remembering, to re-experience, through autonoetic awareness, one's own previous experiences (Tulving, 2002). Accordingly, Irish et al. concluded that AD patients have impoverished autonoetic consciousness. The authors found an analogous profile of impairment for MCI patients' recollective experience (Irish et al., 2010). These findings highlight the complex nature of the recollective experience and suggest that, even when AD patients are able to experience subjective recollection, this process may be qualitatively different from the recollection process in healthy ageing. In particular, AD patients often fail to consciously retrieve details associated with the target memory. Similarly, Tendolkar et al. (1999) have found that AD patients who had smaller hippocampi showed an inability to recollect the study context in a verbal recognition memory task. A few studies have investigated long-term memory for associations in AD patients. Studies that have used the Paired Associate Learning Task (PAL) from the Cambridge Neuropsychological Testing Automated Battery (CANTAB; Cambridge Cognition Place, UK) suggest that patients with AD (Lee et al., 2003) and patients with MCI (De Rover et al., 2011) have deficits in associative memory. Furthermore, Fowler et al. (2002) found that performance on the PAL was impaired in early-stage AD patients, even before standard neuropsychological measures detected any deterioration. Along the same lines, Pariente et al. (2005) found that AD patients' performance was reduced in a face–name association task. One study directly investigated item and associative recognition in AD patients (Hanaki et al., 2011). The authors found that both item and associative recognition memory were impaired in AD patients. Other authors derived item and associative memory scores from standardised tests of recall in patients with MCI (Troyer et al., 2008). They found that both item and associative recall were impaired in MCI patients, but that associative recall was more impaired and more sensitive to MCI than item recall. Interestingly, it appeared that AD patients' difficulties with associations are not specific to long-term memory but are also robustly demonstrated in short-term memory (Della Sala et al., 2012; Parra et al., 2009, 2010a, 2010b; Van Geldorp et al., 2012). Finally, some studies have used an associative recognition task to disentangle recollection- versus familiarity-based memory performance. In most of these studies, participants studied word pairs and were then asked to distinguish between intact pairs (IPs) and recombined pairs (RPs) during a recognition test. In this way, Gallo et al. (2004) showed that AD patients had difficulties engaging controlled recall of an originally studied pair of words in order to reject a rearranged version of the pair. These authors interpreted the impaired recall-to-reject process as evidence of altered recollection-based monitoring in AD patients. Similarly, Algarabel et al. (2012) used the proportion of hits for IPs minus the proportion of false alarms for RPs as an index of recollection in AD patients and patients with MCI. They found that recollection scores were deficient in these patients. Wolk et al. (2008, 2011) applied the PDP to an analogous word-pair recognition task. They found that both recollection and familiarity scores were deficient in patients with MCI and AD. In summary, impaired recollection and associative memory are two robust characteristics of the memory profile in AD. These alterations may be observed even at very early stages of AD, when the criteria for dementia have not yet been met. Recollection is a complex function that engages many processes. Studies that have examined the controlled aspect of recollection in AD patients suggest that this process is impaired, whereas the impairment of automatic memory processes is subject to debate. Studies that have examined the subjective experience accompanying recollection suggest that subjective remembering is impaired, whereas whether the experience of knowing is preserved or impaired is less clear. Furthermore, studies that have investigated the quality of the recollective experience in AD patients suggest that it lacks contextual details and autonoesis. These results are similar to those of the studies of associative memory in AD patients that suggest that memory for associations, whether short- or long-term, is altered in these patients. Together, then, these findings suggest that the controlled retrieval of associations in long-term memory in AD patients should be severely deficient. In addition, one might expect the subjective experience of remembering that accompanies this retrieval to be impaired in AD. Several theories have been proposed to model how associations in long-term memory are supported by the medial temporal lobe (MTL) and therefore are associated with recollection and familiarity. According to Yonelinas (2002; Yonelinas et al., 2010), retrieval of inter-item associations is mainly supported by recollection. However, the authors noted one exception: if the two items have been encoded as a whole item in the study phase (i.e., unitisation)2, familiarity may contribute to associative memory judgements. Since the hippocampus is critical for recollection but plays no role in familiarity-based recognition, one might expect that recognition of