Determinants of activity of brown adipose tissue in lymphoma patients

The determinants of brown adipose tissue (BAT) activity are not yet known in detail but might serve as future therapeutic targets against obesity and the metabolic syndrome. We analyzed 235 datasets of lymphoma patients with two PET/CT examinations at different time points retrospectively. We assessed the anthropometric characteristics, features related to the metabolic syndrome, thyroid dysfunction, season of the PET/CT examination, weight change, prior cancer history, lymphoma subgroups, disease activity, and specific lymphoma-related therapies, and evaluated their association with BAT activity. We found BAT activity in 12% of all examinations, and the incidence of BAT activity after initially negative examinations was 10%. In multivariate regression analysis, the prevalence of BAT activity was associated with age, body mass index, sex, the season of the examination, diabetes mellitus, arterial hypertension, and medication on the beta-receptors. New BAT activity arose more often in patients without preceding lymphoma-related therapy. No specific medication was associated with BAT activity. In conclusion, this study confirms the potential connection of BAT with the metabolic syndrome. Preceding lymphoma-related therapy might have an inhibitory effect on the recruitment of BAT.


Scientific Reports
| (2020) 10:21802 | https://doi.org/10.1038/s41598-020-78419-7 www.nature.com/scientificreports/ capsinoids, glucocorticoids, bile acid chenodeoxycholic acid and further dietary components like tea catechins and ephedrine affecting BAT activity [17][18][19][20][21][22][23][24][25] . Other substances under discussion include thiazolidinedione, resveratrol, thyroid hormones, the intestinal hormone secretin and cytarabine [26][27][28][29][30] . Additionally, active tumor disease itself might be associated with BAT activity 31 . The rationale behind the search for BAT activating drugs is the hope of detecting potential therapeutic targets to fight obesity and metabolic disorders by increasing the energy consumption and elevating the insulin sensitivity of the treated individuals 1,9,16 . Based on these considerations, the present explorative study aimed to identify the association of BAT activity with patient-and disease-specific parameters in subjects with lymphoma. Lymphoma patients represent a clinical collective with relatively high BAT activity, which possibly facilitates the identification of BAT-activating factors. In this context, we re-evaluated the most familiar potential influencing parameters of BAT activity, as well as lymphoma-related features, in a large patient cohort.

Results
Patients. The final dataset contained 235 patients (mean age 50 ± 19 years, 111 females, 124 males). The BMI was higher in males than in females (26.1 ± 4.8 kg/m 2 versus 24.8 ± 5.4 kg/m 2 , p = 0.005). Age did not differ significantly between both sexes (p = 0.70). Age and BMI showed a weak positive correlation (r = 0.13, p = 0.04). According to the WHO classification, 113 patients suffered from mature B-cell lymphoma, 103 from Hodgkin lymphoma, 11 from other lymphoma or leukemia, and eight from mature T-cell lymphoma ( Table 1 shows the detailed histological entities). The mature T-cell lymphoma group contained only males, resulting in a significantly different distribution of the sexes to the groups Hodgkin lymphoma and mature B-cell lymphoma (p < 0.004 and p = 0.01, respectively). The BMI did not differ between the lymphoma subgroups (p = 0.99). Patients with mature B-cell lymphoma (mean age of 60 ± 16 years) were older than patients with Hodgkin lymphoma, other lymphoma, and mature T-cell lymphoma (p < 0.001, p = 0.004 and p = 0.04, respectively).
Prevalence of brown adipose tissue activity. In the whole cohort with 470 PET examinations, BAT activity was present in 55 (12%) examinations, while the remaining 415 (88%) examinations displayed no BAT activity. The median standardized uptake value (SUV) of BAT activity was 2.1, and we categorized accordingly 30 datasets as low BAT activity and 25 as high BAT activity. Table 2 illustrates the relationship between BAT activity and all parameters in detail. Age, BMI, sex, and the season of PET were associated with BAT activity (p < 0.001, p = 0.002, p = 0.01, and p = 0.04, respectively). BAT activity was significantly higher in younger patients, in females, in leaner patients, and in winter compared with summer and autumn (p = 0.02, and p = 0.02, respectively). No patient treated with beta-blockers displayed BAT activity, resulting in a significant difference to patients without beta-receptor therapy (p < 0.001) and receiving beta-receptor agonists (p = 0.03). Diabetes mellitus and hypertension were negatively associated with BAT activity (p = 0.009, and p = 0.003, respectively). All these conditions were independent parameters associated with BAT activity in multivariate regression. In univariate regression, patients with Hodgkin lymphoma had a significantly higher BAT activity than patients

Hodgkin lymphoma 103
Other lymphoma or leukemia 11 Lymphoblastic lymphoma of T-cell type 4 Acute myelotic leukemia 3 Unspecified acute B-cell lymphatic leukemia 2 Post-transplant lymphoproliferative disorder 1

