Hippocampus shape deformation: potential diagnostic biomarker for chronic back pain in women

Abstract

Sex differences in the quality and prevalence of chronic pain are manifold, with females generally presenting higher incidence and severity. Uncovering chronic pain-related sex differences informs neural mechanisms and may lead to novel treatment routes. In a multi-center morphological study (total n=374), we investigated whether the shape of subcortical regions would reflect sex differences in back pain. Given the hormone-dependent functions of the hippocampus, and its role in the transition to chronic pain, this region constituted our primary candidate. We found that the anterior part of the left hippocampus (alHP) presented outer deformation in females with chronic back pain (CBP), identified in CBP in the USA (n=77 females vs. n=78 males) and validated in a Chinese dataset (n=29 females vs. n=58 males with CBP, in contrast to n=53 female and n=43 male healthy controls). Next, we examined this region in subacute back pain (SBP) who persisted with back pain a year later (SBPp; n=18 females vs. n=18 males), and in a subgroup with persistent back pain for 3-years. Weeks after onset of back pain there was no deformation within alHP, but at 1-year and 3-years females exhibited a trend for outer deformation. The alHP partly overlapped with the subiculum and entorhinal cortex, whose functional connectivity, in healthy subjects, was associated with emotional and episodic memory related terms (Neurosynth, reverse inference). These findings suggest that in females alHP undergoes anatomical changes with pain persistence, highlighting sexually-dimorphic involvement of emotional and episodic memory-related circuitry with chronic pain.

1. INTRODUCTION

Ample evidence points to sex differences in the prevalence and severity of clinical pain. Women comprise a majority of people with chronic pain, with several non-sex specific conditions being more common in women [61; 62]. Proposed explanations for such pain-related sex disparities are multifaceted and range from psychological and sociocultural to hormonally and genetically driven differences in brain neurochemistry [30; 62], while rodent model studies of persistent pain indicate differential sex-specific roles of microglia and astrocytes in spinal cord nociceptive processes [21; 29; 33; 50] and in immune responses to injury [48; 49]. Still, the involvement of sex specific brain mechanisms in the transition to chronic pain remain to be uncovered.

Human neuroimaging studies demonstrate that brain anatomy and activity are major determinants of transition to chronic pain; a process which carves brain anatomy and function into a novel state, characterizing distinct chronic pain conditions [5; 6; 8-10; 12; 28; 32; 35; 54; 55; 64; 66; 87]. Complimentary rodent studies confirm and extend these findings by establishing causal relationships between various mesolimbic neuronal states of excitability, receptor expression, functional properties, and anatomical changes, all accompanying or controlling transition to chronic pain [7; 20; 46; 47; 60; 65; 69; 72; 73]. Within this conceptual construct, one expects to observe brain, especially limbic, anatomical differences between male and female patients with chronic pain, either due to differential risk or as a consequence of distinct impact of the persistence of pain in each sex. Here, we study limbic brain anatomy, at the scale of local shape deformations, in chronic back pain (CBP) and in early onset back pain, as a function of sex.

We specifically examine the anatomy of the hippocampus, a limbic brain region involved in the predisposition and maintenance of chronic pain. Earlier studies have shown low hippocampal volume imparting risk for CBP [87], impaired hippocampal neurogenesis in spared-nerve injured animal models [4; 65], functional connectivity changes during transition to chronic pain [51; 64], and hippocampal shape mediating memory-bias in CBP [13]. In addition, the emotional interplay between learning and memory are considered important contributors to the transition to chronic pain [3; 86], and these constitute primary hippocampal functions. Thus, given its hormone-dependent sex-specific responses to stress [36; 57; 71], we investigated the hippocampus as a sex-dependent biomarker for chronic pain, with shape deformations potentially revealing subtle reorganization at the sub-regional level. On exploratory grounds, we additionally examined potential sex and CBP relationship in other subcortical regions, to evaluate alternative theories, including dopaminergic and descending noxious inhibitory control pathways [30; 58].

In this multi-center brain imaging study, we followed a rigorous approach to ensure validity of results. We tested the reproducibility and robustness of findings from data collected in one center by testing replicability in two separate datasets. We found that, relative to the group average, women with CBP displayed an outward deformation in the shape of the anterior left hippocampus (alHP), with properties suggesting that the deformation is a consequence of living with pain, and not a risk-factor for CBP.

2. METHOD

2.1. Study Participants

We used data from three separate studies. The two main data sets are from studies assessing neurobiological mechanisms of CBP: one was used for primary analyses (discovery), and the other was used to test validity of findings as well as verifying specificity of findings to CBP relative to healthy state. The third data set was comprised by individuals with subacute pain (SBP) who had persistent pain 1-3 years later (SBPp), and was used to infer specificity of findings to CBP versus SBP.

