ADF was prepared according to our previously published protocol^[20]. The precursor material was derived from abdominal adipose tissue obtained by liposuction from multiple female donors aged 25-35 years. Eligible donors were women aged 25-35 years undergoing abdominal liposuction who provided written informed consent for tissue use; donors with positive screening results for infectious pathogens were excluded. Tissue processing began immediately after liposuction. Adipose tissue from eligible donors was pooled to prepare a single precursor ADF batch. The exact number of contributing donors was not retained in the available source records. Human adipose tissue collection was approved by the Ethics Committee of the Fourth Military Medical University (No. KY20243548-1), and all procedures were conducted in accordance with the Declaration of Helsinki.

The pooled adipose tissue underwent sequential crushing, washing, and mechanical pressing to form ADF, which was cryopreserved at -80 °C for 12 months. After thawing, the ADF was manually cut into fragments using sterile surgical scissors. Fragments from the same cryopreserved batch were pooled, thoroughly mixed, and aliquoted before physiological saline was added. The aliquots were mixed with sterile physiological saline at four weight-to-volume ratios: 1:5, 1:7.5, 1:10, and 1:15 (w/v), designated as I-ADF-1:5, I-ADF-1:7.5, I-ADF-1:10, and I-ADF-1:15, respectively. Fragment dimensions were not quantitatively standardized or measured, and the fragment-size range and mean particle size were therefore unavailable. All formulations were prepared within 2 h of use.

Each freshly prepared formulation was loaded into the same type of syringe and manually extruded through a 17-gauge needle under the qualitative test conditions used in this study. Successful passage through the needle, continuity and visual homogeneity of the extrudate, and visible water exudation or phase separation were documented photographically. Fragment size, injection force, extrusion rate, viscosity, storage modulus, and loss modulus were not quantitatively measured. Therefore, the assessment was restricted to qualitative extrusion behavior under the specified 17-gauge test conditions and should not be interpreted as a quantitative comparison of injectability.

Implant samples were fixed in 4% paraformaldehyde for 24 h, dehydrated through graded ethanol, cleared in xylene, embedded in paraffin, and sectioned at a thickness of 5 μm.

For hematoxylin and eosin (H&E) staining, the sections were deparaffinized in xylene, rehydrated through graded ethanol, stained with hematoxylin for 8-15 min, differentiated in 1% hydrochloric acid alcohol, blued in running water, and counterstained with eosin for 2-5 min. The sections were then dehydrated, cleared in xylene, and mounted.

For Masson’s trichrome staining, deparaffinized sections were treated with potassium dichromate, stained with Weigert’s iron hematoxylin for 5 min and acid fuchsin for 10 min, differentiated with phosphomolybdic acid, and stained with aniline blue for 30 s. The sections were subsequently dehydrated, cleared, and mounted.

Histological evaluation included both the peripheral rim and central core of the implant. The archived histopathology records did not contain sufficient information to determine the exact number of sections selected for quantitative analysis, their depth or anatomical level within each paraffin block, or the interval between analyzed sections. The records also did not permit verification of whether section-level sampling was applied consistently across experimental groups. In addition, the archived analysis records did not permit reliable reconstruction of the exact number or selection procedure of microscopic fields used for regional quantitative analysis. H&E staining was used for general morphological assessment, whereas Masson’s trichrome staining was used to assess adipose-like areas and collagen-containing tissue within the implants.

Paraffin sections were deparaffinized, rehydrated, and subjected to antigen retrieval using 10 mmol/L citrate buffer at pH 6.0. No separate permeabilization step was documented in the archived protocol for these paraffin-embedded implant sections.

Sections were blocked with 5% normal goat serum and incubated overnight at 4 °C with anti-Perilipin-1 (Abcam, Cat. No. ab3526, 1:200) for adipocyte identification, anti-CD31 (Servicebio, Cat. No. GB11063-2, 1:1,000) for total vessel labeling, and anti-α-smooth muscle actin (α-SMA; Servicebio, Cat. No. GB12045, 1:5,000) to identify α-SMA-positive cells associated with vascular structures. After washing, sections were incubated with the corresponding species-specific Alexa Fluor 488- or Alexa Fluor 594-conjugated secondary antibodies (Invitrogen, 1:500) for 1 h at room temperature, followed by 4’,6-diamidino-2-phenylindole (DAPI) nuclear counterstaining.

