Heterogeneous in vitro effects of doxorubicin on gene expression in primary human liposarcoma cultures

Background Doxorubicin is considered one of the most potent established chemotherapeutics in the treatment of liposarcoma; however, the response rates usually below 30%, are still disappointing. This study was performed to identify gene expression changes in liposarcoma after doxorubicin treatment. Methods Cells of 19 primary human liposarcoma were harvested intraoperatively and brought into cell culture. Cells were incubated with doxorubicin for 24 h, RNA was isolated and differential gene expression was analysed by the microarray technique. Results A variety of genes involved in apoptosis were up and down regulated in different samples revealing a heterogeneous expression pattern of the 19 primary tumor cell cultures in response to doxorubicin treatment. However, more than 50% of the samples showed up-regulation of pro-apoptotic genes such as TRAIL Receptor2, CDKN1A, GADD45A, FAS, CD40, PAWR, NFKBIA, IER3, PSEN1, RIPK2, and CD44. The anti-apoptotic genes TNFAIP3, PEA15, Bcl2A1, NGFB, and BIRC3 were also up-regulated. The pro-apoptotic CD14, TIA1, and ITGB2 were down-regulated in more than 50% of the tumor cultures after treatment with doxorubicin, as was the antiapoptotic YWHAH. Conclusion Despite a correlation of the number of differentially regulated genes to the tumor grading and to a lesser extent histological subtype, the expression patterns varied strongly; however, especially among high grade tumors the responses of selected apoptosis genes were similar. The predescribed low clinical response rates of low grade liposarcoma to doxorubicin correspond to our results with only little changes on gene expression level and also divergent findings concerning the up- and down-regulation of single genes in the different sarcoma samples.


Background
Together with malignant fibrous histiocytoma (not otherwise specified sarcoma, NOS), liposarcoma represents the most common entity of soft tissue sarcomas and accounts for approximately 20% of sarcomas in adults [1][2][3]. Although surgery and radiation therapy could achieve good results concerning local control, distant metastatic disease remains a therapeutic dilemma limiting survival [4,5]. With a maximum response rate of approximately 20% the effects of cytostatics on liposarcoma are still disappointing [6][7][8]. The most favoured chemotherapeutics for treatment of advanced soft tissue sarcoma, including liposarcoma, are ifosfamide and doxorubicin, but the data for ifosfamide differ with respect to improvement of local control and survival [9][10][11]. Although meta-analysis of 14 randomised trials found that doxorubicin treatment was associated with a 10% improvement of recurrence free survival, the overall survival could not be improved [12][13][14]. In carcinomas, multiple mechanisms of drug resistance on the molecular level have been characterized [15,16] including over-expression of p53 [17][18][19][20], MDR1 (multidrug resistance gene 1) [20][21][22], MRP1 (multidrug resistance-associated protein), the induction of DNA repair [20] and many others involving tumor suppressor genes, oncogenes, cell cycle regulators, transcription factors, growth factor receptors, and cell death regulators.
Only little is known about the molecular basis of drug resistance in soft tissue sarcomas and studies on the effect of cytostatics on gene expression, especially in liposarcomas [23][24][25][26][27], are rare. Comprehensive knowledge of the differential expression patterns induced by cytotoxic drugs may be useful for examining the molecular basis of drug effects and also drug resistance. Because of the limited comparability of established purchasable sarcoma cell lines to in vivo tumors, we primarily harvested liposarcoma cells from resection specimens, incubated the cultured cells with doxorubicin and evaluated the changes in gene expression with a focus on genes related to apoptotic pathways. To the authors' knowledge, to date there are no studies that examined the effects of doxorubicin on primary human liposarcoma on a molecular basis.

