Introduction
Major depressive disorder (MDD) is associated with reduced hippocampal volumes (Campbell & Macqueen, Reference Campbell and MacQueen2004; Arnone et al. Reference Arnone, McIntosh, Ebmeier, Munafo and Anderson2012; Palazidou, Reference Palazidou2012). Although there is evidence of reduced hippocampal volumes in first-episode, untreated patients (Zou et al. Reference Zou, Deng, Li, Zhang, Jiang, Huang, Sun and Sun2010; Cole et al. Reference Cole, Costafreda, McGuffin and Fu2011) and in individuals at familial risk of depression (Baare et al. Reference Baare, Vinberg, Knudsen, Paulson, Langkilde, Jernigan and Kessing2010; Chen et al. Reference Chen, Hamilton and Gotlib2010; Amico et al. Reference Amico, Meisenzahl, Koutsouleris, Reiser, Moller and Frodl2011), meta-analyses suggest that volume reductions of the hippocampus are more pronounced with prolonged illness (McKinnon et al. Reference McKinnon, Yucel, Nazarov and MacQueen2009), a higher number of depressive episodes (Videbech & Ravnkilde, Reference Videbech and Ravnkilde2004) and more severe depression (Vakili et al. Reference Vakili, Pillay, Lafer, Fava, Renshaw, Bonello-Cintron and Yurgelun-Todd2000; Arnone et al. Reference Arnone, McIntosh, Ebmeier, Munafo and Anderson2012). Depressed patients also show significant reductions in performance in a large range of cognitive domains (Burt et al. Reference Burt, Zembar and Niederehe1995; Rock et al. Reference Rock, Roiser, Riedel and Blackwell2014), including memory and learning, some of which persist in the remitted state (Hasselbalch et al. Reference Hasselbalch, Knorr and Kessing2011). Given the prominent role of the hippocampus and surrounding brain structures in memory-related functions (Burgess et al. Reference Burgess, Maguire and O'Keefe2002; Lavenex & Banta Lavenex, Reference Lavenex and Banta Lavenex2013), structural changes in this region may therefore play a role at least in part in the cognitive dysfunction seen in MDD (Kaymak et al. Reference Kaymak, Demir, Senturk, Tatar, Aldur and Ulug2010; Trivedi & Greer, Reference Trivedi and Greer2014).
Considering the large number of studies investigating hippocampal structure in MDD and the evidence for abnormalities in neuropsychological functions believed to involve the hippocampus, it is surprising that far fewer neuroimaging studies have investigated hippocampal function in this group. Functional magnetic resonance imaging (fMRI) studies of task-related hippocampal activations during memory tasks have been inconsistent showing both increased (Young et al. Reference Young, Bellgowan, Bodurka and Drevets2014) and reduced activations (Fairhall et al. Reference Fairhall, Sharma, Magnusson and Murphy2010; Milne et al. Reference Milne, MacQueen and Hall2012) as well as no significant differences relative to healthy controls (Werner et al. Reference Werner, Meindl, Materne, Engel, Huber, Riedel, Reiser and Hennig-Fast2009). A magnetoencephalography study using a spatial navigation task found reduced theta band power in the right hippocampus and parahippocampal area in depressed patients, suggesting reduced cortical activity in this region in a spatial task (Cornwell et al. Reference Cornwell, Salvadore, Colon-Rosario, Latov, Holroyd, Carver, Coppola, Manji, Zarate and Grillon2010). Earlier H2 15O positron emission tomography (PET) studies showed increased resting cerebral blood flow (CBF) in the hippocampus of depressed patients (Videbech et al. Reference Videbech, Ravnkilde, Pedersen, Hartvig, Egander, Clemmensen, Rasmussen, Andersen, Gjedde and Rosenberg2002), but reduced blood flow increases during a verbal encoding task (Bremner et al. Reference Bremner, Vythilingam, Vermetten, Vaccarino and Charney2004). A recent study of CBF using arterial spin labelling (ASL) also showed increased resting blood flow in the hippocampus in MDD (Lui et al. Reference Lui, Parkes, Huang, Zou, Chan, Yang, Zou, Li, Tang, Zhang, Li, Wei, Chen, Sun, Kemp and Gong2009). The state of hippocampal function in depression is therefore not clear from the current evidence. Adding to this, the relationships between functional and structural abnormalities of the hippocampus in depressed patients remains largely unexplored. Differences in hippocampal function in depression could be a direct consequence of differences in hippocampal structure but may also emerge independently of such structural differences.