The exact host species and catalog numbers of the secondary antibodies were not retained in the archived service records and could not be reliably reconstructed. The archived records also did not contain sufficient information to verify the exact negative-control configuration used for this specific staining run.

Images were acquired using an LSM 800 confocal microscope (Zeiss, Germany). According to the standardized service workflow, all experimental groups within each staining experiment were imaged using the same acquisition procedure. The exact numerical values for laser power, detector gain, pinhole diameter, scanning speed, and related confocal acquisition parameters were not retained in the archived service records. Images were quantified using ImageJ version 1.53 (National Institutes of Health, USA). The endpoint-specific image-analysis criteria used for adipose regeneration and vascular quantification are described below.

Newly formed adipose tissue was evaluated separately in the peripheral and central zones of each implant. The peripheral zone was defined as the outer 500-μm rim of the implant, whereas the central zone was defined as the remaining inner core region. Quantitative measurements were performed separately for the predefined peripheral and central regions of each implant. For the week-12 quantitative analyses reported here, five implants were analyzed per group, with each implant treated as one independent biological replicate. Thus, five independent biological values were obtained per group for each regional endpoint.

Newly formed adipose tissue was operationally defined as tissue areas containing morphologically recognizable adipocytes on H&E-stained and Masson’s trichrome-stained sections, with adipocyte identity further supported by Perilipin-positive immunofluorescence. The percentage of the analyzed area occupied by adipose-like structures was quantified on Masson’s trichrome-stained sections, while the Perilipin-positive area was independently quantified by immunofluorescence.

Image analysis was performed using ImageJ version 1.53 (National Institutes of Health, USA), following the same analytical framework as in our previous ADF study^[20]. For Masson’s trichrome-stained sections, unstained or blank spaces were not classified as adipose tissue solely on the basis of staining intensity; adipose-like areas were included only when morphologically recognizable adipocytes were present, according to the operational definition described above. For Perilipin immunofluorescence, the Perilipin-positive area was quantified as the proportion of the predefined region of interest occupied by positive fluorescence. The same documented analytical framework was applied to coded images across all experimental groups within each staining experiment. However, the archived analysis records did not retain the exact numerical fluorescence threshold values, detailed segmentation parameters, or predefined rules for manual correction of technical artifacts; therefore, these pixel-level analysis parameters could not be reliably reconstructed retrospectively. For each implant, the region-specific quantitative value was used as the implant-level value for that region. Each implant was treated as one biological replicate for statistical analysis. Images bearing coded identifiers were analyzed by investigators blinded to group allocation.

Total vessel density was operationally defined as the number of identifiable CD31-positive vascular structures per high-power field (HPF). An α-SMA-associated vessel was defined as a CD31-positive vascular structure surrounded by α-SMA-positive cells, and the number of such structures per HPF was used to calculate α-SMA-associated vessel density. This marker-based classification followed the same analytical framework used in our previous ADF study^[20], in which CD31 and combined CD31/α-SMA staining were used to assess total and α-SMA-associated vascular structures, respectively. Vessel densities were quantified separately in the predefined peripheral and central regions of each implant. Five implants were analyzed per group at week 12, with each implant treated as one independent biological replicate.

Images were analyzed using ImageJ version 1.53. The same documented vessel-counting definitions were applied to coded images across all experimental groups. However, the archived analysis records did not retain more detailed criteria for resolving ambiguous vascular profiles, nor did they retain the exact fluorescence threshold or segmentation settings; therefore, these details could not be reliably reconstructed retrospectively. For each implant, the region-specific vessel-density value was used as the implant-level value for that region. Each implant was treated as one biological replicate for statistical analysis. Coded images were analyzed by investigators blinded to group allocation.