2.1.1. Discovery data set (USA CBP, n=155):

The discovery data set is part of a clinical trial investigating mechanisms of placebo in CBP participants, conducted at Northwestern University (Chicago, USA) between 2014 and 2018 (NCT02013427 and NCT02986334) [85]. General inclusion criteria comprised low back pain for 6 months or longer, 18 years or older, average pain intensity of 5 on a 0-10 numeric rating scale, and no additional neurological, chronic pain, or psychiatric comorbidities. In the present study, we included baseline data only, which was collected prior to any study therapeutic interventions.

2.1.2. Validation data set (China CBP and CON, n=183):

The validation data set is part of a clinical trial investigating mechanisms of CBP before and after percutaneous endoscopic lumbar discectomy [38]. It was conducted at Wenzhou Medical University (Wenzhou, China) between August 2016 and April 2019. Inclusion criteria for CBP were similar to the discovery data set except for the minimum length of pain duration being 3 months, and the presence of a lumbar disc pathology confirmed by a radiologist. Age and sex matched healthy controls (CON) were also included for baseline comparisons. Here we included baseline data only, which was collected prior to the intervention, and included 87 CBP and 96 healthy controls (CON).

2.1.3. Subacute back pain data set (USA SBPp, n=36):

SBP data was acquired as part of a longitudinal observational study, conducted at Northwestern University (Chicago, USA), where individuals with SBP were followed during 3 years [11; 35; 54; 87]. All participants were diagnosed by a clinician for back pain and reported pain intensity greater than 40 on a 0-100 visual analogue scale, pain duration between 4–16 weeks, but no back pain during the previous 12 months. Participants were free of other neurological, chronic pain, systemic disease, head injury or psychiatric comorbidities. Here we only included those participants who, one or three years later, were found to have persistent pain levels (SBPp), based on the criterion of having less than 20% pain reduction from baseline levels of back pain.

2.2. Measures

2.2.1. MRI acquisition

2.2.1.1. Discovery data set (USA CBP)

Data were acquired on a clinical 3T Siemens Magnetom Prisma whole body scanner equipped with a receive-only 64 channel head/neck coil.

T1-weighted MRI acquisition:

High-resolution T1-weighted brain images were collected using integrated parallel imaging techniques (PAT; GRAPPA). The acquisition parameters were: voxel size: 1 × 1 × 1 mm³, TR = 2.3 s, TE = 2.40 ms, TI = 900 ms, flip angle = 9°, 176 sagittal slices, acceleration factor = 2, and field of view = 256 mm. Phase encoding direction was anterior to posterior, and the duration of acquisition was ~5 min.

2.2.1.2. Validation data set (China CBP and CON)

Participants were scanned on a 3T GE-Discovery 750 whole body scanner, equipped with a receive-only 8 channel head/neck coil.

T1-weighted MRI acquisition:

High-resolution T1-weighted brain images were acquired with the following parameters: voxel size = 1 x 1 x 1 mm³, TR =7.7 s, TE = 3.4 ms, TI = 450 ms, flip angle=12°, 176 sagittal slices, acceleration factor=2, and field of view = 256 mm. Duration of acquisition was ~5 min.

2.2.1.3. Subacute back pain data set (USA SBPp):

Brain images were acquired with a 3T Siemens Trio whole-body scanner, with echo-planar imaging capability using the standard radio-frequency head coil.

T1-weighted MRI acquisition:

High-resolution T1-weighted brain images were acquired using the following parameters: voxel size = 1 x 1 x 1 mm³, TR = 2.50 ms, TE = 3.36 ms, flip angle = 9°, 160 slices, field of view = 256 mm.

2.2.2. Image processing

T1-weighted images:

Individual T1-weighted structural images were visually inspected for motion artifacts. Images presenting motion artifacts were not included in the analysis (n=10). Subcortical shape analysis was conducted using automated procedures as part of FMRIB’s Software Library (FSL) version 5.0.9. T1-weighted brain images were subjected to subcortical segmentation using FMRIB's integrate registration and segmentation tool (FSL FIRST). The tool automatically segments 15 subcortical regions as parametrized surface meshes based on a training set of n=336 manually traced brain images [67]. More specifically, the method consists of the following steps, as described in the FSL FIRST user guide [fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST]. Manual labels are parameterized as surface meshes and modelled as a point distribution model. Deformable surfaces are then used to automatically parameterize volumetric labels in terms of meshes; where deformable surfaces are constrained to preserve vertex correspondence for the entire training data. Additionally, normalized intensities on the surface are sampled and modelled. The shape and appearance of the model is based on multivariate Gaussian assumptions where shape is expressed as a mean with modes of variation. Based on the learned models, FIRST searches through linear combinations of shape modes of variation for the most probable shape instance given the observed intensities in T1 images.

