Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 300 NMR spectrometer (Bruker Corporation). Reversed phase high-performance liquid chromatography (HPLC)-mass spectrometry (MS) analysis was performed on a Waters Alliance 2695 LC (Waters Corporation) coupled to a Quattro Ultima triple quadrupole MS (Waters Corporation) using the same separation conditions as described previously (46). The separation conditions were follow: Chromatogram column, XBridge™ column (4.6×150 mm, 5 µm); column temperature, 20°C; and injection volume, 1 µl. Elution was performed at a flow rate of 1 ml/min, using as the mobile phase a mixture of water (A) and acetonitrile (B). The samples were eluted using the following gradient: 0 min, 98.0% A; 0-8 min, linear increase to 100% B; 100% A held for 2 min, 11-12 min, return to 98.0% A. The final optimization of ESI source operation parameters were a nitrogen gas flow of 11 l/min, nebulizer of 30 psi, capillary voltage of ±2.4 kV and a drying gas temperature of 250°C (N2). The Quattro Ultima triple quadrupole MS operated in multiple reaction monitoring mode with a resolution of 0.7 m/z. Electrospray ionization (positive mode) and photodiode array detection at 254 nm were performed.

Different batches of A. annua (Luparte^®) were determined by the Department of Pharmaceutical Biology, Johannes Gutenberg University (Mainz, Germany). This blinded approach was used as a quality measure to independently guarantee the presence of artemisinin in Luparte^®. Each batch of A. annua powder (2×10 g) was extracted with two different polarity solvents, methylene chloride and methanol, at room temperature for 24 h. The extract was concentrated in vacuo to obtain a residue of 0.8 and 1.2 g for methylene chloride and methanol, respectively. Each extract was transferred to a small vial and kept dry at room temperature. Small crystals were formed, which were collected and washed with n-hexane and methylene chloride to remove extract residues. Then, ¹H and ¹³C NMR analyses were performed.

The animals were treated between 2010 and 2017. Briefly, 16 dogs (10 males and 6 females) and 4 cats (2 females, 1 male and 1 N/A) were treated with A. annua. The mean age of the dogs was 10.31 years and the mean age of the cats was 12.25 years. The mean weight of the dogs was 30.125 kg and the mean weight of the cats was 5.10 kg. Tissue samples were taken from all animals and initially fixed at room temperature (~22°C) in 4% formaldehyde solution for 48 h. Following fixation, the organs were trimmed, processed and embedded in paraffin wax at approximately 60°C. Sections were cut to a thickness of 4 µm and routinely stained with hematoxylin and eosin at room temperature. Histopathological examination of the hematoxylin and eosin-stained sections was performed under a light microscope in immersion oil with a Zeiss Axioskop. After routine pathological diagnosis, the paraffin blocks were used as excess material for subsequent immunohistochemical studies. Furthermore, the histological hematoxylin and eosin stained tumor sections were assessed for the presence of the tumor-infiltrating lymphocytes (TILs). The clinical data are presented in Tables I and II. The owners of the animals provided written informed consent for this retrospective study. The signed written consent forms are deposited at the Department of Pharmaceutical Biology, Johannes Gutenberg University (Mainz, Germany) and can be inspected upon reasonable demand.

The serum iron content was determined following collection of blood samples; a normal range is between 140 and 170 µg/dl (47). After initial iron content determination using ferrozine color test (48), the tumors were surgically removed with safety margins by a standard protocol after visual inspection and where required after lung radiography. Between blood collection and determination of the serum iron results from the clinical diagnosis laboratory, iron was administered orally [Ferrosanol^® capsules (100 mg/capsule or 40 mg/capsule)]. The dosage was 100 mg/30 kg BW twice daily mixed into vegetable-poor food to avoid iron binding plant molecules, such as phytane, oxalates and/or phosphates. Alternatively, iron was subcutaneously injected (Myofer^®/Ursoferran^®; 100 mg/ml) at a dose of 100 mg/10 kg BW daily, until the serum iron results were obtained back from the clinical laboratory. The iron substitution was individually continued until an iron content of 250±30 µg/dl (~43±5 µmol/l) was stably reached in blood serum. From the fourth day of the initial iron substitution onward, the animals were orally treated with Luparte^® capsules at a dosage of 1,400 mg/m² body surface divided to three fractions per day (Table III). The capsules were administered 1-2 h before the next meal. The serum iron content was regularly monitored and where necessary adapted to 250±30 µg/dl.

