The study protocol was approved by the local ethics committee prior to patient enrollment (Nr. S 6(a)/2020, accepted on 18. February 2020) and the study was conducted according to the principles of the Declaration of Helsinki and its amendments. All subjects provided their informed consent before study participation. Lipedema was diagnosed according to the criteria described in our previous work (1), including (1) pain upon pressure and touch in the affected limbs, (2) negative Stemmer’s sign, (3) bilateral, symmetrical, disproportionate adipose tissue hypertrophy on the limbs, (4) easy bruising, and (5) sparing of the feet. For study inclusion, the diagnosis for all lipedema patients had to be confirmed by an independently practicing non-surgical physician. Conservative treatment methods must have been applied for at least 6 months prior to surgery (19). A total of 32 female patients were included in the study cohort which fulfilled the criteria, of which 9, 16, and 7 individuals could be assigned to the progredient lipedema severity stages I, II and III, respectively ( Table 2A ). The control group consisted of 14 women who did not report lipedema-associated symptoms or met any of the exclusion criteria and who underwent aesthetic liposuction of the extremities and abdomen. Individuals were age- and BMI-matched to the lipedema group. Exclusion criteria were: Previous surgery on the extremities (e.g. Varicose vein stripping, liposuction) or the abdomen (e.g. laparotomy), diabetes of any kind, severe uncontrolled endocrine, cardio-vascular, pulmonary or inflammatory/rheumatic diseases, moderate or severe psychiatric diseases of any kind, lymphedema according to the German Lymphedema Guidelines, current skin infections or other signs of acute inflammation, pregnancy, primary obesity without disproportion or secondary obesity (BMI > 35kg/m²), fat distribution disorders of other genesis (e.g., painless lipohypertrophy, benign symmetric lipomatosis) and fibromyalgia.

Adipose tissue biopsies were obtained from the thigh and the lower abdomen regions of female lipedema patients and age- and BMI-matched control subjects during elective liposuction. Tissue samples of the thigh were harvested in the ventral portion of the dominant affected (or right) thigh midway between the anterior superior iliac spine and the cranial margin of the patella. Samples from the caudal abdomen were taken originating from the umbilical approach. The biopsies were harvested before infiltration of the tumescent solution by surgical dissection through the regular surgical incisions for liposuction. One-half of the freshly obtained biopsy samples was stored in phosphate-buffered saline (PBS) with 3.5% BSA (Sigma-Aldrich, Taufkirchen, Germany) at 4°C immediately after collection and during transport. Subsequently, the specimens were fixed in 4% formaldehyde/PBS before embedding for histology. The remaining biopsy material was immediately transferred into dry 1.5 ml collection tubes and stored without addition of cryopreservant on dry ice. For long-term storage, frozen biopsy specimens and lipoaspirates were transferred to -80°C.

Adipose tissue samples of 14 controls and 9 stage I, 16 stage II and 5 stage III classified lipedema patients ( Table 2B ) were fixed in 4% paraformaldehyde (Carl Roth GmbH, Karlsruhe, Germany) for 24 hours at 4°C, dehydrated embedded in paraffin, and cut into 4-µm sections. To determine adipocyte diameter, area and number, sections were stained with hematoxylin and eosin (H&E) staining. Slices were photographed at 100x magnification and analyzed with Image J software (NIH, Bethesda, MA, USA) as described elsewhere (20). A total of three sections per patients and 2 non-overlapping fields per section were assessed. To determine fibrosis area, Sirius red staining was performed in adipose tissue deparaffinized slides incubated with a 0.1% Sirius red solution dissolved in aqueous saturated picric acid for 1 hour, washed in acidified water (0.5% acetic acid) and mounted for image analysis. All images were captured at 200x magnification and fibrosis area [μm²] was quantified using Image J software (NIH, Bethesda, MA, USA) with the freely available plugins “MRI fibrosis tool” and “Color Deconvolution” (21). A total of three sections per patient and 3 different non-overlapping fields per section were quantified ( Table 2D ).