All segmentations were manually inspectedfor accuracy before extracting the meshes for each subject.

For each subcortical region, vertex locations for each participant were projected onto the group average surface in Montreal Neurological Institute (MNI) space to obtain vertex-wise displacement values for each subject and used in statistical testing.

2.3. Statistical analyses

T1-weighted images:

The T1-weighted images were used for group level analysis. Vertex-wise displacement value group contrast was carried out using FSL's randomise tool for non-parametric permutation inference to test for differences between sexes in the discovery group. Threshold-free cluster enhancement (TFCE) results corrected for repeated measures at a family-wise error rate of P≤0.05 were considered significant [90]. Ventricular volume, calculated with SIENAX [83], was used as a covariate of no interest. Age also was used as a covariate of no interest in confirmatory post-hoc tests.

For each of the identified clusters displaying significant shape differences between males and females with CBP within the discovery data set (USA CBP), mean vertex-wise scalar values were extracted for each participant. The resultant values represent the mean inward (negative) or outward (positive) deformations from the group average shape, in millimeters.

To test whether findings were reproducible, as well as their specificity to chronic pain, similar analysis steps were performed on the Chinese (China CBP and CON) and SBPp (USA SBPp) datasets, except that this time only the vertex-wise displacement values were extracted from the clusters identified in the discovery phase (USA CBP).

Mann Whitney U tests were used to evaluate structural differences between males and females across cohorts. Mixed-design ANOVA was used to evaluate time by sex effects in hippocampal shape in the USA SBPp cohort. Ordinary least squares regression was used to control for covariates of no-interest. All statistical tests were two-sided. Post-hoc tests were performed in python 3.0, using statsmodels v0.9.0 and scipy v1.2.1, as well as Matlab R2016a.

3. RESULTS

3.1. Participant characteristics

Demographic, pain, and mood characteristics were balanced between males and females across data sets: USA CBP, China CBP, and USA SBPp (Table 1).

Table 1.

Demographics and pain characteristics

	USA CBP		China CBP		China CON		USA SBPp
	Female	Male	Female	Male	Female	Male	Female	Male
N	77	78	29	58	53	43	18	18
Age (y), mean ± SE	50.7 ± 1.7	51.8 ± 1.3	46.3 ± 2.1	44.1 ± 1.6	45.6 ± 1.8	47.9 ± 2.0	41.2 ± 1.9	44.8 ± 2.4
Pain duration, mean ± SE	6.0 ± 0.8 y	6.1 ± 0.6 y	4.1 ± 0.8 y	3.0 ± 0.5 y	n/a	n/a	11.1 ± 1.3 wk	8.8 ± 0.7 wk
NRS (0-10), mean ± SE	5.6 ± 0.3	5.7 ± 0.2	4.7 ± 0.3	5.2 ± 0.3	n/a	n/a	7.0 ± 0.4	6.5 ± 0.4
BDI mean ± SE	7.0 ± 0.6	7.8 ± 0.7	8.4 ± 1.3	8.5 ± 1.0	4.2 ± 0.8	5.6 ± 1	6.1 ± 1.1	6.6 ± 1.0
PASS mean ± SE	*36.3 ± 4.3	*34.5 ± 3.6	32.4 ± 3.0	29.6 ± 2.2	NA	NA	NA	NA

CBP, chronic back pain; CON, healthy subjects; SBPp, subacute back pain patients who persist with back pain one year later; SE, standard error of the mean; y, years; wk, weeks; NRS, numerical rating scale; BDI, Beck Depression Inventory II (scale 0-63); PASS, Pain Anxiety Symptoms Scale (0-100); NA, not available.

^*For this dataset, PASS scores were only available to phase I participants of the study. USA, data collected in the USA at Northwestern University; China, data collected in China at Wenzhou Medical University.

3.2. Left anterior hippocampus is sex-dimorphic in CBP

Vertex-wise shape deformation was first tested in right and left hippocampus, searching for a sex difference in the USA CBP data set. Secondarily, searching for additional sex dependent shape deformations in the USA CBP data set, we also bilaterally examined six other subcortical nuclei: thalamus, caudate, putamen, accumbens, amygdala, and globus pallidus. Only three regions were identified displaying significant shape differences between men and women with CBP, within this discovery data set: bilateral medial thalamus and the left ventral anterior hippocampus (alHP) (Figure 1, top row). Post-hoc analysis of mean regional shape deformation indicated that women with CBP displayed outward shape deformation within the three regions, relative to males with CBP (Figure 1, bottom row, Table 2). These group differences were still present after controlling for age via ordinary least squares regression.