In the present study we investigated if hippocampal function, as measured with fMRI, is altered in depressed patients during performance of a spatial memory task. The task was designed to place specific demands on allocentric spatial memory, which is known to be dependent on the hippocampus (King et al. Reference King, Burgess, Hartley, Vargha-Khadem and O'Keefe2002; Lee et al. Reference Lee, Buckley, Pegman, Spiers, Scahill, Gaffan, Bussey, Davies, Kapur, Hodges and Graham2005). In the allocentric condition of this task, locations are remembered independently of an individual's location and orientation, which is in contrast to the egocentric condition in which locations are remembered relative to the position of the individual. We have previously demonstrated that the allocentric and egocentric conditions result in robust differences in blood oxygen level-dependent (BOLD) signal in the hippocampus in young healthy subjects, which motivates its use as a measure of hippocampal function in the present study (Nilsson et al. Reference Nilsson, Ferrier, Coventry, Bester and Finkelmeyer2013). A larger offline battery of neuropsychological tests was also completed to give a complete picture of cognitive performance in the patient group relative to the control group.
To allow an exploration of the relationship between hippocampal function and structure, the hippocampi of all subjects were manually traced and were also considered in the analyses. Furthermore, ASL provided a measure of CBF at rest in the hippocampus to aid in the interpretation of group differences in BOLD signal. Specifically, group differences in BOLD signal could not only be due to differences in neural activity, but may also reflect changes in blood flow, blood volume or oxygen metabolism, all of which contribute to the BOLD signal (Buxton, Reference Buxton2012). Considering previous findings of increased hippocampal CBF in MDD (Videbech et al. Reference Videbech, Ravnkilde, Pedersen, Hartvig, Egander, Clemmensen, Rasmussen, Andersen, Gjedde and Rosenberg2002; Lui et al. Reference Lui, Parkes, Huang, Zou, Chan, Yang, Zou, Li, Tang, Zhang, Li, Wei, Chen, Sun, Kemp and Gong2009), the ASL measure was therefore included to explore whether potential group differences in BOLD signal could be accounted for by differences in resting CBF.
The present study therefore set out to answer the following main questions: (1) do depressed patients exhibit abnormal hippocampal function relative to healthy controls, as evidenced by an altered BOLD signal in response to a spatial memory task that places demands on allocentric and egocentric memory systems? (2) Can such group differences in task-related BOLD signal be explained by differences in hippocampal volume or resting CBF?
Method
Participants
A sample of 20 depressed patients and 20 healthy controls were recruited via their consultant psychiatrists or online advertising to participate in the study. Presence or absence of MDD diagnosis was confirmed via the Mini-International Neuropsychiatric Interview (Sheehan et al. Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavs, Weiller, Hergueta, Baker and Dunbar1998). Patients were excluded if they had any other Axis I disorders other than anxiety, had previously received electroconvulsive therapy or had a change in psychiatric medication in the last 4 weeks. To take part, patients were required to have a score of 16 or above on the Hamilton Depression Rating Scale (HAM-D) (Williams et al. Reference Williams, Kobak, Bech, Engelhardt, Evans, Lipsitz, Olin, Pearson and Kalali2008). Healthy controls were excluded if they or a first-degree relative had a history of psychiatric illness or they had a score of 5 or above on the HAM-D. This meant to ensure that controls were free of even mild depressive symptoms. Patients and healthy controls were excluded from the study if they were dependent on or abusing alcohol or other drugs in the past 12 months. Individuals with conditions contraindicative to MRI were excluded from the study. All participants were right-handed.
In both groups, depressed mood and anxiety were assessed using the Beck Depression Inventory (Beck et al. Reference Beck, Ward, Mendelson, Mock and Erbaugh1961) and the State and Trait Anxiety Inventory (Spielberger, Reference Spielberger1983). For patients, age of onset, number of episodes, illness duration and current medication regimen were determined via retrospective self-report. Pre-morbid verbal intelligence quotient (IQ) was determined using the National Adult Reading Test (Nelson & Willison, Reference Nelson and Willison1991). Table 1 shows patient and healthy control sample characteristics. Groups were matched in terms of age, sex and pre-morbid IQ.
