Background
Lymphoma refers to a type of cancer derived from lymphatic system. Based on the types of involved lymphocyte cells, lymphoma can be roughly divided into two main types: Hodgkin’s lymphoma (HL) and non-Hodgkin’s lymphoma (NHL) [
1]. NHL accounted for approximately 2.82% of 19.3 million new cancer cases and 2.63% of 9.9 million cancer deaths worldwide in 2020 [
2]. As a phenotypically and genetically heterogeneous disorder, diffuse large B-cell lymphoma (DLBCL) was one of the most common subtypes of NHL [
3].
Owing to the advances in diagnosis and therapy, survival or remission rates for lymphoma have improved prominently. Apart from the lymphoma- and chemotherapy-related somatic symptom burden, increasing attention has been drawn to the health-related quality of life [
4,
5]. In particular, the presence of cognitive impairment, negative psychological status (e.g., anxiety, depression and committing suicide) or other emotional symptoms among patients with hematological cancer was supposed to be more likely than those with solid tumor [
6‐
9]. Furthermore, psychological complications were common among NHL patients, especially for the more aggressive subtypes [
10]. Though these brain dysfunctions were assumed to be mainly associated with chemotherapy neurotoxicity, the disease status of non-central nervous system (non-CNS) cancer itself was also considered to be relevant [
11,
12]. Knowledge about the underlying mechanisms for brain dysfunction induced by non-CNS lymphoma itself, would further facilitate the better assessment of exclusive detrimental effect related to chemotherapy or other treatments and, even more importantly, allow for developing tailored therapeutic regimen to alleviate not only the physiological, but also the psychological symptoms.
The application of
18F-fluorodeoxyglucose positron emission tomography-computed tomography (
18F-FDG PET/CT) has been routinely recommended for the staging and response assessment of FDG-avid lymphoma, especially DLBCL [
13,
14]. Besides,
18F-FDG PET/CT imaging could also be applied for the noninvasive appraisal of altered brain functions from the perspective of glucose metabolism. An increasing number of
18F-FDG PET studies revealed that chemotherapy would lead to regional brain glucose metabolic abnormalities among lymphoma patients [
15‐
18]. However, up till now, only a few researches have investigated the brain metabolic impairments in patients with pre-treatment lymphoma [
19‐
21].
Compared with the normal control, lymphoma patients prior to the initiation of chemotherapy showed regional hypometabolism primarily in the parieto-occipital lobes [
20]. However, it is still up for debate whether or not these cerebral metabolic abnormalities related to the disease status of lymphoma will change after receiving chemotherapy [
19,
21]. Moreover, higher FDG uptake in basal ganglia and thalamus or lower FDG uptake in cerebellum relative to liver based on pre-treatment
18F-FDG PET data was supposed to have the potential for predicting a worse prognosis in patients with DLBCL or extranodal natural killer/T-cell lymphoma [
22,
23]. All of these brain metabolic parameters were derived from univariate methods accounting for only the intensity or magnitude. To the best of our knowledge, lymphoma-related pattern of metabolic brain network has not been validated.
The primary aim of this study was to establish and validate the DLBCL-related pattern (DLBCLRP) of metabolic brain network. In order to analyze discrepancies in the expressions of lymphoma-related pattern between DLBCL and HL, the expressions of DLBCLRP in HL patients were also investigated. Moreover, the correlations between DLBCLRP and several indexes of the staging and response assessment were further explored.
Discussion
The current study provided a fresh perspective for the altered metabolic brain network in DLBCL. The established pattern of brain network related to DLBCL could be verified through validation group and was completely different from that of HL. This DLBCLRP was characterized by the increased metabolic activity in brain regions rarely involving cerebrum. The expression of DLBCLRP was associated with the tumor burden of lymphoma, implying a potential biomarker for staging and prognosis. These findings about brain dysfunction induced by non-CNS lymphoma itself, would also facilitate the further exploration of exclusive detrimental effect related to chemotherapy or other treatments.
It was suggested that not only chemotherapy, but also cancer itself could disrupt the functional connectivity measured with resting- or task-state functional magnetic resonance imaging (fMRI) [
11,
32,
33]. For the patients with breast cancer before receiving any treatment, the lower whole-brain functional connectivity, the greater decline in cognitive function [
33]. These aberrant functional connectivities were also considered to be relevant to the memory impairment and fatigue [
11,
32]. The presence of neuropsychiatric symptoms among patients with hematological cancer was supposed to be more likely than those with solid tumor [
6‐
9]. Thus, it is necessary and critical to investigate the underlying mechanisms for brain dysfunction induced by non-CNS lymphoma itself. However, evidence for the association between brain function and metabolic network in patients with extracranial DLBCL is still lacking. Only a few studies have explored the lymphoma-related metabolic features, by using univariate methods accounting for merely the intensity or magnitude of
18F-FDG PET data [
20,
34‐
36]. The glucose metabolic levels of bilateral frontal, temporal and parietal lobes were negatively correlated with the lymphoma burden [
20,
34]. But the discrepancy in metabolic features between different types of lymphoma was not further discussed. Hypometabolism might be localized prominently at bilateral occipital lobes for the young HL patients [
35]. DLBCL with immune effector cell-associated neurotoxicity syndrome after receiving chimeric antigen receptor T-cell therapy showed a diffuse brain hypometabolism [
36]. Moreover, in T-cell lymphoma patient, global hypometabolism and its associated neuropsychiatric symptoms can be reversed by administrating glucose [
37]. To better understand the alteration of metabolic brain network in DLBCL, multivariate decomposition technique with SSM-PCA was introduced to establish DLBCLRP in this study.
