The impact of the ethanolic fraction of Skeletonema costatum (hereafter referred to as SKLT) on bone regeneration was initially evaluated using the zebrafish caudal fin regeneration assay. This model was selected due to its simplicity, high reproducibility, and suitability for studying the coordinated actions of multiple cell types during de novo bone formation. It also allows for precise in vivo cell tracking (45, 47, 48). Following caudal fin amputation, adult zebrafish were exposed to SKLT by immersion in water supplemented with the fraction. Regenerative outgrowth, mineralization, and morphological patterning of the regenerating bony rays were assessed at a single time point: 5 days post-amputation (5 dpa, Figures 1A, B ). Quantitative analysis showed a trend toward reduced regenerative area in SKLT-treated animals, though this reduction did not reach statistical significance ( Figure 1C ). In contrast, SKLT exposure resulted in a significant increase in the area of mineralized tissue ( Figure 1D ), suggesting enhanced mineral deposition. While the width of regenerated rays remained unchanged ( Figure 1E ), a marked distalization of ray bifurcation points along the proximal–distal axis was observed in SKLT-treated fish ( Figure 1F ).

To investigate whether the observed effects on mineralization were associated with antiresorptive activity, the dynamics of osteoclast precursors—identified as rounded cathepsin K (ctsk)-expressing cells – in Tg(Ola.ctsk:FRT-DsRed-FRT-Cre, myl7:EGFP)^mh201 transgenic zebrafish (49), hereafter referred to as Tg(ctsk:DsRed), was monitored in fish continuously exposed to SKLT for 10 days following caudal fin amputation ( Figure 2A ).

In vehicle-treated controls, the number of ctsk⁺ cells peaked between 24- and 48 hours post-amputation (hpa) ( Figures 2B, D ), followed by a progressive decline in cell density within the regenerated area. In contrast, SKLT-treated fish showed significantly fewer ctsk⁺ cells recruited to the blastema at both 24 and 48 hpa ( Figures 2C, D ). A delayed peak was observed at 72 hpa, after which ctsk⁺ cell counts were comparable between SKLT-treated and control animals.

Notably, elongated, tubular ctsk-expressing osteoclasts resembling osteolytic tubules (OLTs), previously described in regenerating fin tissue (48), appeared adjacent to the regenerated rays by 96 hpa in control fish, but were delayed until 120 hpa in SKLT-treated fish ( Supplementary Figure S1 ).

This reduction in osteoclast precursor recruitment correlated with a significant decline in early osteoclastic activity. TRAP (tartrate-resistant acid phosphatase) staining, a marker of mature osteoclast function, was nearly absent at 24 hpa in SKLT-exposed animals but returned to levels comparable to controls by 240 hpa ( Figures 3A, B ). Furthermore, in the 10-day exposure experiment, SKLT-treated fish exhibited a sustained reduction in the area of regenerative outgrowth beginning at 72 hpa and persisting through 240 hpa ( Figure 4 ). This suggests that the ethanolic fraction may interfere with molecular pathways governing blastema expansion during the regeneration process.

To investigate whether immunomodulatory mechanism underlies the effects observed in SKLT-treated zebrafish, tissue-specific transcriptomic analysis was performed on regenerating caudal fin blastemas collected at 24 hours post-amputation — the time point at which SKLT exerted the strongest suppression of osteoclast precursor recruitment.

Bulk RNA sequencing identified 720 differentially expressed genes (false discovery rate, FDR< 0.01), of which 361 were upregulated and 359 were downregulated ( Supplementary Figure S2 ). Gene ontology (GO) analysis was performed using the DAVID platform, and significantly enriched Biological Process (BP) terms were filtered for uniqueness and plotted using REVIGO ( Figure 5A ). Heatmaps were then generated to visualize the expression patterns of genes associated with the most relevant enriched processes.

Among the most significantly affected biological processes were those associated with both innate and adaptive immune responses. Genes involved in inflammatory signaling—including irg1l, nos2a, thbs1, and il20ra—as well as those linked to the interferon system (ifi30, ifng1r, irf4b, irf10) and the complement cascade (c4, c7a) were consistently downregulated in SKLT-treated blastemas ( Figure 5B ). Since interferon gamma (IFN-γ), primarily produced by Th1 cells (50), promotes antigen-dependent T-cell activation and the expression of pro-osteoclastogenic cytokines (50), genes involved in T-cell signaling and antigen presentation were examined. Indeed, multiple genes central to T-cell activation and MHC class II antigen presentation—including socs1a, itk, il7r, batf, cd4, cd74a, cd74b, and jak1 ( 30, 31)—were downregulated following SKLT exposure ( Figure 5B ).