non-unitised associations should be related to the hippocampus whereas item recognition which does not engage associative retrieval should be supported by extrahippocampal medial temporal regions (Yonelinas et al., 2010). According to the Domain-Dichotomy (DD) view (Mayes et al., 2007), unitised associations and within-domain associations, which are associations between identical or very similar kinds of items (e.g., face–face pair), might be supported by familiarity via the perirhinal cortex. In contrast, between-domain associations, which are associations between different kinds of items (e.g., face-name pair), are mainly supported by the hippocampus via recollection. However, to account for divergent findings in the literature, the authors suggested that recognition of within-domain associations, such as word pairs, is supported by the recollection function if the items have been linked at encoding by using a mediator (e.g., a sentence or a mental image). A few years later, the authors proposed a modified version of the DD view, called the Convergence, Recollection and Familiarity Theory (CRAFT; Montaldi and Mayes, 2010) in which they added a role for the parahippocampal cortex in context recognition via familiarity. A very close perspective, the Binding of Item and Context model (BIC), was proposed by Diana et al. (2007; for a review of evidence in favour of this model, see also Ranganath, 2010). Like previous models, the BIC model proposes that the perirhinal cortex is involved in item familiarity-based recognition of items whereas the hippocampus is necessary to link one item to another or to the study context, so that the hippocampus plays a crucial role in the recollection of inter-item associations. Also similarly to previous models, the BIC model predicts that the perirhinal cortex might support associative recognition on the basis of familiarity if the items are encoded as a single unit (i.e., unitised). In addition, the BIC model suggests that the parahippocampal cortex is engaged during recollection of contextual information. In summary, current theories of the implementation of associative memory in the MTL diverge on some points, but convergence can be found in several respects. First, there is agreement that the hippocampus supports the retrieval of inter-item associations via recollection in many conditions. Second, all models suggest that the perirhinal cortex may be engaged in familiarity-based retrieval of associated items if these items were encoded as a single entity. Along these lines, recent findings obtained with various procedures suggest that familiarity, supported by the perirhinal cortex, may support recognition of the source associated with a given item when that source is encoded as a feature of the item (Diana et al., 2008, 2010, 2011). Previous findings highlight the fact that most studies that have examined the brain regions associated with recollection and familiarity have focused on the MTL, mainly because recollection and familiarity were considered in the framework of associative memory versus item memory. Reviews of these studies suggest that generally, within the MTL, recollection seems to be associated with the hippocampus and the parahippocampal cortex whereas familiarity seems to be related to the perirhinal cortex (Diana et al., 2007; Eichenbaum et al., 2007). Regarding the functional magnetic resonance imaging (fMRI) data suggesting that the perirhinal cortex supports familiarity, it is noteworthy that, whereas subsequent familiarity is associated with activation in the perirhinal cortex during encoding (Davachi et al., 2003; Ranganath et al., 2004b), familiarity/item memory at retrieval has been found to be associated with deactivation in this region (Gonsalves et al., 2005; Henson et al., 2005; Weis et al., 2004). However, Skinner and Fernandes (2007) suggested that, in fMRI studies of recollection and familiarity, the involvement of the MTL in relation to these processes depends on the specific demands of the tasks and the type of information involved. In the same vein, in a meta-analysis of fMRI studies examining episodic retrieval, Spaniol et al. (2009) showed that objective (source memory) and subjective (‘remember’ reports) recollection processes share brain activations, but that some regions were specifically involved in either objective or subjective recollection. In particular, the hippocampus was active for subjective but not objective recollection. In a recent study, Slotnick (2010) showed that, during memory retrieval, the inferior parietal cortex supports subjective remembering whereas the hippocampus mediates binding of item-related information. These findings highlight the fact that familiarity and recollection are supported by many different regions and that recollection is a high-level function characterised by multiple processes and consequently is related to different brain regions depending on the specific aspect targeted by the task. However, some constants can be found in the patterns of brain activation related to recollection and familiarity functions, as already reported for the MTL. Within the parietal lobe, familiarity processes are preferentially associated with the superior parietal cortex whereas recollection processes are preferentially associated with the inferior parietal cortex (Kim, 2010; Skinner and