Table 1. Participant characteristics
Data are given as mean (standard deviation) unless otherwise indicated.
IQ, Intelligence quotient; HAM-D, Hamilton Depression Rating Scale; SSRI, selective serotonin reuptake inhibitors; SNRI, serotonin and noradrenalin reuptake inhibitors; AD, atypical antidepressants; TCA, tricyclical antidepressants.
a ‘Other’ includes pregabalin, lithium, aripiprazole and quetiapine.
The research was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent and the study was approved by the local Research Ethics Committee.
Neuropsychological assessment
Participants completed a number of neuropsychological tasks encompassing a range of cognitive functions, many of which are believed to engage the hippocampus. The battery included tasks of visuospatial and verbal memory: the object-relocation task (Kessels et al. Reference Kessels, Postma and de Haan1999), the Newcastle Spatial Memory Test, a computerized adaptation of a task previously described by other groups as ‘box task’ (van Asselen et al. Reference van Asselen, Kessels, Wester and Postma2005) or ‘executive golf’ task (Feigenbaum et al. Reference Feigenbaum, Polkey and Morris1996), the visual patterns test (Della Sala et al. Reference Della Sala, Gray, Baddeley and Wilson1997), digit span, and the Rey Auditory Verbal Learning Test (Schmidt, Reference Schmidt1996). Two additional tasks examined primarily executive function: the Stroop task (Golden & Freshwater, Reference Golden and Freshwater2002) and the Trail Making Test (Tombaugh, Reference Tombaugh2004).
Spatial memory fMRI task
For the fMRI scan participants performed a spatial memory task, which relied on spatial knowledge of a previously learned artificial environment which comprised a circular arena with seven spatial landmarks placed at equidistance on the walls. Prior to the scan session, participants were familiarized with this environment using a scale model. All participants were able to remember all landmark positions before entering the scanner.
During the scan, participants viewed computer renderings of the arena from a viewpoint above and slightly outside the arena (see Fig. 1). Each trial consisted of three phases: encoding, delay and recall. In the encoding phase participants were shown the arena with a single pole marking the spatial location to be remembered. In the recall phase, participants were required to recall this location by making a forced choice between two marked locations. Crucially, participants were told that during the delay phase either the walls of the arena would rotate while their own position remained fixed or that their own position would rotate around the perimeter of the arena. Whereas the wall rotation forced participants to rely on egocentric spatial relations to retrieve the target position, a change in their own position forced them to rely on allocentric spatial relations between the landmarks and the target. Since the rotations were not shown on the screen, participants were made aware of the rotation type via a verbal cue during the delay phase (‘walls’ for wall rotation or ‘you’ for rotation of viewer location, see Fig. 1). A third, control condition had identical trial structure and visuomotor demands but did not require participants to encode or recall spatial locations. Here, the empty arena was shown during the encoding phase and during the retrieval phase one of the two test locations was highlighted and participants simply had to press the button that corresponded to this highlighted location. Participants completed a total of 36 trials per condition (egocentric, allocentric, control) split into two runs. In addition, there were 36 baseline periods (9 s each) throughout the task, during which a fixation cross was presented. The length of each run was approximately 14 min.
Fig. 1. Task design and example trials for the allocentric and egocentric conditions. Not shown is that the verbal cue during the delay was displayed on top of a scrambled version of the scene. In both examples the darker dot marks the correct location and the white dot is the distractor. During ‘control’ trials, no pole was shown at encoding and during recall one of the two dots was marked with a pole.