Castleman disease 1
Mature T-cell lymphoma 8 Unspecified T-cell lymphoma 3 Extranodal natural killer/T-cell lymphoma 1 Peripheral T-cell lymphoma 4 Scientific Reports | (2020) 10:21802 | https://doi.org/10.1038/s41598-020-78419-7 www.nature.com/scientificreports/ with mature B-cell lymphoma (p = 0.004), and chemotherapy with ABVD was associated with a higher prevalence of BAT activity (p = 0.01). Both conditions were no significant factors in multivariate regression. The other assessed features were not associated significantly with BAT activity (details see Table 2). No patient in our cohort received thiazolidinediones.
Incidence of brown adipose tissue activity. A total of 214 patients had one BAT negative PET/CT examination at initial diagnosis or during disease and a subsequent PET/CT examination in the further course, e.g., for therapy evaluation or in suspected relapse and hence were included in this analysis. Incidence of "new" BAT activity in the second examination occurred in 21 cases (10%), of whom 13 patients developed low BAT activity and eight patients high BAT activity. A total of 193 patients remained BAT negative in the follow-up www.nature.com/scientificreports/ examination. Table 3 illustrates the relationship between the incidence of BAT activity and all parameters in detail. The incidence of BAT activity was associated negatively with age (p < 0.001) and positively with the condition of no preceding lymphoma-related therapy (p = 0.002) in univariate and multivariate regression. New BAT activity occurred more often after therapy with BEACOPP and less common after therapy with rituximab (p = 0.02, and p = 0.02, respectively), but both were not significant in multivariate regression. All other investigated therapies and chronic diseases were not associated with BAT activity in this context (see Table 3). Also, the amount of weight change between the first and second PET/CT examination was not linked with the incidence of BAT activity. Mature T-cell lymphoma 6 0 1 Other lymphoma 10 0 1

Discussion
In this study, we evaluated the prevalence and incidence of BAT activity in lymphoma patients. We assessed various potential determinants of BAT activity to gain further insights into the modulation of BAT as a potential therapeutic target against obesity. While the BAT prevalence might relate primarily to strong known impact factors and chronic conditions, the investigation of new incidence of BAT activity might reveal the potential impact of temporary factors. Our patient cohort showed a similar distribution of age, sex, and BMI to prior unselected studies on BAT activity 6,10,32 . Therefore, our cohort might compose a representative sample of a lymphoma population. The prevalence of BAT activity was 12% in our study, and the incidence of BAT activity in initially BAT negative patients was 10%. Both rates are higher than reported ratios of about 4-5% BAT activity in mixed populations 6,10,32 . In agreement, other studies with pediatric and adult lymphoma patients observed a comparatively high prevalence of BAT activity 6,30,33,34 . As expected, BAT activity was associated with its well-known determinants age, sex, BMI, and season 6,10 . Additionally, our results strengthen the potential role of BAT in the metabolic syndrome by showing a negative association with diabetes mellitus and arterial hypertension 6,10,35 . We found an analogy between BAT and betareceptors' activity, confirming prior findings where treatment with beta-blockers and beta-receptor agonists affected BAT 6,17 . On the other hand, BAT activity was not independently associated with other factors under discussion, like hyperlipidemia, thyroid dysfunction, prior cancer disease, and the specific lymphoma type 1,16,28,33,35 .
We found no association of BAT and lymphoma's metabolic activity in our cohort. Former studies reported variable results on the role of cancer vitality in the regulation of BAT activity 31,33,36 . Also, BAT activity was not linked with the extent of weight change during the disease, in contrast to prior reports on cancer cachexia 35 . Thus, other disease accompanying conditions, particularly the application of a specific therapy, might be relevant. The effect of preceding chemotherapy on BAT activity is controversial 10,30,37 . The present results plead for an inhibitory role of chemotherapy or radiotherapy on BAT since the BAT recruitment was higher in patients without preceding lymphoma-related therapy. Steroids have been reported to stimulate BAT activity in the short-term but inhibit its function during long-term use. The real impact of steroids on energy metabolism via regulation of BAT activity is unclear, and we did not find an association with BAT activity in the present study [21][22][23]30,38 . Also, cytarabine application did not affect BAT activity, in contrast to a previous study with a comparable patient cohort and although it shares the signaling pathway with BAT 30 . The impact of cytarabine on BAT activity might be minor and prone to interference by other influencing factors. Furthermore, the investigated patient numbers are low so far, and the potential role of this medication on BAT activity can be answered only after larger studies. The ABVD and BEACOPP regimen contain doxorubicin and vincristine, potential BAT recruiting agents 30,39 . Although we saw a tendency towards altered BAT activity by applying these chemotherapies, we could not independently prove an association from confounding factors. Further studies are needed on these medications as well.
Our study has some limitations due to its retrospective design and the limited numbers of patients in specific subgroups. Therefore, we could not consider specific lymphoma entities, all single drugs, and all potential confounders in our analyses. Furthermore, the patients did not undergo a cooling protocol or a preparatory high fat diet before the PET/CT examinations, which both increase BAT activity. In the assessment of BAT activity by FDG PET, the differentiation between active BAT and other conditions with metabolic activity, e.g., lymph nodes, can be challenging due to the unspecific nature of FDG uptake. Our approach, according to the BARCIST criteria, aimed to minimize this bias by a fixed lower cut-off value and a specified HU range in CT for defining BAT as well as drawing the VOI in a sufficient distance to visible lymph nodes 7 . Continuous quantification of BAT activity is not possible using a fixed cut-off value. Therefore, we differentiated two BAT activity levels based on the median SUV of patients with BAT activity, but they are still artificial. Finally, including only two PET/CT examinations per patient might introduce selection bias, but it results in a more balanced dataset than including different examination numbers per patient.
In conclusion, BAT activity in lymphoma patients is related to the well-known anthropometric factors, conditions of the metabolic syndrome, and the beta-receptors' activity. While we did not find a connection between BAT activity and a specific medication of lymphoma, preceding lymphoma-related therapy might have an inhibitory effect on BAT recruitment.