Volume retention was assessed at weeks 4 and 12 after injection. The initial implant volume (V₀) was defined as the delivered formulation volume of 0.5 mL. At each prespecified terminal endpoint, the implants were carefully excised with the implant boundary preserved, and macroscopically identifiable surrounding tissues were removed without disrupting the implant structure. Harvested implant volume (V_t) was determined using a physiological saline displacement method. Briefly, the saline volume was recorded before (V_before) and after (V_after) complete immersion of the implant using a graduated syringe with a resolution of 0.01 mL. Visible air bubbles were removed before the final reading, and harvested implant volume was calculated as V_t = V_after - V_before. Volume retention (%) was calculated according to:

Volume retention (%) = (V_t /0.5 mL) × 100%.

After harvest, implants were assigned coded identifiers that did not disclose the formulation group. Volume measurements and calculations were performed by investigators blinded to group allocation. Group codes were disclosed only after completion and verification of all volume measurements. Each implant was measured three times, and the mean of the three measurements was used as the implant-level volume value.

All animal experiments were approved by the Ethics Committee of the Fourth Military Medical University (No. 20240726). Forty 8-week-old female BALB/c-nu/nu nude mice were allocated to four formulation groups using a random-number table (n = 10 per group): I-ADF-1:5, I-ADF-1:7.5, I-ADF-1:10, and I-ADF-1:15. Because sample collection at weeks 4 and 12 required terminal harvest, five mice within each formulation group were prospectively assigned to the week-4 terminal cohort and five to the week-12 terminal cohort. No animal contributed data to more than one terminal time point. Each mouse received a single 0.5-mL dorsal subcutaneous injection of the assigned formulation through a 17-gauge needle.

Blinding was not feasible during injection because differences in formulation consistency were visually apparent. However, all subsequent assessments, including gross volume measurement, histological evaluation, and immunofluorescence analysis, were performed on coded samples by investigators blinded to group allocation.

No formal a priori power calculation was performed, and no numerical effect size was prespecified. Sample-size selection was informed by our previously published study of the ADF platform^[20], which used a closely related dorsal subcutaneous implantation model in nude mice and reported relatively limited within-group variability in week-12 implant-volume measurements. Because that study evaluated membrane-form ADF and cryopreserved ADF rather than injectable formulations with different dilution ratios, its data could not provide a directly transferable numerical effect-size estimate for the present formulation comparison. The present study was designed as an exploratory comparison of four I-ADF formulations. Based on the variability observed in the previous ADF study, the intention to identify relatively large formulation-dependent differences, and the reduction principle of the 3Rs, five mice per formulation at each terminal time point were selected.

Predefined exclusion criteria included postoperative death, severe infection, wound complications, implant loss or displacement, and unsuccessful sample collection.

All data are presented as mean ± standard deviation (SD). Before performing the parametric analyses, normality and homogeneity of variance were assessed separately for each endpoint. Normality was evaluated using the Shapiro-Wilk test, and homogeneity of variance among the four formulation groups was evaluated using Levene’s test. A p value greater than 0.05 was considered to indicate no statistically significant violation of the corresponding assumption. No statistically significant departure from normality or homogeneity of variance was detected for any dataset analyzed using ordinary one-way analysis of variance (ANOVA). Comparisons among multiple groups were therefore performed using ordinary one-way ANOVA followed by Tukey’s multiple-comparisons test.

For the volume-retention analysis, the week-4 and week-12 datasets were obtained from independent terminal cohorts and were assessed and analyzed separately within each time point. No repeated-measures analysis or direct statistical comparison between the two time points was performed. Because the two analyses addressed separate, prespecified endpoint-specific questions and no joint or longitudinal inference across time points was made, no additional multiplicity adjustment across time points was applied.

Pearson’s correlation analysis was used to evaluate the linear association between central α-SMA-associated vessel density and the central Perilipin-positive area using paired central-zone values from individual biological samples. Before the Pearson correlation and exploratory linear regression analyses, linearity was examined using scatterplots, and the normality and homoscedasticity of the residuals were assessed. No major violation of these assumptions was identified. The correlation coefficient and corresponding p value were reported. Exploratory linear regression analysis was used to evaluate the overall association between ADF concentration and week-12 volume retention. A two-tailed P value < 0.05 was considered statistically significant. All statistical analyses, including the assumption checks, were performed using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA).