Image acquisition and analysis
Scans took place on a 3T Achieva MR scanner (Philips Healthcare), using an eight-channel head coil as receiver. High-resolution T1 weighted anatomical images were acquired using a standard clinical three-dimensional MPRAGE sequence [repetition time (TR)/echo time (TE) = 8.5/4.6 ms, 320 × 320 matrix size, 225 slices, voxel size 0.8 mm isotropic]. This was followed by two runs of the functional scan, using a gradient-echo echo-planar-imaging sequence (TE = 30 ms, TR = 2600 ms, flip angle = 65°, voxel size 2.5 × 2.5 × 3.5 mm, 40 axial slices, 325 volumes). Lastly, a set of flow-sensitive alternating inversion recovery (FAIR) ASL scans that used an improved inversion pulse (He & Blamire, Reference He and Blamire2010) was performed (TR = 4000 ms, TE = 23 ms, 4 × 4 × 6 mm voxel size, 64 × 64 matrix size, inversion time = 1700 ms, 40 tag-control pairs; additional inversion times were used for M0 calculations), with four axial slices positioned along the length of the hippocampus.
Manual tracings of the hippocampus in both hemispheres were performed in Analyse 12.0 (Brain Imaging Resource, USA) by an experienced tracer (J.J.M.) who was blind to the status of the scans as patient or control. Tracings were performed on coronal slices and verified from axial and sagittal perspectives. The hippocampus was divided into an anterior region (including the body and head) and a posterior region (Maller et al. Reference Maller, Daskalakis and Fitzgerald2007). Hippocampal volumes were then calculated for the different subregions. To account for differences in head size, hippocampal volumes were divided by total intracranial volume which was determined from segmentation of the anatomical scan using SPM8 (Wellcome Centre for Neuroimaging, University College London, UK) in Matlab R2010b (The MathWorks Inc., USA). Tracings also served as region-of-interest (ROI) definitions for the analysis of the fMRI and ASL scan data.
SPM8 was also used for the fMRI image analysis. Pre-processing included slice-time correction, realignment/unwarping and co-registration with the anatomical scan. First-level models were calculated in native space to allow for the extraction of contrast estimates from the individually traced hippocampus ROIs. They included two regressors for the encoding phase (encoding v. control), three regressors for the delay phase and three regressors for the recall phases (allocentric recall, egocentric recall, control). The regressors for egocentric and allocentric recall further included the trial-specific reaction time as parametric modulator. All regressors were constructed as (short) boxcar functions of their respective event onsets and durations (3 s encoding, 3.75 s delay, 5 s retrieval) convolved with the canonical haemodynamic response function implemented in SPM8. Movement parameters were included as regressors of no interest. One patient had excessively moved during the fMRI scans (volume-to-volume motion exceeded half a voxel size) and was excluded from further fMRI data analysis.
The primary interest of the present paper was the difference between the allocentric and egocentric recall phase. Following the estimation of the first-level model, we therefore computed contrast images that subtracted the egocentric from the allocentric recall phase estimates. Average values for the hippocampus ROIs were then extracted from these allocentric-v.-egocentric (AvE) contrast images for statistical analysis. An additional whole-brain analysis for this contrast is reported in the online Supplementary material.
ASL images were first re-sliced to 2 mm isotropic voxel size, then re-aligned to the first scan and manually co-registered with the anatomical scan using rigid-body transformations. All subsequent ASL processing was based on extracted signal levels from the hippocampus ROIs and from grey matter and white matter segments. The average tag-control signal level difference of the FAIR scan was used to calculate blood perfusion values according to a standard kinetic model (Buxton et al. Reference Buxton, Frank, Wong, Siewert, Warach and Edelman1998). Additional parameters for these calculations (e.g. T1 of arterial blood) were based on published findings (Roberts et al. Reference Roberts, Rizi, Lenkinski and Leigh1996; Lu et al. Reference Lu, Clingman, Golay and van Zijl2004; MacIntosh et al. Reference MacIntosh, Filippini, Chappell, Woolrich, Mackay and Jezzard2010).
Statistical analyses
Statistical analyses were performed in SPSS Statistics 21 (IBM, USA). Initial group comparisons of the AvE contrast, volumes and some neuropsychological measures were performed using two-tailed, independent-sample t tests, adjusted for unequal variance if necessary. Neuropsychological measures that provided counts out of a fixed number of trials or items as a dependent measure were analysed using generalized linear models with a binary logistic, binomial distribution model. Because sex differences in CBF have previously been reported (Parkes et al. Reference Parkes, Rashid, Chard and Tofts2004), any analysis involving regional CBF included sex as a covariate. Additional group comparisons included either (regional) hippocampal volumes (AvE contrast, neuropsychological measures) or regional CBF (AvE contrast) as covariates to eliminate their potential impact on the group differences.