Conversely, several immunoregulatory genes such as fn1a, fn1b (fibrin isoforms), and il13ra2 (interleukin 13 receptor alpha 2) were upregulated, consistent with a shift toward anti-inflammatory immune signaling. Chemokine-related paracrine communication appeared differentially modulated: genes encoding Ccl34 (ccl34a.3) and Cxcl8 (cxcl8a) were upregulated, whereas Ccl19 (ccl19a.1) and the chemokine receptor Cxcr3 (cxcr3.1) were downregulated.

Together, these transcriptomic data support the hypothesis that S. costatum induces complex immunomodulatory effects in the regenerating fin blastema, suppressing both inflammation and T-cell–mediated antigen presentation. This shift in immune signaling likely contributes to reduced recruitment and differentiation of osteoclast precursors.

Cluster analysis of genes involved in cell cycle regulation and cytoskeletal organization revealed that SKLT exposure may suppress cell proliferation in fin blastemas at 24 hpa ( Figure 5C ).

Downregulated genes included those associated with proliferation (ttk, cdca8, dlgap5, tacc3, nuf2, ndc80), mitosis (bub1, ccnb1, ccnb3, ccna2, cdk1, tmpoa, mad2l1, foxm1), and cytokinesis (ect2, aurka, aurkb, kif23, racgap1, anln), indicating a broad inhibitory effect of SKLT on cell cycle progression. In contrast, transcriptomic analysis revealed increased expression of genes involved in energy metabolism ( Figure 5D ), suggesting a metabolic shift. Glycolysis-related genes—including eno3, eno1a, gapdhs, aldocb, tpi1b, and gpia—as well as fatty acid oxidation genes (gcdh, asah1a, sdhb, sdhaf2, idh1, sdha, idh3a) were significantly upregulated in SKLT-treated blastemas.

Interestingly, aldh1a2, encoding aldehyde dehydrogenase 1 family member A2, was downregulated in SKLT-treated fish. These molecular signatures are consistent with the reduced regenerative outgrowth observed at later stages in SKLT-treated animals ( Figure 4 ).

Whether the immunomodulatory properties of SKLT could prevent bone loss in a disease setting was investigated in the inducible osteoporosis medaka model Tg(rankl:HSE: CFP)^TG1135, in which bone resorption is triggered by conditional overexpression of RANKL (19). Successful induction is monitored via cyan fluorescent protein (CFP) fluorescence, which is co-expressed with RANKL under the control of a bidirectional heat-shock promoter. Higher CFP intensity corresponds to greater RANKL expression and increased osteoclastic activity, resulting in bone loss.

To evaluate the protective effects of SKLT, medaka larvae were heat-shocked and then exposed to SKLT or vehicle for 6 days. RANKL expression levels were estimated by measuring CFP fluorescence intensity ( Figures 6A, B ). Because the distribution of CFP fluorescence intensity was bimodal or right-skewed, larvae were stratified into two subgroups based on CFP intensity ( Figure 6C ). Bone loss was assessed by measuring the mineralized area of the vertebral column following alizarin red staining. In the subgroup with low CFP fluorescence—indicative of low RANKL expression—no difference in vertebral mineralization was observed between control and SKLT-treated fish ( Figure 6D ). However, in fish with high CFP (and thus high RANKL) expression, exposure to SKLT significantly increased vertebral mineralization ( Figure 6E ).

Additionally, these fish exhibited a greater number of mineralized neural arches ( Figures 6F–H ). These findings suggest that SKLT provides protection against RANKL-induced bone loss in cases of strong RANKL induction, but has no effect when RANKL expression is low. Thus, SKLT appears to counteract the osteoporotic phenotype in a severity-dependent manner.