For immunohistochemistry, a meticulous selection process was implemented to minimize the potential influence of age and BMI as factors affecting macrophage numbers. Initially, the 14 control samples were divided into three groups, ensuring equitable distribution while excluding the samples with the lowest and highest BMI values. Subsequently, from each stage, six lipedema patients with the lowest BMI were selected to match the control subgroups in terms of age and BMI. Regrettably, due to substantial damage during the paraffin embedding process, only five sections from stage III lipedema patients were accessible. As a solution, a matched stage II patient was included in the analysis to maintain consistency and comparability. The paraffin-embedded sections were deparaffinized and incubated with antibodies to pan-macrophage (anti-CD68, Abcam,Cambrige, United Kingdom), M1-like macrophages (anti-CD86, Abcam, Cambrige, United Kingdom), and M2-like macrophages (anti-MRC1/CD206, Abcam, Cambrige, United Kingdom) at 4°C overnight. Secondary antibodies were either Anti-rabbit (AlexaFluor-594-labelled, Thermo Fisher Scientific, Dreieich, Germany) or Anti-mouse (AlexFluor-488-labelled, Thermo Fisher Scientific, Dreieich, Germany). 4’,6-Diamidino-2-Phenylindole, Dihydrochloride (DAPI, Biolegend, London, United Kingdom) was used for nuclei staining. Images were captured at 200x, and labelled cells were counted manually using BZ-II Viewer software (Keyence, Ōsaka, Japan).

RNA from 30mg of minced and frozen adipose tissue samples ( Table 2C ) was homogenized using a bead-based micro-blender with in QIAzol lysis reagent (Qiagen, Hilden, Germany for 5 min at 4°C followed by centrifugation for 10min at 4°C and 12,000 rpm to separate lipid layer from aqueous extract. Phase separation was initiated by adding 100 µl chloroform. After 15 min of centrifugation at 12,000 rpm and 4°C, RNA was precipitated from the upper clear phase with 250 µl of isopropanol. RNA was purified with 500 µl of 75% ethanol, dried and solubilized in 30 µl sterile-filtered water treated with di-ethyl-pyrocarbonate (DEPC). Purified RNA was reversely transcribed into cDNA using a high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific, Dreieich, Germany). Quantitative real-time PCR was performed using Maxima SYBR Green/ROX qPCR Master Mix (Thermo Fisher Scientific, Dreieich Germany) on a CFX384 Touch instrument (Bio-Rad, Munich, Germany). Primers were designed as intron spanning sequences to specifically amplify cDNA (primer sequences: Supplementary Table S1 ). The mRNA expression levels were determined using the 2^(–ΔΔCT) method and normalized to mRNA levels of human beta actin (hACTB) as housekeeping gene.

For the assessment of the metabolic risk profile, blood plasma samples were collected preoperatively after 6-8 h of starvation in an S-monovette (Sarstedt, Nuernbrecht, Germany). Blood samples were stored on ice until further processing, subsequently centrifuged for 10 min at 6,000 rpm at 4°C, after which plasma was collected and stored at -80°C until analysis. Plasma profiles of 14 controls were analyzed and compared to 8 stage I, 16 stage II and 7 stage III classified lipedema patients ( Table 2E ). Plasma insulin was determined using a human insulin ll ELISA kit (BioVendor R&D, Czech Republic) according to the supplied protocol. Blood parameters were measured on a Cobas Mira Analyzer (Roche Diagnostics, Mannheim, Germany) by using commercially available assay kits for free fatty acids (FFA), (Wako chemicals, Neuss, Germany), glycerol (Randox, Crumlin, UK), triglycerides, cholesterol, high- density lipoprotein (HDL), and glucose (Axonlab, Stuttgart, Germany).

Analyses were performed using GraphPad Prism, version 8.0.0 for Windows (GraphPad Software, San Diego, CA, USA). All data were presented as mean ± standard error of the mean (SEM), or median (interquartile range, 25-75%). Sample sizes as well as the statistical analyses are indicated in the respective figure legends. Welch corrected t-test and multiple t-tests were applied for comparisons between two groups. Statistically significant differences are assumed for p-values < 0.05.

After verification of previously published diagnostic criteria (1) and evaluation of exclusion criteria, a total of 32 lipedema patients were recruited in our study and divided into severity stages (stage I: n=9; stage II: n=16; stage III: n=7). Moreover, 14 controls subjects were enrolled that were matched to the patient cohort for BMI ( Table 2A ; control: 26.7 ± 4.8 kg/m² versus lipedema: 28.9 ± 4.0 kg/m², p=0.1498) and age ( Table 2A ; control: 38.7 ± 9.9 versus Lipedema: 37.4 ± 12.4, p=0.7131).

Adipose tissue morphology was assessed in thigh and abdominal biopsies for all patients included in this study (stage I: n=9; stage II: n= 16; stage III: n=5) and compared to morphologies of the 14 non-lipedema control subjects ( Table 2B ). In general, individual adipocyte size was larger in thigh biopsies compared to abdominal regions in both groups. In lipedema patients a significantly elevated average adipocyte size was observed in the thigh region while no differences were detected in abdominal biopsies ( Figures 1A, B ).