Results
fMRI results
Table 2 shows the results of the extracted AvE contrast for each of the four hippocampus ROIs. There were significant group differences for the AvE contrast in the two anterior ROIs, with statistical trends in the same direction for the posterior regions (see Table 2). As shown in Fig. 2, which shows BOLD signals of the two conditions relative to the control condition, this group difference appears to be driven by the control group. Compared with the control condition, controls showed reductions in BOLD signal in the allocentric condition in bilateral anterior regions (left: p = 0.030, right: p = 0.001) and right posterior ROI (p = 0.006), but not the left posterior ROI (p = 0.467), whereas no significant change from the control condition was observed in any ROI during the egocentric condition (all p > 0.211). The patient group did not show significant differences in either the egocentric or the allocentric condition relative to the control condition in any of the ROIs (all p > 0.570). A whole-brain analysis across both groups showed widespread BOLD signal differences in the AvE contrast, but no significant group differences (see online Supplementary material). In the control group, clusters of negative AvE contrast can be found in the bilateral anterior hippocampus at an uncorrected level of p < 0.001 (see online Supplementary material).
Fig. 2. Blood oxygen level-dependent signal β coefficients of the four hippocampus regions of interest for major depressive disorder (MDD) patients and healthy control (HC) participants for the allocentric (ALLO) and egocentric (EGO) retrieval phases. Error bars represent standard errors.
Table 2. Region-of-interest fMRI results of the AvE contrast of the retrieval phase
Data are given as mean (standard deviation).
fMRI, Functional magnetic resonance imaging; AvE, allocentric-v.-egocentric.
a The condition effect column shows the result of a one-sample t test of this contrast against 0 (across both groups).
b The group difference column shows the result of an independent-sample t test for the difference between patients and healthy controls. This t test did not assume equal variances.
To investigate if the group differences are specific to the hippocampus regions, we extracted contrast estimates from three additional ROIs that also showed negative BOLD signal during the allocentric condition and a negative AvE contrast (at least p < 0.001, uncorrected, in whole-brain analysis). These included a cluster in the posterior cingulate cortex [Montreal Neurological Institute (MNI): −8, −52, 30; 137 voxels], a cluster in the rostral anterior cingulate cortex (MNI: −6, 38, −6; 206 voxels) and a cluster in the medial prefrontal cortex (MNI: −2, 64, 10; 211 voxels). Comparing the AvE contrast in these regions between patients and controls showed no significant differences (all p > 0.385).
Behavioural performance
Test results of the neuropsychological tasks are shown in Table 3. Significant differences between the groups were seen in the majority of the tasks measuring verbal and spatial memory. In contrast to our earlier study in young healthy controls, in which performance in the egocentric condition of the fMRI task was near perfect (96% correct), three patients and five controls had near or subrandom task performance in this condition in the present study (less than 23 of 36 trials correct). This probably indicates non-compliance with or a misunderstanding of the instructions for this particular condition. We therefore excluded these participants from the analysis of behavioural performance of the fMRI task. As this may have also influenced the main fMRI findings, we repeated the above ROI analysis with these participants excluded. This did not fundamentally alter the results presented above (see online Supplementary information). There were no significant group differences in the behavioural performance of the spatial memory fMRI task, but the otherwise expected task effects (see online Supplementary material for detailed results).
Table 3. Results of neuropsychological testing
Data are given as mean (standard deviation).
a Group differences were calculated with independent-samples t tests, adjusted for unequal variance if necessary, or with Wald χ2 tests as part of a generalized linear model with logit link function if the measure represented a count and was limited by the task.
b Effect sizes represent coefficients of determination of ordinary-least-squares general linear models and their equivalent based on maximum likelihood for generalized linear models.
* Significant effects following correction for multiple comparison using a false discovery rate of 5%.
Hippocampal volumes
The comparison of normalized hippocampal volumes showed no significant differences between patients and controls for any of the individual ROIs or for any combined volume (all p > 0.274). Detailed results are given in the online Supplementary material.