The treatment details and the follow-up information, including survival times, have been recorded in the clinical documentation files. All ethical issues have been discussed prior to evaluation of clinical and experimental data with the Regierungspräsidium (Government Presidium) Freiburg (Germany). According to the design of this retrospective study, the Regierungspräsidium (Government Presidium) Freiburg gave written permission for the present study (Az. 35-9185.81/1, dated from February 4th, 2019).

The D-GBM cell line was generated by Dr George Stoica (Department of Veterinary Pathobiology, Texas A&M University, College Station, TX, USA) from a brain tumor of an 8-year-old male Boxer dog (49). The D-GBM cells used in the present study were obtained from Dr Pablo Steinberg (Institute for Food Toxicology and Analytical Chemistry, University of Veterinary Medicine Hannover, Hannover, Germany), who obtained it from Dr G. Stoica.

The histiocytic DH82 cell line was generated by Dr M.L. Wellman (Department of Veterinary Pathobiology, College of Veterinary Medicine, Ohio State University, CO, USA) (50). The DH82 cell line used in the present study was also obtained from Dr Pablo Steinberg, who purchased it from the European Tissue Culture Collection (catalog no. 94062922).

A panel of 54 human tumor cell lines of the Developmental Therapeutics Program (DTP) of the National Cancer Institute consisted of cell lines derived from leukemia, melanoma, non-small cell lung cancer, colon cancer, renal cancer, ovarian cancer, breast cancer and prostate carcinoma, as well as central nervous system tumors (51). Previously, the cytotoxicity of cells treated with artemisinin for 48 h was evaluated using a sulforho-damine B assay (52). The log₁₀ IC₅₀ values for artemisinin and mRNA expressions for Ki-67 and TfR were available publicly in the DTP database (https://dtp.cancer.gov/default.htm).

A resazurin reduction assay (Promega Corporation) was performed as previously described (53) to assess the cytotoxicity of artemisinin, artesunate and dihydro-artemisinin towards the dog tumor cell lines DH82 and DGBM. The concept of the assay is based on the metabolic reduction of the non-fluorescent dye by living cells to the strongly-fluorescent dye resorufin (54). IC₅₀ values were calculated from dose-response curves using a non-linear regression analysis tool in GraphPad Prism 7 software (GraphPad Prism, Inc.). All IC₅₀ values are expressed as the mean ± standard deviation. Each assay was performed thrice independently, with six replicates each.

A total of 17 tumor tissues representing different tumor types in pets were collected to evaluate the expression of Ki-67 and TfR in animal tumors (Table V). Briefly, tumor tissues were obtained from 15 dogs (7 males and 8 females) and 2 cats (both female). The mean age of the dogs was 9.33 years and the mean age of the cats was 4 years. The mean weight of the dogs was 27.266 kg and the mean weight of the cats was 4.00 kg. All the chosen animals had been kept in pet animal housing conditions. The immuno-histochemical staining of tumor tissues was performed as described previously (55). The slides were washed twice with xylene (98.5% xylene for 5 min each at room temperature) to remove paraffin. Then, sample tissues were rehydrated through graded washes with isopropanol in water. Heat-induced epitope retrieval was performed using a pressure cooker as a heating device. Ultra-vision protein block and UltraVision Hydrogen Peroxide Block (catalog nos. TA-060-UB and TA-060-H2O2Q, respectively; Thermo Fisher Scientific, Inc.) were added to block endogenous proteins and endogenous peroxidase activity, respectively, to avoid non-specific background staining. Overnight incubation at 4°C was performed following the addition of monoclonal primary antibodies. For the detection of Ki-67, the SP6 clone (catalog no. ab16667; Abcam) was used at a dilution of 1:200. For the determination of TfR expression, the H68.4 clone (catalog. no. 136800; Thermo Fisher Scientific, Inc.) was applied at a dilution of 1:100. Subsequently, horseradish peroxidase-labeled polymers conjugated with secondary antibodies specific for both mouse and rabbit primary antibodies were added (catalog nos. TL-060-QPH and TL-060-QPB; Thermo Fisher Scientific, Inc.) at room temperature for 1 h, according to the manufacturer's protocol. The final staining reaction was performed with diaminobenzidine and slides were counterstained with hematoxylin for 3 min at room temperature.