All 40 animals completed the prespecified experimental protocol. No animal met the prespecified animal- or implant-level exclusion criteria. Accordingly, five animals per formulation group were included at each terminal time point. After 12 months of cryopreservation at -80 °C, the ADF retained its gross morphology and histological architecture after thawing, indicating preservation of its structural features under the evaluated storage conditions [Figure 1A]. The ADF was then cut into fragments and mixed with physiological saline at four dilution ratios (1:5, 1:7.5, 1:10, and 1:15, w/v), designated as I-ADF-1:5, I-ADF-1:7.5, I-ADF-1:10, and I-ADF-1:15 [Figure 1A and B]. Gross examination showed progressively greater free-water separation as the saline proportion increased. I-ADF-1:5 appeared relatively dry, I-ADF-1:7.5 appeared moister, and visible free water was present in I-ADF-1:10 and I-ADF-1:15 [Figure 1B]. Representative H&E-stained and Masson’s trichrome-stained sections showed no obvious qualitative differences in histological architecture among the four formulations before implantation [Figure 1C and D]. These observations indicated broadly comparable baseline histological appearances among the four formulations. We next compared their qualitative extrusion behavior and in vivo performance.

All four formulations could be manually extruded through a 17-gauge needle under the qualitative test conditions used in this study [Figure 2A]. I-ADF-1:5 and I-ADF-1:7.5 produced relatively continuous and visually homogeneous extrudates. In contrast, I-ADF-1:10 and I-ADF-1:15 showed visible water exudation before extrusion of matrix fragments, resulting in less uniform material delivery. However, because fragment size, injection force, extrusion rate, and rheological properties were not quantitatively measured, these observations should not be interpreted as a quantitative comparison of injectability.

At week 4, the I-ADF-1:5 and I-ADF-1:7.5 groups showed similar volume retention rates, with no significant difference between them. Both groups showed higher volume retention than the I-ADF-1:10 group (P < 0.01) and the I-ADF-1:15 group (P < 0.001; Figure 2B and C). At week 12, the volume retention rates were 86.86% ± 3.65% (I-ADF-1:5), 88.26% ± 3.98% (I-ADF-1:7.5), 50.44% ± 6.22% (I-ADF-1:10), and 41.78% ± 10.52% (I-ADF-1:15), showing a similar intergroup pattern to that observed at week 4. However, because the two time points were represented by independent terminal cohorts, the present data do not demonstrate within-animal volume stability over time.

At week 12, ADF concentration was positively associated with volume retention across the four tested formulations [Figure 2D]. Volume retention increased from 41.78% in the I-ADF-1:15 group to 88.26% in the I-ADF-1:7.5 group, whereas the I-ADF-1:5 and I-ADF-1:7.5 groups showed comparable retention rates (86.86% and 88.26%, respectively; P > 0.05).

H&E staining at week 12 revealed abundant adipocyte-like cells in the peripheral zones of all four groups [Figure 2E], consistent with broadly comparable peripheral adipose-like tissue formation among the tested formulations. In contrast to the peripheral zones, central zone adipose regeneration showed marked differences among groups [Figure 2E]. Central-zone regeneration therefore provided greater discrimination among the four formulations than peripheral regeneration. To systematically evaluate this differential regenerative capacity, we next performed quantitative histomorphometric analysis using Masson’s trichrome staining and Perilipin immunofluorescence.

Masson’s trichrome staining showed that the newly formed adipose structures were surrounded by organized collagen tissue, consistent with local ECM remodeling [Figure 3A]. Consistent with H&E findings, the peripheral zones of all groups showed abundant adipose-like structures. Quantitative analysis showed that adipose-like structures occupied approximately 80% of the analyzed peripheral area across all groups, with no statistically significant differences among the four formulations (P > 0.05) [Figure 3B]. These findings indicate broadly comparable peripheral adipose regeneration among the four formulations. At week 12, the I-ADF-1:7.5 group exhibited near-complete filling of the outermost peripheral zone with adipose tissue, closely resembling normal adipose tissue architecture. In the central zone, however, the I-ADF-1:7.5 group displayed large, continuous areas of adipose tissue with mature structural characteristics, whereas the other groups showed scattered or discontinuous adipose distribution [Figure 3A].