ASL results
Group had no significant effect on CBF in any of the hippocampal regions (all p > 0.287) or for the overall grey and white matter segments of the imaged slab. There were significant main effects of sex in all but one region (left posterior, p = 0.105), with females showing higher CBF than males. See online Supplementary material for detailed ASL results.
Controlling for volume or CBF
Including the respective regional volume as a covariate in the group comparisons of the AvE contrast did not alter the earlier findings strongly. The significant group effect in the two anterior ROIs (left: p = 0.029, right: p = 0.023) and the statistical trend in the right posterior ROI (p = 0.096) remained. Furthermore, the previous trend-level group difference in the left posterior ROI was statistically significant (p = 0.038). Controlling for regional CBF and sex had little influence on the earlier AvE group effects although their p values were generally slightly increased (left anterior: p = 0.048, right anterior: p = 0.074, left posterior: p = 0.138, right posterior: p = 0.083). Previously significant group differences in the neuropsychological tasks remained significant when covarying for differences in the normalized volume of the entire hippocampus.
Discussion
The present study investigated hippocampal function and its relationship to hippocampal structure in major depression using multimodal MR imaging and neuropsychological testing. Our ROI analysis of the BOLD signal during the fMRI task showed that the contrast between allocentric and egocentric recall conditions was significantly different between the two groups in the bilateral anterior hippocampus: Healthy controls showed a reduced BOLD signal during the allocentric condition relative to the egocentric condition. In contrast, in patients there was no reduction in BOLD signal in the allocentric compared with the egocentric condition in any of the ROIs, which was also not different from the control condition.
The reduction in BOLD signal in the allocentric condition relative to the egocentric condition in the control group is consistent with our previous findings in healthy young participants (Nilsson et al. Reference Nilsson, Ferrier, Coventry, Bester and Finkelmeyer2013). This demonstrates that the level of hippocampal engagement differs between these two conditions. This BOLD signal decrease may signify a task-related reduction in hippocampal neuronal activity in response to the allocentric condition, though we have previously discussed alternative interpretations (Nilsson et al. Reference Nilsson, Ferrier, Coventry, Bester and Finkelmeyer2013). Thus, whilst this negative-going BOLD signal is difficult to interpret, it is nonetheless evidence that the different task conditions result in differential activity in the anterior hippocampus in control subjects.
In contrast, depressed patients showed no change in BOLD signal from the control condition in either the allocentric or egocentric condition in any of the hippocampal regions. Accordingly, this finding can be described as a lack of modulation of hippocampal activity in response to task demands in the patient group. Similar reduced modulations of hippocampal BOLD in response to task demands have been observed in previous studies in depression although they are often described as reduced increases in activity in the patient group (Bremner et al. Reference Bremner, Vythilingam, Vermetten, Vaccarino and Charney2004; Fairhall et al. Reference Fairhall, Sharma, Magnusson and Murphy2010; Milne et al. Reference Milne, MacQueen and Hall2012). Whilst the relative nature of the BOLD signal prevents any conclusions regarding the absolute levels of neuronal activity, the current results demonstrate that the systematic, task-related differences in underlying hippocampal activity in healthy controls are absent in depressed patients.
To test the specificity of this finding to the hippocampus we investigated three other ROIs that also showed strong negative BOLD signal in the allocentric condition, both compared with control and the egocentric condition. These regions are frequently associated with the so-called default mode network (Raichle & Snyder, Reference Raichle and Snyder2007), which shows reductions in BOLD signal during a wide variety of tasks and has been argued to be dysfunctional in depression (Dutta et al. Reference Dutta, McKie and Deakin2014). Analyses of these additional ROIs did not indicate a difference between patients and controls in the amount of reduction of BOLD signal during allocentric recall. The observed group effect in the hippocampus therefore does not appear to be the reflection of a more generic alteration of default-mode network activity in MDD. Furthermore, analyses that included hippocampal volumes or resting blood flow as a covariate did not fundamentally alter the results, indicating that the differential pattern of BOLD signal in the patient group cannot be explained by volumetric differences or by alterations in hippocampal blood flow.