Methods
Study design and patients. We retrospectively evaluated all lymphoma patients who underwent two or more 18 F-FDG-PET/CT examinations between January 2012 and December 2018 in the PET/CT center of our university hospital. In patients with more than two PET/CT examinations, we included either the last PET/CT without BAT activity and the first examination with BAT activity, or-if not applicable-the first two examinations in the disease course. We assessed age, sex, and BMI and categorized the lymphoma entities according to the WHO classification in the four groups mature B cell-lymphoma, Hodgkin lymphoma, mature T cell-lymphoma, and other lymphoma 40 . We identified the lymphoma disease as active if vital manifestations with FDG uptake above the mean liver uptake were present. Cases with unclear faint FDG uptake in morphologic residual caused by minimal residual activity or reactive changes were not considered. We assessed the season at the PET/ CT examination and calculated the weight change between the two examinations. We noted, which therapy regimen was applied before the respective PET examination since the beginning of the disease or a former PET examination. We considered specific medications if more than thirty individuals received them. This criterion applied to the chemotherapy regimens ABVD (doxorubicin, bleomycin, vinblastine, dacarbazine), CHOP/ CHLIP (cyclophosphamide, doxorubicin, conventional/liposomal vincristine, steroids), and BEACOPP (bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine, steroids), as well as the separate drugs rituximab, cytarabine and steroids. We waived infrequently used drugs due to the limited statistical power. We searched the patients' history concerning the following further factors potentially linked with BAT activity: www.nature.com/scientificreports/ prior cancer disease, arterial hypertension, hyperlipidemia (known disease or treatment with statins), medication on the beta-adrenergic receptors (separate categories for beta-blockers, or beta-receptor agonists), and thiazolidinediones. We noted thyroid dysfunctions (differentiation in hypo-or hyperthyroidism) and diabetes mellitus if they were reported as diseases, a specific medication was applied, or the corresponding blood values were abnormal at the PET/CT examination (normal range of thyroid-stimulating hormone 0. . CT data were reconstructed in the axial direction (slice thickness 3 mm, increment 2.5 mm) and used for image analysis and attenuation correction. The commercial software SyngoVIA (Siemens Healthcare, Knoxville, USA) served as the study's viewing platform. We used a standardized uptake value (SUV) normalized on lean body mass for PET quantification. We reviewed the fused PET/CT images of all patients for BAT activity in the right and left nuchal, supraclavicular, and mediastinal regions according to the BARCIST criteria. We quantified BAT activity using isocontour volumes of interest (VOI) with a lower cut-off SUV of 1.2 if the region's density corresponded to fat in CT (Hounsfield Units − 10 to − 190) 7 . FDG positive lymph nodes near the areas of BAT assessment are a potential finding in lymphoma patients. We took special care to avoid spillover bias by adhering to a sufficient distance between visible lymph nodes and each VOI. We did not quantify regions with an SUV below the cut-off value, as these do not represent BAT activity according to BARCIST. In each patient, we noted the number of measurable VOI (active BAT depots) and averaged the mean SUV (SUVmean) of all active BAT depots. Subsequently, we classified the extent of BAT activity in three categories to reflect the continuous nature of BAT activity: high BAT activity (≥ 4 active BAT depots and SUVmean above the median SUVmean of all patients with BAT activity), low BAT activity (measurable BAT activity, but below the criteria for high BAT activity), no BAT activity (examples see Fig. 1). EG and CB read all PET/CT images in consensus (1 and 10 years of experience in hybrid imaging analysis).

Statistical analysis.
Shapiro-Wilk test denied the normal distribution of the data. The variables were categorical with a nominal or ordinal scale, except for the continuous variables age, BMI, and weight change. We www.nature.com/scientificreports/ tested the association between continuous and categorical variables with univariate logistic regression and Wilcoxon test, and the association between categorical variables with the Chi-Square test. We used Spearman correlation to test the relation of continuous variables. We identified independent parameters associated with BAT activity by multivariate regression with all significant parameters. We set the significance level at a p value ≤ 0.05 and used JMP 13.2 as statistical software (SAS, Cary, USA).