The stage dependent analysis revealed no significant differences between lipedema and controls in both biopsies comparing stage I patients to the control group ( Figure 1C , Supplementary Figure S1 ). Conversely, a higher abundance of hypertrophic adipocytes in thigh and abdominal biopsies of lipedema patients of stage II was observed, showing elevated counts of smaller adipocytes and reduced numbers of large-area adipocytes in biopsies of the control group ( Figure 1D , Supplementary Figure S1 ). In stage III lipedema patients, a similar pattern emerged, altogether shifting adipocyte distribution towards larger cells in lipedema thigh regions, albeit not reaching statistical significance in the abdominal samples ( Figure 1E , Supplementary Figure S1 ).

In summary, the phenotypical changes of the pathology were most pronounced in stage II and III patients, in comparison to the control group and stage I patients. A potential explanation for the seemingly milder phenotype in stage III is that this subgroup of patients also presented with the highest BMI ( Table 2B ). This distinction could act as confounding factor resulting in increased adipocyte sizes in all depots of stage III patients independently of lipedema itself.

The appearance of hypertrophic adipocytes is frequently accompanied by remodeling of the extracellular matrix (ECM). We therefore conducted a histological analysis of fibrosis levels in controls and lipedema patients. No difference of pericellular fibrosis was detected in the non-symptomatic lower abdominal region when comparing the control group to all lipedema patients while a significant increase of fibrotic areas was detected in the lipedema-affected thigh compared to healthy controls ( Figures 2A, B ). Evaluation of stage-dependent fibrosis levels in the abdominal region overall confirmed no evidence for elevated fibrosis in any stage, but unexpectedly revealed significantly lower fibrosis levels in stage III ( Figure 2C ). In contrast, a significant stage-dependent increase of fibrosis was found in stage I and II patients and a similar trend was observed in stage III when comparing histologies of thigh biopsies to the control group ( Figure 2C , Supplementary Figure S2A ).

Taken together, our findings show that pathological ECM-remodeling of thigh, but not abdominal adipose tissue depots manifests early on in the disease, i.e. at stage I, thus preceding adipocyte hypertrophy and altogether suggesting that this aspect of lipedema warrants further investigation as a marker with diagnostic value.

To define alterations of adipose tissue morphology in lipedema on the transcriptional level, we measured the gene expression profiles of markers reflecting the processes of adipogenesis, inflammation, fibrosis and angiogenesis. Additionally, the marker peroxisome proliferator-activated receptor gamma-coactivator 1 alpha (PPARGC1A, also known as PGC1A) was assessed as a general marker of mitochondrial function. In abdominal and thigh biopsies comparing control subjects to all lipedema patients, several marker genes of inflammation and angiogenesis were broadly and significantly upregulated, suggesting that this process is a target of lipedema pathology ( Figure 3A , Supplementary Figures S3 , S4 ). We subsequently re-analyzed the data by comparing the control group to the three stage groups individually. Here, no statistically different gene expression was recorded at stage I in the lipedema-affected thigh region except for hypoxia-inducible factor 1-alpha (HIF1A, Supplementary Figure S4 ), while expression of inflammatory marker gene interleukin-6 (IL6) and fibrosis marker collagen 6a1 (COL6A1) were elevated in abdominal biopsies of stage I patients ( Figure 3B , Supplementary Figure S3B ). In abdominal biopsies, these changes, in particular pertaining to IL6, remained relatively stable throughout the more severe stage II, but only limited, if any, further differences became apparent, which was consistent with the relatively mild changes in the all group-comparison of abdominal biopsies ( Supplementary Figure S3 ). Conversely, thigh biopsies started featuring several significantly differentially expressed genes in stages II and III, such as reduced expression of the adipokine leptin (LEP), and increased expression of angiogenesis marker gene vascular endothelial growth factor C (VEGFC) Supplementary Figures S4C, D ) and several inflammatory marker genes, including IL6, tumor necrosis factor-alpha (TNF), and macrophage markers CD86 and mannose receptor c-type 1 (MRC1; also known as CD206) ( Figures 3C, D ). In none of the conditions, PPARGC1A ( Figure 3 ) expression was altered, suggesting that mitochondrial biogenesis is normal during lipedema disease progression. In summary, these data suggest that non-affected adipose tissue regions, such as the abdominal depot, may display altered gene expression patterns as an early indicator of lipedema, whereas expression of inflammatory markers genes is stage-dependent in affected adipose depts of the extremities and correlates with disease severity. Decreased expression of leptin also suggests an altered endocrine profile of the affected adipose thigh regions.