Beyond the hippocampus, the fMRI task recruits a large network of brain regions, including large parts of the parietal lobe, all of which contribute to the task (Nilsson et al. Reference Nilsson, Ferrier, Coventry, Bester and Finkelmeyer2013). At the whole-brain level we did not observe any significant group differences in BOLD signal. While this could be due to the lower sensitivity of a whole-brain analysis, it suggests that patients did not use fundamentally different brain regions to perform the task. In line with such an argument, there were also no significant behavioural differences between the groups in any of the conditions of the fMRI task, but both groups showed large differences between the two conditions and (expected) increases in error rate and response time with rotation angle in the allocentric but not the egocentric condition. Thus, the observed group difference in hippocampal BOLD signal did not translate into differential performance in the task. This also suggests that group differences in BOLD signal were not due to other task-related factors (e.g. motivation, perceived difficulty). It is worth noting that the fMRI task was developed primarily to investigate differences in hippocampal function and not to detect behavioural differences.
In support of the limited sensitivity of the task for detecting behavioural group differences in the scanner, the patient group demonstrated impaired performance on a number of spatial and verbal memory tasks performed outside the scanner, which is consistent with previous research (Ravnkilde et al. Reference Ravnkilde, Videbech, Clemmensen, Egander, Rasmussen and Rosenberg2002; Hinkelmann et al. Reference Hinkelmann, Moritz, Botzenhardt, Riedesel, Wiedemann, Kellner and Otte2009). One explanation for this discrepancy could be that compared with the fMRI task, the offline spatial measures required participants to remember a greater number of locations (‘object-location binding’, ‘position-only memory’) and to remember locations over longer time periods (‘between-search errors’). It is conceivable that the higher cognitive load in the offline spatial memory measures resulted in an increased sensitivity for detecting group differences. Furthermore, it should be emphasized that the present study is not the first to demonstrate an altered modulation of hippocampal activity in MDD in the absence of behavioural group differences (Bremner et al. Reference Bremner, Vythilingam, Vermetten, Vaccarino and Charney2004; Fairhall et al. Reference Fairhall, Sharma, Magnusson and Murphy2010).
Contrary to previous findings we did not observe decreases in hippocampal volume in the patient group. This is perhaps not surprising given that a recent meta-analysis on this topic has shown only small to medium effect sizes for hippocampal volume reductions in depression (effect size approximately −0.3) with significant heterogeneity across studies (Arnone et al. Reference Arnone, McIntosh, Ebmeier, Munafo and Anderson2012). It has also been suggested that pharmacological treatment may counteract hippocampal volume reductions in MDD (Arnone et al. Reference Arnone, McKie, Elliott, Juhasz, Thomas, Downey, Williams, Deakin and Anderson2013). Since all but one patient of the current study were medicated this may have contributed to the fact that we did not find volume differences. Furthermore, it has been suggested that such reductions mostly occur in depressed patients who have been ill for some time or who have had multiple episodes (McKinnon et al. Reference McKinnon, Yucel, Nazarov and MacQueen2009). While a number of our patients certainly fell into this category, there was considerable heterogeneity in terms of clinical characteristics and age in the present sample. This, combined with the relatively moderate sample size means that the power to detect small volume differences in the present study was probably insufficient.
It is also noteworthy that, similar to the previous neuroimaging studies of hippocampal function in depression (Bremner et al. Reference Bremner, Vythilingam, Vermetten, Vaccarino and Charney2004; Fairhall et al. Reference Fairhall, Sharma, Magnusson and Murphy2010; Milne et al. Reference Milne, MacQueen and Hall2012), our group difference in BOLD signal was located in anterior regions of the hippocampus, whereas volumetric findings in depression tend to show differences in more posterior regions of the hippocampus (Neumeister et al. Reference Neumeister, Wood, Bonne, Nugent, Luckenbaugh, Young, Bain, Charney and Drevets2005; Maller et al. Reference Maller, Daskalakis and Fitzgerald2007, Reference Maller, Daskalakis, Thomson, Daigle, Barr and Fitzgerald2012; Cole et al. Reference Cole, Toga, Hojatkashani, Thompson, Costafreda, Cleare, Williams, Bullmore, Scott, Mitterschiffthaler, Walsh, Donaldson, Mirza, Marquand, Nosarti, McGuffin and Fu2010). It is possible that the different spatial resolutions of the anatomical and functional scans play a role in this apparent discrepancy. The narrow and elongated shape of the hippocampus may make the posterior part of the hippocampus more sensitive to detecting morphological changes in anatomical scans with high spatial resolution, whereas it would increase the influence of partial volume effects in lower resolution functional scans. If the same pathophysiological process that is responsible for the structural changes seen in other studies was also responsible for the functional changes seen in the current study, it may mean that functional changes can precede changes in structure, which is why we were able to detect the former but not the latter. Alternatively, it is possible that independent pathological processes are at work, with some responsible for volume changes in posterior parts of the hippocampus, while others are responsible for functional changes in the anterior hippocampus. To reduce the potential influence of partial volume effects, replications with higher-resolution functional scans would be necessary. Such research should also involve a functional task that produces task-related BOLD changes in the posterior hippocampus.