The immunostaining of DH82 and DGBM cell lines was performed without the rehydration or epitope retrieval steps. Briefly, six cover slips were placed in a 6-well plate, then 5×10⁵ cells/well were seeded over the cover slips and incubated in humidified 5% CO₂ atmosphere at 37°C overnight to let the cells attach to coverslips. Subsequently, the cells were fixed in 4% paraformaldehyde for 15 min at room temperature, and the blocking and staining steps were performed as described for the tissue section. However, the Ki-67 and TfR antibodies were diluted at 1:500 and 1:1,000, respectively. The immunostained slides were scanned using Pannoramic Desk (3DHISTECH Ltd.) and the amount of Ki-67- or TfR-expressing cells was quantified by Pannoramic Viewer software version 1.15 (3DHISTECH Ltd.).

Statistical analysis. Significance values and correlation coefficients were calculated using Pearson's correlation coefficient and Fisher's exact test with the WinSTAT software program version 2012.1 (www.winstat.com/). Linear regression was performed using Excel 2016 (Microsoft Corporation) for the log₁₀ IC₅₀ values for artemisinin and mRNA expressions for Ki-67 and TfR. P<0.05 was considered to indicate a statistically significant difference.

¹H NMR spectrum of a colorless crystal contained signals typical of artemisinin. The spectrum displayed signals for three methyl groups: Two secondary at δ_H 1.00 (d, J=6.5, H-14), 1.17 (d, J=7.2, H-13) and one tertiary methyl signal at δ_H 1.40 (s, H-15). Additionally, two aliphatic methylene signals were observed at δ_H 2.12 (brddd, J=14.9,4.2,2.6) and 2.40 (ddd, J=14.9, 13.1,4.1) for H-3a and H-3b, respectively. A characteristic downfield singlet signal at δ_H 6.01 indicated the presence of oxygenated proton for H-5. As presented in Fig. 1B, The ¹³C NMR spectrum displayed 15 major carbon signals, including significant carbons signals confirming the presence for artemisinin as the following: δ_C 173.3 (s, C-12), 105.2 (s, C-4), 94.3 (d, C-5) and 79.5 (s, C-6) (56).

Reversed phase HPLC-MS was used to quantitatively analyze the presence of artemisinin. Fig. 1C presents the chro-matogram obtained after dissolving 2 mg/ml MeOH extract, which indicated that artemisinin represents 3.46% of the A. annua (Luparte^®) extract.

In total, 23 dogs and 8 cats were included in the present study. Among these, 20 were treated with A. annua in addition to their standard therapy (Table I) and 11 were subjected to standard treatment alone (Table II). The A. annua-treated group consisted of 10 carcinomas (8 dogs, 2 cats) and 10 sarcomas (8 dogs, 2 cats). All animals were treated between 2010 and 2017 (Table I). In the group of animals without A. annua treatment, 5 animals presented with carcinoma (3 dogs, 2 cats) and 6 animals had sarcoma (4 dogs, 2 cats) (Table II). The blood iron content and the survival times of all pets were recorded.

The effect of A. annua treatment on overall survival time was assessed. In total, 13 A. annua-treated animals and 11 non-treated animals exhibited a survival time of <18 months following therapy. Whereas, 7 animals in the A. annua-treated group had a survival time of >18 months after A. annua treatment, but no animals in the non-treated group survived >18 months. This difference in survival times between the groups was statistically significant (P=0.033; two-tailed Fisher's exact test; Table IV).

Using a second collection of 17 tumors, the histology of the tumors was determined by hematoxylin and eosin staining of formalin-fixed and paraffin-embedded tumor sections. Representative photographs of different tumor histology are presented in Fig. 2, including a tubulopapillary breast adenocarcinoma, soft tissue sarcoma, squamous cell carcinoma and neuroendocrine carcinoma.