The verified pre-treatment DLBCLRP was characterized by the relatively increased metabolic activity in bilateral cerebellum, brainstem, thalamus, striatum, hippocampus, amygdala, parahippocampal gyrus and right middle temporal gyrus. Put another way, hardly any cerebrum was involved in this compensatory increase in metabolic network. Based on the parameters of brain SUV
max, SUV
mean and TBG, however, the FDG uptake level of whole brain in DLBCLRP patients was reduced compared with that of the control [
34]. This metabolic decline might be attributed to the competition from the tumor tissue with huge energy demand [
21]. We speculated that the brain basic functions (e.g., heartbeat and breathing) would be reserved as priority to cope with this relative energy shortage. Accordingly, DLBCLRP demonstrates compensatory and relatively elevated glucose consumption almost exclusively located at lower-level brain structures, including cerebellum, brainstem and diencephalon.
Accounting for two-thirds of the total brain volume, the cerebrum should be responsible for processing a series of complex brain functions, such as emotion, language, cognition and so on [
38‐
40]. Principal component analysis in the current study also revealed that metabolic activity remained relative stable in the frontal and temporal lobes, but declined in the parietal and occipital lobes. Thus, the latter brain regions involving relatively less significant functions were considered more likely to be sacrificed in the remodeling of metabolic brain pattern for DLBCLRP. The occipital lobe was considered as the center for visual processing, while the parietal lobe, the center for sensation information processing related to cognition and speech. Besides, the anterior cingulate gyrus, midcingulate cortex and the adjacent medial frontal cortex should be responsible for regulating the emotion, memory storage and behavior. Consequently, these dysfunctional brain regions might be involved in the formation of DLBCLRP-associated neuropsychiatric symptoms, including visual hallucinations, depression, anxiety, cognitive and speech dysfunctions, fatigue, etc. [
4,
10,
37,
41].
To determine whether there is a difference in the patterns of metabolic brain network between DLBCL and HL, the expressions of DLBCLRP in HL patients were also investigated. We found that the established DLBCLRP could not be applied to the patients with HL. Though HL group was significantly younger than DLBCL group in our study, no significant difference in the pattern expression was found between HL patients < 40 yrs and those ≥ 40 yrs. Therefore, the discrepancy in the patterns of metabolic brain network between two types of lymphoma should not be attributed to the disparity in their age of onset [
2]. Though no significant difference in the TMTV was found between DLBCL and HL groups, HL patients had significantly lower SUV
max and TLG than DLBCL patients. Thus, it was speculated that the expression of DLBCLRP might be associated with the tumor burden of lymphoma.
To confirm this assumption, correlation between the expression of DLBCLRP and the parameters of tumor burden (TMTV and TLG) was further explored. The Z scores of pattern expression were significantly positively correlated with TMTV and tended to be positively correlated with TLG in the form of log transformation. Moreover, our results also suggested that the higher level of whole-brain FDG uptake was associated with the lower pattern expression, TMTV and TLG. In consideration of the competition for capturing glucose between the tumor and the brain tissue, the tumor burden of lymphoma might indirectly affect the expression of DLBCLRP via whittling down the whole-brain “input” of glucose/FDG. Then, the metabolic brain network would undergo remodeling to accommodate the reduced energy supply as discussed above. Moreover, there was an inverse relationship between plasma glucose levels and brain FDG uptake [
42]. Based on this assumption, the declined brain “input” of glucose and its associated DLBCLRP might be reversed by the administration of glucose supplementation [
37].
Considering no significant difference in age and gender between DLBCL and control groups, this DLBCLRP can still be established, but might vary in some extent with age or gender. One possible explanation might be the brain compensatory function declining with age; subsequently, the metabolic features and network of DLBCLRP are prone to be tempted among the elderly [
43,
44]. Both brain function and its glucose metabolism could be altered in a sex-dependent manner [
45]. This feature may be attributed to the sex differences in genes, transcriptomes and hormones [
46]. More detailed investigations are needed to verify the necessity to establish age-specific or gender-specific DLBCLRP. Furthermore, we are very curious to further explore the correlation and disparity between DLBCLRP and senile neurodegenerative diseases-related brain metabolic pattern.