Fernandes, 2007). Recollection processes are also preferentially associated with medial prefrontal and posterior cingulate cortex (PCC)/precuneus activation (Kim, 2010). In a recent meta-analysis, Kim (in press) has shown that the inferior portion of the posterior precuneus was more associated with high-confidence Remember responses than with low-confidence Know responses, whereas superior portions of the precuneus show the reverse pattern. Therefore, one can assume that the PCC and the inferior part of the precuneus form a whole entity involved in the recollection network. Little is known about the brain substrates of the recollection/associative memory and familiarity/item memory processes in AD patients. Regarding associative memory, it has been found that patients with MCI and AD have altered hippocampal activation when compared to healthy controls (HCs) during face-name recognition tasks (Pariente et al., 2005; Petrella et al., 2006). Furthermore, it has been found that patients with aMCI have impaired dynamic signal attenuation related to associative learning in the hippocampal region compared to HCs (Johnson et al., 2008). Interestingly, it has also been found that patients with MCI had impaired hippocampal activation when compared to HCs in an adaptation of the PAL task but only in the higher memory load condition (De Rover et al., 2011). Wolk et al. (2011) examined the relationships between the volume of MTL structures (hippocampus and extrahippocampal MTL) and the recollection/familiarity estimates derived with PDP from the word-pair recognition task in AD, MCI and healthy older participants. They found that recollection performance was more related to hippocampal volume than to extrahippocampal volume whereas familiarity performance was more associated with extrahippocampal volume than with hippocampal volume. However, no study has yet investigated how specific aspects of recollection and familiarity are related to brain functioning outside the MTL in AD patients. Bastin et al. (2010) performed cognitive–metabolic correlations between controlled memory performance and brain metabolism in questionable AD patients who were known either to remain stable or to meet the criteria for AD after 36 months. They found that CER was preferentially correlated with activity in the medial prefrontal cortex (MPFC) and the PCC in questionable AD patients who subsequently received a diagnosis of Alzheimer's dementia. This study was a first attempt to identify brain regions related to the controlled aspect of recollection in the very early stages of AD. However, several questions remain unanswered. First, the brain regions supporting familiarity for individual items in AD patients are unknown. Second, whole-brain regions associated with the recollection function, in particular CER, in demented patients are not well defined. The main objective of this study was thus to measure brain activation during the CER aspect of recollection and during item familiarity in AD patients and healthy older participants, using fMRI. An objective recollection procedure focussing on the controlled retrieval process was preferred to a subjective memory procedure for two reasons. First, the R/K procedure measures recollection and familiarity on the basis of subjective reports, which can be inaccurate in some participants, particularly in those who have impaired cognitive abilities (Baddeley et al., 2001; Yonelinas, 2002). As the task was performed in the scanner, the accuracy of patients' choices could not be assessed. In the same vein, assessment of the qualitative aspects of the recollective experience, such as quality of reliving, amount of contextual details retrieved, quality of imagery and self-involvement, should preferentially be performed using fine-grained scales, which is not easy to do in a scanner with demented patients. Since current theories and previous work suggest that retrieval of associations is a core process of the recollection function and that this process is impaired in AD patients, we applied the logic of the PDP to a word-pair recognition task in order to disentangle the recollection and familiarity functions during memory retrieval, as previously done in some behavioural studies (Wolk et al., 2008, 2011). After incidental encoding of unrelated word pairs, participants saw intact, recombined and new pairs. They were asked to make an old/new judgement on each pair. Participants were explicitly instructed to answer ‘old’ only if they had seen both words associated in the same pair during the encoding session, that is, only in case of an IP. The RPs provided a condition in which controlled retrieval of the association was opposed to item familiarity (i.e., an exclusion condition). As the participants had seen both words previously, they were familiar with these items and thus, in the absence of controlled retrieval of the original association, they might be driven to incorrectly endorse the pair as an old one. In contrast, the IPs provided a condition in which controlled retrieval was congruent with familiarity. Thus, our task allowed associative CER to be disentangled from item familiarity in recognition performance. It targeted the controlled and associative aspects of episodic retrieval, but it did not exclude other aspects of the recollective experience (i.e., imagery, autonoesis and