Furthermore, immunohistochemical analyses were performed. The membrane-bound expression of TfR (CD71) and the nuclear expression of the proliferation marker Ki-67 were determined. The percentage of stained cells was quantified using a computer-based quantification system. Six different areas were selected from each tumor section to provide representative expression values. The percentage of TfR-expressing cells ranged between 43.5 (±37.9%) and 99.2% (±0.2%), whereas the percentage of Ki-67-positive cells was in a range between 6.6 (±1.8%) and 85.7% (±4.7%) (Fig. 3A and B). As tumor heterogeneity plays an important role in the response to anticancer therapy, the present study also focused on heterogeneous and homogeneous staining patterns among the different tumor biopsies. Certain tumors were very heterogeneous in the immunohistochemical staining, while others revealed uniform staining. This can be seen in the standard deviations, which ranged between 0.1 (cases no. 3, 8 and 13) and 37.9% (case 10) for TfR, and between 1.8 (case 17) and 16.3% (case 8) for Ki-67 (Table V). Subsequently, the correlation between the expression of TfR and Ki-67 with the set of tumor biopsies was investigated, which revealed a statistically significant correlation (P=0.025; r=0.469; Fig. 3C). No significant correlation was identified between blood iron content and survival time of the animals (P=0.386; r=-0.062; Fig. 3D). Notable, the association between survival time and A. annua treatment was significant. Pets treated with standard therapy plus A. annua has significantly longer survival times compared with those treated with standard therapy alone (P=0.033; Fig. 3E).

As the direct association between TfR and Ki-67 expression and the cytotoxicity of artemisinin could not be assessed in these tumor samples, the TfR and Ki-67 microarray-based mRNA expression in 54 human tumor cell lines was analyzed compared with the log₁₀ IC₅₀ values, as determined by sulforhodamine assay. Both TfR and Ki-67 expression were significantly negatively correlated with the cytotoxicity of artemisinin in these cell lines (P<0.05; r<-0.20), which indicates a higher expression of these two markers was associated with a higher sensitivity of the cell lines to arte-misinin (Table VI). The expression levels of TfR and Ki-67 were positively correlated (P=0.008; r=0.317). This panel of 54 human tumor lines consisted of cell lines derived from eight tumor types (leukemia, melanoma and brain tumor, and carcinoma of colon, breast, ovary, kidney and prostate). The number of cell lines of each tumor type (n=1-9) was too low to reveal significant correlations except for the three following results. In lung cancer cell lines, the log₁₀ IC₅₀ values of arte-misinin were inversely correlated with the Ki-67 expression (P=0.009; r=-0.759). In melanoma cell lines, the expression levels of Ki-67 and TfR were significantly correlated (P=0.010; r=0.787). In renal cancer cell lines, the expression levels of Ki-67 and TfR were also significantly correlated (P=0.019; r=0.781) (data not shown). Since all other associations between artemisinin response and the expression of Ki-67 and TfR were not statistically significant, reliable conclusions on the role of Ki-67 and TfR may be drawn from the cell line panel as a whole, but not from tumor type-specific subsets.

Representative images of immunostaining for TfR expression are presented in Fig. 4A-D. Strong TfR expression was identified in biopsies of tubulopapillary breast adenocarci-noma, soft tissue sarcoma and squamous cell carcinoma. By contrast, the negative control did not demonstrate any reactivity, indicating the specificity of the immunostaining procedure. Furthermore, examples of Ki-67 expression in tubulopapillary breast adenocarcinoma, soft tissue sarcoma, squamous cell carcinoma and neuroendocrine carcinoma are presented in Fig. 4E-H. Furthermore, assessment of the presence of TILs in hematoxylin and eosin stained tumor tissues revealed an absence of TILs in all tumor slides.

For comparison, the Ki-67 and TfR expression was analyzed in two dog cell lines. DH82 histiocytic sarcoma and DGMB glioblastoma cells were immunostained for both markers. Indeed, Ki-67 and TfR were overexpressed in both cell lines as presented in Fig. 5. Furthermore, the present study investigated the cytotoxicity of artemisinin, artesunate and dihydroartemis-inin towards the dog cell lines. As expected, artemisinin and its derivatives were also active in these tumor cell lines. The IC₅₀ values for artemisinin, artesunate and dihydroartemisinin towards DH82 cells and DGBM cells are presented in Fig. 6.