The ASL-based measure of hippocampal resting blood flow did not show differences between the two groups. In contrast, sex had a significant effect on CBF in nearly all investigated regions. Such differences between males and females have consistently been reported using both MR-based (Parkes et al. Reference Parkes, Rashid, Chard and Tofts2004; Liu et al. Reference Liu, Zhu, Feinberg, Guenther, Gregori, Weiner and Schuff2012) and PET methodologies (Henriksen et al. Reference Henriksen, Kruuse, Olesen, Jensen, Larsson, Birk, Hansen, Wienecke and Rostrup2013). This suggests that our ASL methodology was sensitive enough to detect some group differences, but if any differences in CBF between our patient and control group existed, they were too small to be detected in our sample. The previous study showing hippocampal hyperperfusion in depression using PET (Videbech et al. Reference Videbech, Ravnkilde, Pedersen, Hartvig, Egander, Clemmensen, Rasmussen, Andersen, Gjedde and Rosenberg2002) studied a patient sample that was considerably different from the present study. Whereas patients in the PET study were all in-patients and described as ‘rather acutely depressed’ (p. 38), our sample consisted of out-patients that tended to have been ill for at least some time. The PET study also showed that the differences were more pronounced in patients who were either unmedicated or had only received medication for less than 1 week (Videbech et al. Reference Videbech, Ravnkilde, Pedersen, Hartvig, Egander, Clemmensen, Rasmussen, Andersen, Gjedde and Rosenberg2002). A more recent PET study that found increased resting blood flow in the left parahippocampal gyrus in a whole-brain analysis also only studied unmedicated patients (Monkul et al. Reference Monkul, Silva, Narayana, Peluso, Zamarripa, Nery, Najt, Li, Lancaster, Fox, Lafer and Soares2012). In our sample, apart from one unmedicated patient, all patients had been on stable medication for at least 4 weeks prior to the scan. It can be speculated that blood flow changes in the hippocampus are restricted to rather short periods at the beginning of an acute depressive episode and are reversed by medication. In support of this, a large study using ASL to measure CBF in acutely depressed patients found hippocampal hyperperfusion only in a group of non-treatment-refractory patients, both compared with treatment-refractory patients and with controls (Lui et al. Reference Lui, Parkes, Huang, Zou, Chan, Yang, Zou, Li, Tang, Zhang, Li, Wei, Chen, Sun, Kemp and Gong2009). It was outside the scope of our study to determine any treatment response of our patients; thus our sample is probably a mix of treatment-refractory and non-refractory patients.
In summary, we have presented evidence for reduced modulation of hippocampal activity during a spatial memory task in depressed patients relative to healthy controls. This reduced modulation could not be accounted for by differences in hippocampal structure or resting CBF in the same region, suggesting that volumetric or baseline haemodynamic differences are not underlying the functional differences. Whether or not these differences in hippocampal function are reflective of an early degenerative process that has yet to result in observable volumetric reductions can only be answered with longitudinal investigations, which could also answer the question if and when such functional abnormalities may be reliably linked to behavioural impairments.
Supplementary material
The supplementary material for this article can be found at http://dx.doi.org/10.1017/S0033291716000702
Acknowledgements
We graciously acknowledge the help of the staff of the Newcastle Magnetic Resonance Centre, Newcastle University, in conducting this research. A.F. is supported by Research Capability Funding from the Northumberland-Tyne-and-Wear NHS Foundation Trust awarded to R.H.M.-W.
Declaration of Interest
None.