By using the principal component analysis approach, the disease-related metabolic covariance patterns have been explored in a number of neurodegenerative diseases, including Parkinson's disease (PD), Alzheimer's disease (AD), dementia with Lewy bodies (DLB), mild cognitive impairment (MCI), fronto-temporal dementia (FTD), REM sleep behavior disorder (RBD), spinocerebellar ataxia (SCA), amyotrophic lateral sclerosis (ALS) and Parkinson plus syndromes (PP) [
47‐
55]. Among these verified disease-related metabolic brain pattern, the most well-known PD-related pattern (PDRP) was characterized by relative hypermetabolism in pallidothalamic and pontocerebellar regions, accompanied by relative hypometabolism in the premotor and parieto-occipital association regions [
56]. This spatial covariance PDRP exhibited a high degree of repeatability in PD population [
26,
47]. Beyond the differential diagnosis of PD, the expression of PDRP was found to be correlated to motor manifestations and could serve as a biomarker for assessing and monitoring the therapeutic efficacy in PD patients [
57]. However, in addition to the DLBCL-specific changes, a wide range of overlapped brain regions was detected between the canonical PDRP and our proposed DLBCLRP. Other neurodegenerative disorders would exhibit greater variability with DLBCL in the disease-related metabolic pattern [
49,
51‐
53]. Furthermore, though we found that the established DLBCLRP could not be applied in another common type of lymphoma, HL, it is still unable to easily draw such a conclusion that this metabolic brain pattern can serve as a robust metabolic marker for the diagnosis of DLBCL. Further straightforwardly investigating the similarities and differences in metabolic brain pattern, and in pattern-associated determinants between DLBCL and PD (or other neurodegenerative disorders), would be very attractive.
Since the expression of DLBCLRP could be associated with the tumor burden of lymphoma, another implication for the establishment of DLBCLRP was whether it could serve as a potential metabolic brain biomarker for the prognosis of DLBCL. Apart from tumor burden, our findings indicated that the pre-treatment DLBCLRP expression was also associated with response assessment and IPI in patients with DLBCL. In view of the competition for capturing glucose/FDG between the tumor and the brain tissue, we speculated that the tumor burden of lymphoma might indirectly affect the expression of DLBCLRP through undermining the brain glucose supply. Moreover, several PET-based studies recommended that the semi-quantitative parameters of tumor metabolic burden should be independent prognostic factors for the DLBCL [
38‐
40]. Thus, it was reasonable to evaluate the prognostic value of the brain glucose metabolic level and the expression of DLBCLRP in patients prior to any treatment. The SUV ratio of the cerebellum over the liver extracted from baseline PET data might be a predictor of progression-free survival [
23]. Besides, it was suggested that low metabolic volume products in not only cerebellum but also basal ganglia were associated with a significantly poor prognosis [
19]. Another study using voxel-wised analysis, however, indicated that the higher pre-treatment metabolic level in basal ganglia and thalamus might predict a relatively worse outcome in patients with extranodal natural killer/T-cell lymphoma [
22]. These discrepancies can be attributed to the differences in pathological types or data processing methods between these studies. The herein proposed expression of DLBCLRP in the form of Z-transformation provided a novel covariance parameter reflecting the whole-brain metabolic pattern, and more investigations are needed to define its prognostic value and psychological symptoms implication.
Limited evidence showed that the chemotherapy-induced aberrant metabolic network mainly involved the prefrontal cortex and cerebellar areas [
58]. It is still up for debate whether or not cerebral metabolic abnormalities related to the disease status of lymphoma will change after receiving chemotherapy [
19,
21]. In the current study, we found that there was a significant reduction in the DLBCLRP expressions of post-treatment, compared with those of baseline. More importantly, the post-treatment declines of DLBCLRP expression were significantly positively correlated with Ann Arbor staging and IPI as well. As this is an emerging field of research, further implications of these altered brain network should be explored in patients with extracranial DLBCL or other cancer.
Several limitations of the current study should be acknowledged. Due to the retrospective study design, the assessment of cognitive function or neuropsychiatric inventory was absent. Accompanied by the more and more attention to brain function in patients with lymphoma or other cancer, the data for neuropsychiatric symptoms would be routinely collected and applied to explore its correlation with the brain functional/metabolic network. By the same token, the data of progression-free survival and overall survival are not yet available owing to insufficient follow-up. Further long-term follow-up would be essential for the reliable assertion regarding the prognostic value of DLBCLRP. In addition, though our findings revealed that the metabolic brain pattern of HL patients was distinct from that of DLBCL patients, other types of FDG-avid lymphoma (e.g., follicular lymphoma) were not investigated. It is also fascinating to determine whether there is a difference in metabolic brain pattern between lymphoma and other common solid tumors.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.