Methods
Primary human liposarcoma tumor samples of at least 1 cm 3 were harvested intraoperatively from patients undergoing resection of an already diagnosed liposarcoma and immediately processed under sterile conditions. Seven atypical lipomas (low grade sarcomas), four dedifferentiated, four pleomorphic, three myxoid/round cell, and one myxoid liposarcoma were included. The grading of the tumors ranged from GI to GIII (GI: 4, GII: 8, GIII: 7). The probes were derived from primary tumors in 12, from recurrent tumors in six, and from metastasis in one case. Nineteen primary human liposarcoma cultures were isolated by dissecting the tumor and digesting the minced samples enzymatically with 10 ml each of collagenase and dispase (10 mg/ml). The single cell suspension was depleted of red blood cells and cellular debris by centrifugation through a Ficoll-Hypaque density gradient. Liposarcoma cells were diluted and cultured during the whole experiment with Leibovitz's L-15 medium, supplemented with 2.0 mM glutamine and 10% fetal bovine serum in a humidified atmosphere in free air exchange with atmospheric air. Cells were seeded at a density of 2 × 10 6 in 25 cm 2 flasks; 24 h later, after having grown to a subconfluent layer, cell cultures were incubated with doxorubicin (0.5 μg/ml) for 24 h and equal volume of PBS as control [28][29][30].

Oligonucleotide microarray analysis
For microarray analyses we used the Affymetrix Gene Chip platform employing a standard protocol for sample preparation and microarray hybridization that has been described in detail previously [26,31]. Briefly, total RNA was converted into double-stranded cDNA using an oligodeoxythymidine primer containing the T7 RNA polymerase binding site (5'-GCATTAGCGGCCGCGAAAT-TAATACGACTCACTATAGGGAGA -(dT) 21 V-3') (MWG Biotech, Ebersberg, Germany) for first strand synthesis. After generation of double-stranded cDNA from the firststrand cDNA, biotinylated cRNA was synthesized by in vitro transcription using the BioArray High Yield RNA Transcript Labeling Kit (Enzo Diagnostics, New York, USA). Labeled cRNA was purified on RNeasy columns (Qiagen, Hilden, Germany) and fragmented and hybridized to HG-U133A microarrays (Affymetrix, Santa Clara, USA). The arrays were washed and stained according to the manufacturer's recommendation and finally scanned in a GeneArray scanner 2500 (Agilent, Santa Clara, USA).
Array images were processed to determine signals and detection calls (Present, Absent, Marginal) for each probeset using the Affymetrix Microarray Suite 5.0 software (MAS 5.0; statistical algorithm). The clustering was performed unsupervised. Pairwise comparisons of treated versus control samples were carried out with MAS 5.0, which calculates the significance (change p-value) of each change in gene expression based on a Wilcoxon ranking test. To limit the number of false positives, we restricted further target identification to those probesets, which received at least one present detection call in the treated/ control pair. Each single microarray analysis was derived from one cell culture. The treated cells were compared to the control. Probesets exhibiting a signal log2 ratio > 1.0 and a change p-value < 0.004 or a signal log2 ratio < -1.0 and a change p-value > 0.996 (corresponding to 2-fold up-or down-regulation) were identified by filtering using the Affymetrix Data Mining Tool 3.0 (Table 1). Additionally unsupervised clustering was performed between the 19 control tumor samples. Expression changes were correlated to the grading and the histological sub-entity of the tumors. Only tumor samples were included in the final analysis whose gross sections were diagnosed as liposarcomas by an experienced soft tissue pathologist.

Real-time PCR for microarray data validation
Microarray data validation was performed for selected gene products with relevant up-regulation in more than ten out of the 19 liposarcoma probes (CD40, CDKN1A, FAS, GADD45A, NFKBIA, PAWR, TNFAIP3, and TNFRSF10B) or relevant down-regulation in at least ten probes (YWHaH, PPP3CA, and ITGB2). CD14 and TIA were not tested because no high quality PCR assays were purchasable.
Total RNA (2 μg) was reverse transcribed using the High Capacity cDNA Archive Kit (Applied Biosystems). Realtime PCR was done with a 7900HT SDS system (Applied Biosystems) in 20 μl reaction volume containing 1× Master Mix, 1 μl assay and cDNA equivalent to 2 ng total RNA. All reagents and realtime PCR assays (m1 type only; CD40_Hs00386848_1, CDKN1A_Hs00355782_m1, FAS _Hs00236330_m1, GADD45A_Hs00169255_m1, ITGB2 _Hs00164957_m1, NFKBIA_Hs00153283_m1, PAWR_H s00169332_m1, PPP3CA_Hs00174223_m1, TNFAIP3_H s00234713_m1, TNFRSF10B_Hs00366272_m1, YWHaH _Hs00607046_m1) used were purchased from Applied Biosystems. Reactions were performed in duplicates and analysed by the deltadeltaCT method. Human GAPD was used for normalization. Pearson correlation of the microarray and the rtPCR data was calculated using SPSS Version 15.0 for Windows (SPSS Inc., Chicago, USA).