Figure 1. Sex differences in subcortical shape in 155 USA CBP (discovery data set).

USA CBP females presented outward shape deformation in the left anterior ventral hippocampus (A), left thalamus (B) and right thalamus (C), relative to male CBP patients. The 3D renderings (top row) depict, in orange, clusters with vertex-wise displacement differences between USA CBP men and women, ventricular volume was included as a covariate of no-interest (non-parametric permutation test, with threshold-free cluster enhancement, TFCE, corrected for multiple comparison at a family-wise error, FWE, rate of P≤0.05). Extracted mean vertex displacement values (bottom row) from these regions, post-hoc analysis, confirms outward shape deformation in CBP women in comparison to men in the left hippocampus (u-stat = 4102, p<0.001), left thalamus (u-stat = 4354, p<0.001), and the right thalamus (u-stat = 4513, p<0.001). Group differences remained statistically significant after controlling for ventricular volume and age, using least squares regression and Mann-Whitney U test. Box plots show median, quartiles, and ranges; while numerals indicate number of subjects. CBP, chronic back pain; A, anterior; P, posterior; R, right; L, left. Numbers of subjects indicated for females and males, on each plot.

Table 2.

Mean vertex displacement values from significant clusters, USA CBP discovery data set.

		Left hippocampus	Left thalamus	Right thalamus
USA CBP	Female, mean ± SE (n=77)	0.33 ± 0.13 mm	0.25 ± 0.06 mm	0.26 ± 0.05 mm
	Male, mean ± SE (n=78)	−0.33 ± 0.11 mm	−0.24 ± 0.08 mm	−0.24 ± 0.07 mm

Mean vertex displacement values from significant clusters, USA CBP discovery data set. SE, standard error of the mean, CBP, chronic back pain. Positive values are outward deformations; negative values are inward deformations.

Visually, figure 1 post-hoc distributions hint that the alHP statistical differences between men and women with CBP may be due to outliers (suggested by a reviewer). We explored this notion by performing a formal outlier analysis: First we set an outlier threshold (1, 2, or 3 standard deviations (STD) from group mean), exclude alHP values above and below set threshold in both groups, and then perform Mann-Whitney test (MW test). We obtain the following results: For threshold set at 1 STD, included women and men were n=55 vs n=57, MW test statistic=2566.0, p-value=6.3x10⁻⁹; at 2 STD n=72 vs n=73, MW test statistic=3713.0, p-value=1.8 x10⁻⁵ ; at 3 STD n=76 vs n=77, MW test statistic=4025.0, p-value=6.1x10⁻⁵. This outlier analysis demonstrates that the sex-dependent alHP differences are not due to outliers.

Validity of findings from the discovery group was tested by verifying reproducibility of obtained results in the Chinese data set. In the latter cohort, we extracted mean shape displacement values only from vertices found to be significantly different between sexes in the USA CBP data set: identified vertices in bilateral thalamus and alHP. Out of the three CBP USA dimorphic regions, both left hemispherical findings were reproduced (Figure 2). After controlling for ventricular volume, the right thalamus finding was also reproduced in individuals with CBP (u-stat=1108; P=0.02). Additionally, given the availability of healthy subjects within this data set, we also tested whether results were specific to CBP or shared with CON. We found that the alHP shape difference between sexes was only present in CBP and not in healthy controls, while thalamic differences were also present in healthy individuals (Figure 2). However, after controlling for ventricular volume this right thalamic finding did not survive (P=0.23).

Figure 2. Validating sex-dependent shape displacements in Chinese CBP and control (CON) participants.

The validity of findings from the discovery group was tested in a data set collected at Wenzhou University in China, by computing the mean shape displacement only from vertices identified in the discovery cohort. Both the left hippocampus (A) and left thalamus (B) findings were reproduced, showing outward shape formation in females with CBP in comparison to males with CBP (u-stat=1066, P=0.04 and u-stat=1070, P=0.04, respectively, with only a trend being observed for the right thalamus (C). From the three regions, only the left hippocampal shape was not sex-dimorphic in healthy controls, implicating specificity to CBP. These findings remained statistically significant after controlling for ventricular volume and age, except for the right thalamus in healthy subjects which became non-significant. After controlling for these confounds, in CBP the right thalamus differences between sexes became statistically significant, with females presenting outward formation (P=0.02). However, relative to healthy females, there was no indication for CBP-related female outward formation in this area (u-stat=899, P=0.27). Thus, while the three shape regions consistently show outward formation in females with CBP, the left anterior hippocampal shape is the only region with indication of specificity to CBP. Positive values represent outward deformations; negative values are inward deformations. Mann-Whitney U tests were used. Box plots show median, quartiles, and ranges; while numerals indicate number of subjects. CBP, chronic back pain; CON, healthy subjects. Numbers of subjects are indicated for females and males, on each plot.