The ability of SKLT to inhibit osteoclast differentiation was evaluated using the murine macrophage cell line RAW 264.7, which undergoes osteoclastic differentiation upon exposure to RANKL. To assess the effects of SKLT on cell viability and proliferation, RAW 264.7 cells were exposed to various concentrations of SKLT (50, 100, and 200 µg/mL) for 24 and 48 hours. No cytotoxic effects were observed at any concentration ( Figure 7A ). However, significant inhibition of cell proliferation was detected after 1, 3, 5, and 7 days of treatment, even at the lowest concentration (50 µg/mL) ( Figure 7B ). To evaluate osteoclastogenesis, cells were co-treated for 6 days with RANKL (50 ng/mL) and SKLT (50 µg/mL), and stained with DAPI and TRAP ( Figures 7C, D ). No large multinucleated (>2 nuclei) TRAP⁺ cells were detected in the negative control group (no RANKL), while the RANKL-treated group produced an average of 14.3 ± 2.6 osteoclasts per 5.0x field. Co-treatment with SKLT significantly suppressed osteoclast differentiation, reducing the number of multinucleated TRAP⁺ osteoclasts to 0.78 ± 0.67 per 5.0x field ( Figures 7C, E ).

These results confirm that SKLT exerts anti-proliferative and anti-osteoclastogenic effects on murine macrophages without causing cytotoxicity, supporting its cross-species activity and potential as an immunomodulatory inhibitor of osteoclast formation.

The freeze-dried biomass of Skeletonema costatum (Necton S.A., Olhão, Portugal) was macerated with 96% ethanol (Laborspirit Lda, Lisbon, Portugal) at a biomass–solvent ratio of 1 g:40 mL (M/V) by gently stirring at 24 °C for 18 h. The macerate was centrifuged for 5 min at 1,000 × g via an Allegra 6R centrifuge (Beckman Coulter Inc., Brea, USA), and the supernatant was collected. The pellet was washed twice with 96% ethanol, and all the supernatants were pooled and then vacuum filtered sequentially through 0.45 µm and 0.22 µm nylon membranes (Labbox Labware S.L., Barcelona, Spain). The filtrate was concentrated with an RV 10 digital rotary evaporator (IKA-Werke GmbH & Co. KG, Staufen im Breisgau, Germany), with the temperature set at 40 °C and pressure at 178 mbar, until a dense, paste-like extract was obtained. The extraction yield (37.9 ± 2.4%) was calculated from 2 mL aliquots (n = 3) placed under a gentle flow of 99.8% nitrogen until complete evaporation of the solvent.

Zebrafish wild-type line AB and transgenic line Tg(Ola.ctsk:FRT-DsRed-FRT-Cre, myl7:EGFP)^mh201,48, referred to as Tg(ctsk:DsRed) throughout the manuscript, were maintained in a water recirculating system ZebTEC (Tecniplast, Buguggiate, Italy) with the following conditions: temperature 28 ± 0.1 °C, pH 7.5 ± 0.1, conductivity 700 ± 50 μS, ammonia and nitrites at levels below 0.1 mg/L, nitrates lower than 50 mg/L, and a photoperiod of 14:10 h light-dark. The medaka transgenic line Tg(rankl:HSE: CFP)^TG1135, hereafter referred to as Tg(rankl:HSE: CFP), was purchased from the National BioResource Project Medaka (NBRP Medaka) (89) and maintained in a water recirculating system with the following conditions: temperature 27 ± 0.1 °C, pH 7.0 ± 0.1, conductivity 300 ± 100 μS, ammonia and nitrites at levels below 0.1 mg/L, nitrates lower than 50 mg/L and a photoperiod of 14:10 h light–dark. For zebrafish and medaka, system water was prepared by supplementing reverse osmosis water with a salt mixture (Instant Ocean, City, USA) and sodium bicarbonate (Sigma–Aldrich, St. Louis, USA). All the fish were fed commercial dry food Zebrafeed (Sparos Lda, Olhão, Portugal) daily.

The caudal fin of wild-type or transgenic adult zebrafish aged 3–4 months was amputated 1–2 segments anterior to the bifurcation of the most peripheral branching lepidotrichia, as described by Cardeira et al. (2016) (45). For the first experiment, AB wild type zebrafish were sampled at a single time-point, 5 days post-amputation (dpa). After finectomy, fish (n = 14 for SKLT, and n = 32 for CTRL) were placed at 33 ± 1 °C in 3 L plastic containers at a density of 5 fish/L. The fish were exposed to the S. costatum ethanolic fraction (SKLT) at 56 µg/mL or ethanol (CTRL) at 0.1% supplemented in system water for 5 days, to evaluate their mineralogenic performance.