retrieval of contextual details such as thoughts are likely to be associated with the controlled retrieval process). We examined brain regions specifically activated during item familiarity and during CER of associations. It is likely that some regions, such as the PCC and the MPFC, mainly support the controlled aspect of the recollection function and therefore might be significantly activated specifically during CER. Other regions may be significantly activated in our healthy population sample during CER since they play a role in the recollection function, such as the inferior parietal cortex, which is said to support the subjective experience of remembering (Slotnick, 2010). The design of our task does not allow an examination of how the retrieval of the associations is personally experienced by the participants and which region supports this process, since our task was designed to disentangle controlled retrieval of a pair from item familiarity. Intuitively, we can assume that healthy participants should feel that they have personally seen the association and, on the basis of current opinion regarding this process, one may assume this to be related to the inferior parietal cortex (Simons et al., 2010). According to the wide range of studies and reviews showing that the hippocampus plays a crucial role in retrieval of associated details, we should expect it to be involved in the recollection function. If these regions are related to recollection, they may be functionally connected to the regions supporting associative CER during recollection. To account for these assumptions, functional connectivity analyses were performed during recollection. At the behavioural level, we expected that associative CER would be impaired in AD patients. We also expected that item familiarity might be impaired since some studies have found it to be already altered at the MCI stage. Regarding the brain substrates of familiarity, we expected, at least in control participants, signal changes in brain regions such as the superior parietal cortex, the intraparietal sulcus (IPS) and the perirhinal cortex (Diana et al., 2007; Eichenbaum et al., 2007; Wagner et al., 2005). Regarding the recollection function, little is known about the specific role of different brain regions in the different processes. First, the brain regions that specifically mediate effortful and controlled retrieval have not yet been identified. Most studies that have examined controlled processes in episodic retrieval have focused on post-retrieval monitoring (e.g., Henson et al., 1999b). One study used an exclusion task in an fMRI scanner and suggested that this task engaged the DLPFC via post-retrieval monitoring. However, although the engagement of this region was greater for the exclusion task, it was not specific to this condition (Rugg et al., 2003). To our knowledge, Bastin et al.’s (2010) study is the only published work which has investigated brain regions related to controlled memory retrieval by using exclusion and inclusion conditions thanks to the PDP. According to that study and congruently with the findings of previous fMRI studies of episodic memory, CER is likely to be related to activations in the MPFC and the posterior cingulate gyrus (Bastin et al., 2010; Kim, 2010) in healthy participants. In addition, one might expect that concomitant subjective remembering and retrieval of item-related information will be related to engagement of the inferior parietal cortex and the hippocampus, respectively, in healthy older participants (Slotnick, 2010). Finally, we hypothesised that associative CER would be related to greater engagement of these regions in the healthy participants than the AD patients since recollection abilities are known to be reduced in AD patients. It is plausible that AD patients and healthy older participants will share brain activations in some regions, allowing for residual CER in AD patients. Moreover, one can hypothesise that, even when AD patients show some CER abilities, their experience may be impoverished, due to lower activations in brain regions supporting other aspects of the recollective experience such as autonoesis and retrieval of contextual details.

This is the first study to directly examine whole-brain regional activity engaged during one core aspect of recollection, namely associative CER contrasted to item familiarity, in AD patients. We found that CER is severely impaired whereas item familiarity is preserved in our patients. The preserved automatic detection that an item is old (but that does not permit patients to accurately retrieve encoded information) is related to a parietal region centred on the IPS in patients as in the healthy population. In contrast, residual associative controlled memory retrieval in AD patients is related to inferior precuneus/PCC activation, also as in healthy participants. However, we found altered functional connectivity between the inferior precuneus/PCC and the inferior parietal cortex, the hippocampus and the DLPFC in AD patients during recollection, suggesting that residual recollection in these patients might be impoverished by the lack of some recollection-related aspects such as autonoetic quality, episodic details and verification. Therefore, future studies are needed to carefully examine the different aspects of recollection in AD patients.