When testing for region (alHP, left thalamus, right thalamus) by sex (male, female) by condition (cbp, healthy) interaction, there was not a significant effect (F(7,540)=0.98, P=0.45, ANOVA), but only a strong main effect for sex was observed (F(1,540)=20.2, P<0.0001). Consistent with a lack of interaction we observe that alHP shape deformation presented a tendency for difference between females with CBP and healthy females (u-stat=955, P=0.1), while there was no indication for a difference in mean shape deformation between male CBP and male CON (u-stat=1287, P=0.79). Comparing healthy female shape displacement in the right thalamus area to that in females with CBP, we found no indication for a difference (u-stat=899, P=0.27). Similarly left thalamic findings were not specific to CBP. Only the alHP shape presented specificity to CBP (Figure 2 and Table 3).

Table 3.

Mean vertex displacement for clusters, derived from the discovery analysis, tested in the Chinese validation data set.

		Left hippocampus	Left thalamus	Right thalamus
China CBP	Female, mean ± SE (n=29)	0.30 ± 0.13#+*	0.10 ± 0.06	0.09 ± 0.07
	Male, mean ± SE (n=58)	−0.07 ± 0.10	−0.07 ± 0.05	−0.07 ± 0.06
China CON	Female, mean ± SE (n=53)	0.06 ± 0.07	0.10 ± 0.04+	0.11 ± 0.05+
	Male, mean ± SE (n=43)	−0.08 ± 0.08	−0.12 ± 0.06§	−0.14 ± 0.07§

SE, standard error of the mean; CBP, chronic back pain; CON, healthy controls. Positive values are outward deformations; negative values are inward deformations.

^#P<0.05, difference compared to healthy participants

⁺P<0.05, difference compared to male healthy participants

^*P=<0.1, difference relative to healthy females

^§P=<0.05, difference relative to healthy females; Mann Whitney U-test.

We again explore outlier dependence of the group difference between sexes for alHP shape distortion in China CBP data. We use the same procedure as for the USA CBP data. At an outlier threshold of 1 STD, we retain n=22 and n=40 female and male CBP, resulting in a MW test statistic=623.0, p-value=0.007; at 2 STD, we retain n=27 and n=55 female and male CBP, with a MW test statistic=923.0, p-value=0.07; while at 3 STD, n=22 and n=40 male and female CBP were retained, with a MW test statistic=623.0, p-value=0.007. Here too we observe group differences in China CBP male and female alHP shape distortion is not driven by outliers.

Thus, these results validate the main observation that left anterior hippocampus shape deformation is distinct between female and male CBP, it also suggests that this difference is a reflection of shape deformation in females CBP in relation to healthy female subjects. Left anterior hippocampus mean shape displacement associations with mood, anxiety, pain intensity and pain duration were not observed in the discovery CBP group, when tested for female-only, male-only, and whole-group.

3.3. Left anterior hippocampus as a putative sex-dependent diagnostic biomarker for chronic back pain

In trying to differentiate the hippocampus finding between sex-dependent prognostic versus diagnostic marker for CBP [70], we tested whether the shape difference would be also present when comparing males and females with early onset back pain who, one year later, presented persisting back pain (SBPp). We did not find a significant difference in shape between sex groups in this cohort at baseline, but over time differences do emerge (ranksum = 375, P=0.19, at one year; ranksum = 119, P=0.07, at three years; Mann-Whitney U test) (Figure 3A, Table 4). Group mean shape displacement trajectories over time show how the differences become stronger over time (Figure 3B). We observe that the mean outward displacement in women doubles at each measurement time (from entry to 1 year and then at 3 years), and the alHP displacement matches that seen in the discovery (USA CBP) and china validation (China CBP) measures, when the USA SBPp patients are examined at 3 years after onset of back pain. When examining only the sub-sample of participants (n = 20) who could be followed during the entire three-year period, we found a significant sex effect (F(1,18) = 4.58, P<0.05, mixed model ANOVA), with no significant time by sex effect (F(2,36)=1.25, P=0.30). Although the number of subjects who could be followed during the entire 3-year period is small, the observed patterns both resemble that seen in participants with CBP, and reinforce the idea that alHP shape deformation is a sex-dependent consequence of persisting pain.

Figure 3. Evolution of left anterior hippocampus shape deformation over time, in individuals with early onset back pain with persisting back pain over 1-3 years.