Quantitative analysis of the central adipose-like area showed that the I-ADF-1:7.5 group had the highest value (60.50% ± 6.31%), significantly exceeding the other groups (P < 0.001) [Figure 3B]. The I-ADF-1:10 and I-ADF-1:15 groups showed comparable central adipose-like area values (30.91% ± 5.49% and 26.64% ± 7.46%, respectively; P > 0.05), whereas the I-ADF-1:5 group had the lowest mean value (19.66% ± 3.23%). The disparity between peripheral and central regeneration was minimal in the I-ADF-1:7.5 group but most pronounced in the I-ADF-1:5 group, indicating that I-ADF-1:7.5 was associated with the highest central adipose regeneration among the four tested formulations.

Perilipin immunofluorescence supported the histological findings [Figure 3C]. In the peripheral zones, all groups showed extensive peripheral Perilipin-positive areas, with no significant differences among groups [Figure 3D]. In the central zone, the I-ADF-1:7.5 group had the largest Perilipin-positive area (59.94% ± 4.28%), significantly greater than those in the I-ADF-1:10 (29.86% ± 4.10%), I-ADF-1:15 (27.94% ± 3.01%), and I-ADF-1:5 (20.18% ± 3.19%) groups (P < 0.001) [Figure 3D]. The I-ADF-1:7.5 group showed large, continuous Perilipin-positive areas, while the I-ADF-1:10 and I-ADF-1:15 groups showed a mixture of continuous and scattered Perilipin-positive areas, and the I-ADF-1:5 group showed a predominantly scattered distribution. The concordance among H&E, Masson’s trichrome, and Perilipin staining further suggests that the 1:7.5 formulation was associated with more uniform adipose tissue formation within the analyzed areas.

Given that the 1:7.5 formulation showed the highest central adipose regeneration among the tested ratios, we next investigated whether this advantage was associated with enhanced angiogenesis, particularly in the central zone.

In the peripheral zone, CD31 immunofluorescence staining revealed no significant differences in total vessel density among the four groups (P > 0.05). Similarly, CD31/α-SMA co-staining showed no significant differences in α-SMA-associated vessel density among groups (P > 0.05) [Figure 4A-C]. The comparable peripheral vascularization among groups was consistent with similar levels of peripheral adipose regeneration.

By contrast, vascularization in the central zone differed among the four formulations. The I-ADF-1:7.5 group had the highest mean total vessel density (25.00 ± 7.21 vessels/HPF). This value was significantly higher than that in the I-ADF-1:5 group (13.60 ± 3.58 vessels/HPF; P < 0.05), whereas the differences from the I-ADF-1:10 (18.80 ± 3.89 vessels/HPF) and I-ADF-1:15 (16.20 ± 4.09 vessels/HPF) groups were not statistically significant [Figure 4B]. The I-ADF-1:7.5 group also had the highest mean central α-SMA-associated vessel density (20.80 ± 6.38 vessels/HPF). It was higher than that in the I-ADF-1:5 group (8.00 ± 2.83 vessels/HPF; P < 0.001) and the I-ADF-1:15 group (12.60 ± 4.19 vessels/HPF; P < 0.05), but did not differ significantly from that in the I-ADF-1:10 group (14.40 ± 2.30 vessels/HPF) [Figure 4C]. Thus, among the four formulations tested, I-ADF-1:7.5 had the highest mean central total and α-SMA-associated vessel densities.

In an exploratory pooled analysis across all 20 biological samples, Pearson’s correlation analysis showed a positive association between central α-SMA-associated vessel density and the central Perilipin-positive area at week 12 (r = 0.719, R² = 0.5174, P = 0.0004; Figure 4D). Because samples from four formulation groups were combined, this association may partly reflect between-formulation differences and does not establish causality.