For the second experiments to track osteoclastic cells, transgenic Tg(ctsk:DsRed) fish (n = 18) were amputated and then exposed to 56 µg/mL (SKLT) or 0.1% ethanol (CTRL).

In all experiments, moderate water dynamics and air-water exchanges were facilitated by bubbling the fish tanks with an air pump. Water quality parameters were monitored daily and maintained stable for the duration of the experiment as follows: dissolved oxygen, 7.0 ± 0.5 mg/L; pH, 7.4 ± 0.2; and conductivity, 680 ± 20 μS. Wild-type fish were sacrificed at 120 hours post-amputation (hpa) via lethal anesthesia with 0.6 mM tricaine methanesulfonate (MS-222, pH 7.0; Sigma–Aldrich), immersed for 30 min in 0.03% alizarin red S (AR-S, pH 7.4; Sigma–Aldrich) and washed two times for 5 min with system water. The stained fish were imaged for fin morphometric analysis. Transgenic For the osteoclast-tracking experiment, At 24 hours post-amputation, half of the transgenic Tg(ctsk:DsRed) (n = 9) for each group were given a lethal anesthesia with MS-222, imaged for ctsk signal, then caudal fins were dissected and collected for TRAP staining. The other half (n = 9) were continuously kept in 56 µg/mL (SKLT) or 0.1% ethanol (CTRL), and fluorescence was at different time points (48, 72, 96, 120 and 240 hpa). Fish were immersed for 30 min in 0.2% calcein (pH 7.4; Fluorexon, Sigma–Aldrich), and washed twice in system water for 10 min. After staining, the fish were anesthetized for 5 min in tricaine and imaged.

For bulk the RNA sequencing (RNA-seq) analysis, fin blastemas were collected at 24 hpa, pooled (n = 5, 10 blastemas per pool), and stored at -80 °C until further processing.

Caudal fins were amputated at the level of the caudal peduncle, washed once with 1X phosphate-buffered saline (PBS, pH 7.4) and fixed for 4 h in 4% paraformaldehyde solution (PFA, solubilized in PBS, pH 7.4) at 24 °C. Tartrate-resistant acid phosphatase (TRAP) staining was performed as previously described by Blum & Begemann (2015) (78), and the fins were imaged as described below.

Fins were imaged under an MZ10F fluorescence stereomicroscope (Leica, Wetzlar, Germany) coupled to a DFC7000T color camera (Leica). Bright-field images were collected to assess the progression of fin regeneration. Fluorescence images were collected to assess (i) de novo bone formation in wild-type fish stained with AR-S or calcein and (ii) the involvement of ctsk-expressing cells in transgenic fish labeled with DsRed. Bright-field images were acquired with an exposure time of 1 ms. Fluorescence images were acquired with the filter set ET560/40x - ET630/75m and an exposure time of 600 ms for mCherry, and the filter set ET470/40x - ET525/50m and an exposure time of 80 ms for GFP. Other image parameters were as follows: gamma, 1.00; image format, 1920×1440 pixels; and binning, 1×1. The fluorescence images were analyzed using ImageJ software version 2.0.0-rc-69/1.52p and processed with the ZFBONE toolset for caudal fin morphometrics (90). Fin regeneration and mineralization were assessed following the method described by Cardeira et al. (2016) (45) by calculating the regenerated area (REG), the stump width (STU), the mineralized area (MIN), and the average width of the rays before the amputation (RAYs). Fin ray patterning was also assessed by calculating the average ray width ratio ( W ¯ ) and the average bifurcation ratio ( B t ¯ ), as shown in Supplementary Equations a, b . For the quantification of the ctsk signal in transgenic fish, the DsRed-positive area was measured using a color threshold on fluorescence images and normalized with REG/STU. To quantify the TRAP signal, TRAP-positive areas were measured using a color threshold on bright-field images and subsequently normalized with REG/STU.

Total RNA was extracted from pools of blastemas at 24 hpa (n = 5) using NZYol (NZYTech, Lisbon, Portugal) and quantified using a NanoDrop OneC spectrophotometer (Thermo Fisher Scientific, Waltham, USA). The RNA integrity was confirmed using an Experion Automated Electrophoresis system (Bio-Rad, Hercules, USA). Only RNA with an RNA integrity number (RIN) greater than 7 was used.