(A) In early onset, no sex-dependent shape distortions were seen in SBPp (ranksum = 370, P=0.25, at baseline), but over time differences start to become visible (ranksum = 375, P=0.19, at one year; ranksum = 119, P=0.07, at three years). (B) Depicts mean shape over time for all available samples (n=36), showing clearly how the differences become stronger over time. (C) Mean shape over time only for the sub-sample of participants who could be followed during the entire three-year period (n=20). Mixed model ANOVA for time and sex indicated a significant sex effect (F(1,18) = 4.58, P=0.046). Although the number of subjects with data for the whole three-year period is small, the observed patterns both resemble that seen in participants with CBP, and reinforce the idea that these left hippocampal changes in shape are a sex-dependent consequence of persisting pain. Positive values are outward deformations; negative values are inward deformations. Box plots show median, quartiles, and ranges; while numerals indicate number of subjects. Mann-Whitney U-tests were used. Mean shape displacement values were corrected for age via ordinary least squares regression. Line plots’ error bands represent standard error of the mean for each time-point.

Table 4.

Mean vertex displacement of the left anterior hippocampus, alHP cluster over time, in SBPp.

alHP	USA SBPp
	Baseline (n=36)		~1yr (n=36)		~3yr (n=20)
	Female	Male	Female	Male	Female	Male
Mean ± SE (mm)	0.07 ± 0.09	−0.07 ± 0.10	0.14 ± 0.13	−0.14 ± 0.15	0.28 ± 0.15	−0.23 ± 0.17

Mean values were corrected for age using ordinary least squares regression. Positive values are outward deformations; negative values are inward deformations. SE, standard error of the mean; SBPp, early onset back pain patients, who one year later had persisting back pain; alHP, left anterior hippocampus

3.4. Meta-analytic estimate of magnitude of validation for hippocampal shape deformation

The sex-dependent outward displacement of alHP observed in China CBP and USA SBPp are valid and unbiased estimates of the results discovered in in USA CBP dataset. These findings in the validation datasets closely match that observed in the discovery dataset. Both the mean and standard errors of aiHP in females CBP in USA CBP, China CBP, and USA SBPp at 3 years of pain persistence are nearly exactly the same values (+0.3+/−0.1 mm), while the corresponding values in males are always negative, and in the female controls (China) they are positive but much closer to zero (0.06+/−0.07 mm). Given that USA SBPp dataset is derived from a longitudinal study and the results at 3 timepoints are from observations collected years apart, all 3 observations can be viewed as independent instances for validating the main hypothesis. Therefore, we can state that we have in fact 4 instances where the validity of outward displacement of alHP in females with backpain was tested. We formalize the significance of reproducibility of obtained results by calculating Fisher’s combined probability test, a meta-analytic probability measure for independently measured corresponding outcomes: Female vs. male back pain China CBP x USA SBPp data at time 0 x at 1 year x at 3 year = 0.04 x 0.25 x 0.19 x 0.07 = 0.00013. The latter statistic estimates overall probability of validation. If we remove the early onset SBP data (time 0) from the meta-analysis then we get 0.04 x 0.19x 0.07 = 0.000532, which still indicates that the probability of the observed result being a consequence of chance is very low.

3.5. Functional networks and associated terms related to the deformed hippocampal region

To uncover potential functions of the identified hippocampus subregions, we used a reverse inference term-based meta-analytic approach from Neurosynth [91]. We chose this approach as it is a more objective procedure for linking anatomy to brain function than the more commonly used practice of simply searching the literature regarding functions associated with the brain region of interest. First, functional connectivity maps from 3 mm radius spheres in the regions of interest (Figure 4a) revealed functional connectivity between deformed portion subiculum (Figure 4b, coordinates: −22, −12, −28), restrospinal/precuneus cortex, and right hippocampus (Figure 4c). For the alHP region within entorhinal cortex (Figure 4b, coordinates: −16, −8, −24), additional connectivity was seen with the medial prefrontal cortex (Figure 4d). According to similarity between these functional connectivity maps and the Neurosynth whole-brain reverse-inference maps for each term, terms most strongly associated with these maps (association test) were: “autobiographical, episodic memory, default, and memories” for the subiculum network, and “autobiographical, emotional, memories, and emotion” for the entorhinal network.

Figure 4. Neurosynth resting-state functional connectivity for seed regions centered in the identified sex-related shape distortion area in the anterior left hippocampus, alHP, in a sample of 1000 participants, and associated non-anatomical terms.