To corroborate the increased gene expression markers associated with inflammation and adipose tissue macrophages in thigh adipose tissue of lipedema patients, immunohistochemistry of CD68+ pan-macrophages, CD86+ M1-like (pro-inflammatory) macrophages ( Supplementary Figure S5A ), and CD206+ M2-like (anti-inflammatory) macrophages were performed ( Figure 4A , Supplementary Figure S5B ). Crown-like structures (CLS) as hallmarks of proinflammatory processes in adipose tissue were present in all stages of lipedema ( Figure 4A ), but no significant differences in the total numbers of CD68+ and pro-inflammatory M1 macrophages were detected when comparing all patients to the control group ( Figure 4B ). In contrast to this, the number anti-inflammatory M2-like macrophages, expressing the surface marker CD206 was significantly increased in lipedema patients in the stage-independent analysis ( Figure 4B ) As before, this analysis was re-assessed by comparing the control subjects to the subgroups representing stages I through III. For this analysis, age- and BMI-matched individuals were selected from the control group for comparisons to rule out these two parameters as potential determinants of altered macrophage numbers. Although no significant differences regarding total macrophages were evident in the all group-comparison, we detected a progressive increase of CD68+ macrophages from stage I o stage III whereas no such increase was found in the matched control subsets ( Figure 4C ). The staining for specific macrophage subtypes revealed a stage-dependent increase of M1-like macrophages in lipedema patients from stage I to stages II and III. This effect was less pronounced in the control groups, where the stage I control group only differed significantly from the stage III control group, presumably due to the increase in BMI between these subgroups ( Figure 4D , Table 2D ). As a result of this trend in patients, M1-macropage levels were significantly reduced at stage I compared to the control group and a marked trend was found for increased M1-levels at stage III compared to the control group ( Figure 4D ). Numbers of CD206+ anti-inflammatory, M2-like macrophages were found to be increased in thigh adipose biopsies of stage II lipedema patients compared to the matched control groups, again confirming the observations form the gene expression and all group-comparison analyses ( Figures 4B, E ). As expected, the same stage-correlating increase of anti-inflammatory macrophages from stage I to III lipedema was found, while no alterations in the control subgroups were found ( Figure 4E ). In summary, a stage-dependent increase of macrophage levels in lipedema was observed while cell numbers remained relatively stable in age- and BMI-matched controls assigned to the respective disease stages. Somewhat unexpectedly, paralleling trends towards increased M1- and M2-macrophage accrual was detected in this analysis. Calculating rations of the two distinct macrophage subsets (M2:M1) is a tool frequently used to assess whether pro- versus anti-inflammatory signals are more prevalent in the adipose tissues. Accordingly, the M2:M1 macrophage ratio was significantly increased in stage I patients compared to the ratio for matched controls, and showed a similar trend (p=0.076) in stage II patients ( Figure 4F ). Conversely, no difference in the ratios of stage III patients and matched controls was evident, altogether suggesting a stage-dependent decline of the M2-driven anti-inflammatory milieu in late-stage disease ( Figure 4F ).

As lipedema-associated changes in our cohort, such as adipocyte hypertrophy, interstitial fibrosis, and inflammation, are commonly associated with an altered metabolic profile in patients with primary obesity, we analyzed plasma samples to assess glucose and lipid metabolism in patients and controls ( Table 3 ). While no differences were found in parameters reflecting glucose homeostasis, including the homeostatic model assessment of insulin resistance (HOMA-IR), the plasma lipid profile displayed several differences. Total cholesterol and triglyceride levels were found to be normal in lipedema patients, but elevated high-density lipoprotein (HDL) cholesterol and reduced plasma glycerol were detected. In addition, the low-density lipoprotein (LDL)/HDL cholesterol and the triglyceride/HDL ratios, representing potential predictors for the risk of metabolic diseases, were lower in lipedema patients. The group breakdown into disease stages and the corresponding control subgroups led to loss of statistical significance, but the data show that these lipedema-associated changes in plasma lipids were robust and evident at all stages of disease severity, pertaining to HDL, glycerol and the LDL/HDL and triglycerides/HDL ratios, while no additional stage-specific differences occurred in any of the other parameters ( Figure 5 , and not shown).