Bulk RNA sequencing was outsourced to STABVIDA Lda (Caparica, Portugal). DNA libraries were constructed using a Stranded mRNA Library Preparation Kit (STABVIDA) and sequenced on a NovaSeq platform (Illumina, San Diego, USA) to generate 150 bp paired-end sequencing reads. The raw sequence data were processed using CLC Genomics Workbench 12.0.3. Trimming was performed in 3 steps: quality trimming on the basis of quality scores (error probability of 0.01), ambiguity trimming (ambiguous limit of 2 nucleotides), and length trimming to discard reads shorter than 30 nucleotides. The quality-checked sequencing reads were mapped against the zebrafish reference genome GRCz11 (GCF_000002035.6) using length fraction and similarity fraction equal to 0.8. The TPM (transcripts per million) and RPKM (reads per kilobase of transcript per million mapped reads) values were then determined from the mapped data. Differential expression analysis was performed with the multifactorial EdgeR method in R. Overall, the gene expression results were represented with a clustering heatmap of gene expression, with a principal component analysis (PCA) plotting the amount of variance explained by the two principal components and a volcano plot representing overall gene expression with a fold change on the x-axis and significance of expression on the y-axis ( Supplementary Figure S2 ). These overall representations of gene expression were performed to assess variation patterns in the gene expression dataset and identify outlier samples for quality control. For functional annotation enrichment analysis, the lists of upregulated and downregulated genes were analyzed with the online resource Database for Annotation, Visualization and Integrated Discovery (DAVID) (91, 92) for Gene Ontology–Biological Processes (GO: BP) with FDR< 0.01 and enriched biological processes.

Eggs of the medaka line Tg(rankl:HSE: CFP) were produced following an in-house breeding program and maintained in Petri dishes with 40 mL of system water supplemented with 0.0002% (w/v) methylene blue until 8 days post-fertilization (dpf). At 8 dpf, the hatched larvae were incubated at 39 °C for 2 h (heat shock) and screened for CFP (cyan fluorescent protein) signal using a fluorescence microscopy (see parameters below). CFP-positive fish were distributed into a 6-well plate at a density of 5 larvae/well. Each well was filled with 10 mL of system water supplemented with SKLT at 56 µg/mL or 0.1% ethanol (vehicle). The treatment was renewed 100% daily. Six days after heat shock, the larvae (n = 17 for CTRL, and n = 12 for SKLT) were sacrificed with a lethal dose of tricaine (see above), stained with AR-S and imaged as described above. Bright-field and fluorescence images were collected for each fish. Bright-field images were acquired with an exposure time of 2 ms. AR-S fluorescence images were acquired using the filter set ET560/40x - ET630/75m and an exposure time of 700 ms. CFP fluorescence images were acquired using the filter set ET436/20x - ET480/40m and an exposure time of 200 ms. Other image parameters were as follows: gamma 1.00, image format 1920×1440 pixels, and binning 1×1. Images were analyzed using ImageJ using two ad hoc macros ( Supplementary Files 1 , 2 ). CFP fluorescence images were transformed into 8-bit images, and the CFP-positive area was measured using a color threshold (min intensity of 7 and max intensity of 255). The pixel mean intensity inside the CFP-positive area was calculated and used as a proxy for CPF fluorescence intensity. AR-S fluorescence images were transformed into 8-bit images, and AR-S positive areas were measured using a color threshold (min intensity 5 and max intensity 255). The AR-S positive area was normalized using the total body area (determined manually from bright-field images) to correct for inter-specimen size variation. The AR-S positive area in the abdominal and caudal vertebrae was used as a proxy for mineralization of the vertebral column. The nomenclature proposed by Di Biagio et al. (2022) (83) was used to identify vertebrae from 4-29. The number of mineralized neural arches was manually counted in each fish from the AR-S images. Following the collection of the data, the frequency distribution of the CFP intensity and tested data normality was determined through an Anderson–Darling test (p< 0.05). Normality was not met in the control group (p = 0.0035), but CFP intensity was normally distributed in SKLT-exposed fish (p = 0.220). The data distribution for CFP intensity was right skewed (median< mean) for the control group and bimodal for the SKLT group. These findings implied that clustering was necessary to proceed with the statistical analysis. The fish were therefore grouped according to CFP intensity into “High CFP” (mean pixel intensity > 50) and “Low CFP” (mean pixel intensity< 50) for further analysis.