(A) High resolution image displays, in orange, the shape dimorphic area in the left hippocampus (represented in blue). (B) Locations of seeds used to identify functional connectivity of the distorted alHP region with the rest of the brain. A 3 mm radius sphere centered in the affected subiculum area (MNI152 coordinates: −22, −12, −28, displayed in green) and the entorhinal area (MNI152 coordinates: −16, −8, −24, depicted in pink). (C) Functional connectivity map for the subiculum area shows activity synchronization with bilateral retrosplenial cortex and right hippocampus (not displayed), when connectivity is thresholded to r>0.2. Similarity of this connectivity map with the Neurosynth database indicated highest similarity with brain maps encoding terms such as “autobiographical” and “episodic memory”. (D) Functional connectivity map for the affected border with entorhinal cortex. Additional connectivity with medial prefrontal cortex is seen. Terms associated with this map show a stronger association with “emotional”. World cloud (top 25 non-anatomical terms of a total of 1335 terms) sizes of words correspond to the Pearson correlation coefficient between the whole-brain (uncorrected, based on n=1000 subject maps) reverse-inference map for each term, and the functional connectivity maps (0.14 < r < 0.41; r, Pearson correlation coefficient).

4. DISCUSSON

Our primary results are the following: 1) We discover in CBP with mean duration of back pain of 6 years, that females in contrast to males (USA CBP) exhibit an outward deformation of the left anterior hippocampus (alHP), a region spanning the subiculum and entorhinal cortex. 2) In a new cohort (China CBP and CON) we test whether the specific regional findings from the discovery cohort are reproduced, and observe again outward deformation of alHP in female CBP in contrast to male CBP; when female CBP are contrasted to female controls we only observe a trend for larger deformation in female CBP. Importantly the mean duration of back pain in the China CBP cohort was around 3-4 years. 3) In a third and longitudinal cohort of early-onset back pain persisting up to 3 years (USA SBPp), examining only the alHP region, and comparing between female SBPp and male SBPp over time, we observe outward deformation in the females, in contrast to males, that becomes more prominent over time, reaching statistical significance only with repeated-measures analysis. 4) To objectively identify the functional properties of alHP, we use meta-analytic tools to uncover concepts associated with alHP circuitry. Therefore, in a multi-site study design with relatively large cohorts, we were able to detect hippocampus morphological differences between males and females with CBP.

Due to hippocampal functional heterogeneity and anatomical complexity, we focused on examining its shape, as the approach can reveal specific sub-regional subtleties which may otherwise be overlooked in volumetric analysis approaches. The robust methodology employed provides adequate evidence for the left anterior hippocampuslaHPalHP shape to be considered as aa potential biomarker for CBP chronic back pain in women. The finding was mostly generalizable across studies, ethnicities, image acquisition parameters and hardware. It is alsoseems specific to the chronic back pain state, as it was not minimally present in healthy participants nor and in early onset back pain, and appears to be developing in time with the transition from acute or early onset back pain to persistence of painto chronic back pain over 1-3 years. Although the longitudinal analysis suggests that the hippocampal deformation is outward in females and inward in males, in the larger discovery and validation cohorts the most robust result was the outward deformation in females with CBP, and thus this is considered the most consistent result.Our Neurosynth-based analyses indicated functional connectivity of the deformed regionslaHPalHP with default-mode network hubs (based on n=1000 resting state scans), namely the precuneus/retrospinal cortex and the medial prefrontal cortex (mPFC), and . The identified connections were meta-analytically related (based on >12,000 task-based studies) to autobiography, emotions, and memories, implyingsuggesting a biologically grounded distinction between females and males with CBPchronic back pain in emotional and cognitive regulation of mnemonic processes. Note that meta-analytic results derive laHPalHP properties in non-specifichealthy individualssubjects;, how these properties become distorted in chronic pain remains to be studied.

Previously we proposed a four-stage mechanistic concept for the transition to chronic pain [5], which gives rise to stage specific biomarker types: prognostic, nociceptive, preventive and diagnostic [70]. To determine its biomarker classification, we examined alHP shape deformation in early onset back pain subjects, and found that the sex differences were becoming more prominent in time. Therefore, the identified hippocampal shape deformation seems to emerge in women over time during transition and with persistence of pain, and as such we classify it as a sex dependent potentially diagnostic biomarker for chronic pain.

Subcortical volume changes have been examined in multiple pain conditions, in humans and rodent models of persistent pain [14; 22; 25; 26; 39; 42; 44; 53; 56; 65; 68; 88]. Most, but not all, identify decreased regional volumes, yet the majority of these studies are based on small samples and without reproducibility testing, and thus their generalizability remains uncertain. The two largest cohort studies (>1000 subjects each), examined musculoskeletal/joint pain [22] and headache pain [39] influence on brain anatomy, and conclude disparate results. Influence of headache was hypothesized to lead to decreased accumbens volume, which could not be confirmed [39]. Only one study has used a validation approach [53], where accumbens low volume was shown as a predictor for chronic pain, expanding on and complementing our earlier similar observation [11]. The authors validated this observation using some of the same data presented here (available in OpenPain.org). Here we examined the potential sex-dependent shape deformation of accumbens in CBP, from the viewpoint of dopaminergic control of chronic pain (striatal dopamine release in healthy subjects is 50% lower in females, [63]). We could not identify such an effect, but this was at least in part due to high measurement variability as a consequence of the contamination of accumbens borders with adjacent ventricles. The study most directly related to the current analysis regards the examination of brain regional volumes in 3892 community dwelling individuals with chronic joint pain [22]. The primary observation of the study was decreased hippocampal volume in females only, after correcting for age, depression and intracranial volume. Our results extend this study, revealing the specific hippocampal regional deformation in females with CBP. Importantly, the opposite effect of pain relief on hippocampal volume has been recently reported too: when chronic patients undergo significant recovery from the condition they exhibit increased hippocampal volume [31], similar results are also seen in cortical grey matter density recovery post successful knee surgery in osteoarthritis [75] and post-back surgery in back pain [80]. Decreased hippocampal volume, or regional deformation, with persistent pain coupled with volume restoration with pain relief is consistent with the notion that the stress of the pain state, as well as the effect of relief from pain, must be coupled with synaptic reorganization processes uniquely in females, and complement our observation that alHP shape is most likely a diagnostic biomarker.

It is suggested that chronic pain can be viewed as an anhedonic central nervous system stress state or response [16; 18; 23; 24]. Within this viewpoint there is good evidence that men and women differ in their susceptibility and prevalence for stress-related illnesses. Women exhibit higher prevalence in disorders like anxiety and depression, which seems to emerge with adolescence, and also exhibit higher prevalence in periods of drastic hormonal perturbations, like puberty, pregnancy, and postpartum [1; 2; 17; 19; 43; 59; 82]. Whether hormonal fluctuations influence rates or prevalence of chronic pain remains unclear. However, negative mood disorders such as major depression (MDD) [15; 45; 76; 77] and rodent models of chronic pain (see introduction) seem both linked to altered structure, function and neurogenesis processes within the hippocampus, and associated with decreased hippocampal volume [52; 79; 89]. Thus, the current results provide human hippocampal specific regional structural deformation for chronic pain and the observation is consistent with the concept of persistent pain imposing a stress response within the hippocampus. Despite the apparent similarities between MDD and chronic pain, we expect structural deformation for MDD to be distinct from that seen for chronic pain, and the location of deformation may even differ between types of chronic pain, implying that alHP location may be unique to CBP. It is also noteworthy that sex specificity of hippocampal structure changes in MDD remain unconfirmed although expected. With regards to sex/gender differences in the healthy human grey matter structure and white matter connectivity, there is convincing evidence that these properties are highly overlapping between sexes, and the differences observed are variable and inconsistent [41].

The fact that the observed sex differences in shape deformation were solely left hemispherical, is not surprising, as the hippocampus is known to present laterality specialization. While the left hippocampus is associated with verbal memory processes, the right is more linked with spatially-dependent memories [74]. Studies of transient pain and associated memory report preferential left hippocampal activity [27]. Exploring functional connectivity laterality of the identified cluster in Neurosynth, we qualitatively observed less connectivity from the right hemisphere with default mode network regions. Even in MDD associated volume decreases are observed more prominently in the left hippocampus [52; 81; 92].

Estrogens and its endogenous specimen 17β-estradiol (E2) are important sexual hormones. There is compelling evidence that the hippocampus can synthesize E2 [78; 84], which suppresses hippocampal inhibitory synapses only in females [37], and induces hippocampal activity-dependent potentiation of excitatory synapses with distinct molecular pathways employed in males and females [40]. Recent evidence also shows that glutamatergic ventral tegmental area inputs to the hippocampus provide a valence signal to episodic memories, and its synaptic efficacy is stronger in females [34]. Thus, the differential E2 and glutamatergic responses in females may be prime mechanistic routes underlying the alHP deformations that we observe in female CBP, and potentially provide molecular targets for the development of novel sex-specific therapies for chronic pain.

The present study adheres to a strict, unbiased approach in studying subcortical sex-dependent shape deformation. It demonstrates that, even when the discovery results show strong statistical differences, secondary validation studies can exhibit weak or borderline statistical significances. The latter in part depends on specific, often subtle, deviations in cohort properties (that require future studies). Moreover, such statistical variations are expected based on probability theory, even when effects are truly present [93-96]. Importantly, the entirety of our results shows a consistent pattern. Our finding evidences the dimorphic adaptations of the hippocampal circuitry with persistent pain, a potential diagnostic biomarker. Because this hippocampal circuitry interacts with the default mode network and is engaged in the emotional and cognitive regulation of mnemonic processes, the results complement and advance previous work on the importance of the limbic system in chronic pain maintenance and thus has implications for treatment.