RAW 264.7 cells were cultured in 10 cm cell culture dishes with Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Sigma–Aldrich), 1% of penicillin–streptomycin (100x), 1% of 200 mM L-glutamine, and 0.2% of 250 µg/mL amphotericin B at 37 °C in a humidified atmosphere containing 5% CO₂. PreconfluenT-cell cultures were sub-cultured at a ratio of 1:4 every other day using trypsin-EDTA solution (0.2% trypsin, 1.1 mM EDTA: prepared in 1x PBS, pH 7.4). All cell culture reagents were obtained from GIBCO-Thermo Fisher Scientific, unless otherwise stated.

An LDH cytotoxicity assay kit and a XTT-cell proliferation assay kit (Canvax Biotech, Córdoba, Spain) were used to evaluate the effects of SKLT on cellular toxicity and proliferation, respectively. For cytotoxicity, RAW 264.7 cells were seeded in a 96-well plate at 1.0 × 10⁴ cells/well (n = 4) in 100 µL of culture medium supplemented with either the SKLT at 50, 75, 100, 200 µg/mL or 0.1% ethanol (vehicle) and cultured for 24 and 48 h. Relative cytotoxicity was calculated as a percentage of LDH lysis control. For cell proliferation, RAW 264.7 cells were seeded in a 96-well plate at 500 cells/well (n = 6) in 100 µL of culture medium supplemented with either the SKLT at 50, 75, 100, 200 µg/mL or vehicle and cultured for 1, 3, 5 or 7 days. The culture medium was renewed every other day. Relative cell proliferation was calculated as a percentage of the negative control.

RAW 264.7 cells were seeded in 12-well plates at 1.5 × 10⁴ cells/well in 1 mL of culture medium (as described above, n = 9) supplemented with either SKLT at 50 µg/ml or 0.1% ethanol (vehicle) and treated for 6 days with 50 ng/mL RANKL (Preprotech, London, UK) prepared in 0.1% bovine serum albumin (BSA) to induce their differentiation into multinucleated osteoclasts. The experimental conditions were as follows: undifferentiated control (0.1% ethanol); differentiated control (50 ng/mL RANKL + 0.1% ethanol); differentiated cells exposed to SKLT (50 ng/mL RANKL + 50 µg/mL SKLT in 0.1% ethanol). The culture medium was freshly prepared and replaced daily. Osteoclast differentiation was assessed by counting the number of tartrate-resistant acid phosphatase (TRAP)-positive cells and the number of nuclei following 4′,6-diamidino-2-phenylindole (DAPI) staining. For this purpose, the cells were fixed in 4% PFA at 24 °C (RT) for 10 min and stained according to the protocol of Blum & Begemann (2015) (78). The cells were subsequently stained with DAPI for nuclear detection and counting. The cells were imaged using an Axio Vert. A1 inverted microscope (ZEISS, Jena, Germany) coupled with (i) an Axiocam 202 monocolor camera (ZEISS) for DAPI images, or (ii) a VWR VisiCam 5 Plus (VWR, Radnor, USA) for bright-field TRAP images. The number of multinucleated osteoclasts per field (5.0X magnification; n = 9) was calculated considering only TRAP+ cells with at least two nuclei.

For all the experiments, normality was tested with a D’Agostino–Pearson omnibus normality test or with an Anderson–Darling test (p< 0.05). Homoscedasticity was tested through the Brown–Forsythe test (p< 0.05). When the distribution of the data of all the experimental groups resulted in normal and homogeneous, statistical differences between the control and the treated groups were tested with either an unpaired t test or one-way ANOVA, followed by Dunnett’s multiple comparison test (p< 0.05). If the distribution of the data of any of the experimental conditions resulted in non-normal or nonhomogeneous, significant differences between the control and the treatments were tested with a Mann–Whitney test or a nonparametric test followed by Dunn’s multiple comparison test (p< 0.05). Statistical analyses were performed using Prism version 9.00 (GraphPad Software Inc., La Jolla, United States).