Results
Sixteen tumors were located at the extremities, three at the trunk. The tumors diameters ranged from 1 cm to 38,5 cm. Twelve primary cultures were harvested from primary tumors, 6 from local recurrences, and 1 from a metastatic tumor. The majority of tumors were high grade (G2-3); only 4 were diagnosed low grade (G1). Atypical lipoma or highly differentiated liposarcoma was the most common histological subentity (n = 7) followed by dedifferentiated (n = 4) and pleomorphic (n = 4), myxoid/rundcell (n = 3) and myxoid (n = 1) liposarcoma. All low grade tumors were diagnosed as atypical lipoma, whereas 3 atypical lipomas were categorized as G2 tumors because of aggressive growth or localized areas of dedifferentiation. The other subentities were all categorized as high grade tumors (G2-3). A summary of the patients' data is given in table 1.
Hierarchical clustering of expression profiles from untreated samples did not clearly separate according to histological subtype or grading, but revealed two major branches, which showed some enrichment for G3 (left) and G2 tumors (right) ( fig. 1).
According to the number of differentially expressed genes n after incubation with doxorubicin, the 19 liposarcomas were categorized into high (n > 2000), intermediate (100 < n < 1000) and low responders (n < 100). All poorly differentiated (G3) tumors were high responders; G2 tumors were predominantly intermediate responders; and, most of the G1 tumors were low responders. A heatmap of all differentially expressed genes after 24 h of doxorubicin tratment shows heterogeneous response patterns ( fig. 2).
Correlating the overall expression changes with the histological subtype showed that most of the atypical lipomas were low responders (5 low, 1 medium, 1 high), dedifferentiated were predominantly medium responders (3 medium, 1 high), most myxoid/roundcell and myxoid liposarcomas as well as the pleomorphic liposarcomas were high responders (3 high, 1 medium each). The high grade sarcomas (G3) clustered closely together.
The alteration of gene expression related to apoptotic pathways correlated to the categorisation given above. Low responders also did also not respond with relevant gene expression changes of "apoptosis genes" whereas the high responders showed a significantly different gene expression profile concerning apoptosis related genes compared to the untreated control. In all, we found 464 genes with expression changes that are related to apoptotic pathways. The single genes that were differentially expressed in the medium and high responder group only partly overlapped with the low responder group. The heterogeneity of the response patterns of apoptotis related genes is illustrated in figure 3.
Although the diversity of changes in gene expression was large, some apoptosis related genes showed similar expression changes in the tumor samples, especially the high grade tumors (G2 and G3) or high responders. Figure  4 focus on expression changes in these genes. The apoptosis related genes most often affected by doxorubicin treatment are mentioned below. Due to their large number, we only refer to the genes that were differentially (up-or down-) regulated in more than 50% of the probes (samples). Some of the genes that were found up-regulated in the majority of the probes (tumors) could be found down-regulated in some other samples and vice versa. The heatmaps provided illustrate the similarity of the expression of these selected genes in correlation to responder group, grading, and histological subtype ( fig. 4. The intrinsic and the extrinsic apoptotic pathway, as well as transcription factors and genes, so far only marginally associated with apoptotic pathways. A summary of the genes and their expression changes is given in table 2 and 3. The results of the microarray analysis were validated using rtPCR. The general Pearson correlation of the expression changes measured in all selected genes in the tumor probes was high (0.913). The correlation coefficients for the single candidate genes are given in table 4.

Discussion
Gene expression profiling has already been helpful in categorizing distinct subtypes of sarcomas by profile cluster- Interestingly the myxoid and the myxoid/round cell liposarcomas clustered together in figure 2 and 3 except from the tumor 7 that was pre-treated with chemotherapy. If that is a coincidence or may be interpreted as a kind of selectional process that could have eliminated the high grade parts of the tumor leaving the residual to cluster closer to the low grade sarcomas can not be determined.
The predescribed low clinical response rates of low grade sarcoma correlate to our findings that low grade liposarcoma, especially atypical lipoma, showed almost no response to doxorubicin on gene expression level [61]. However, some expression changes in response to doxorubicin treatment, observed especially in the high responder and in the high grade group were similar and are focussed on the figure 4. According to the large number of apoptosis related genes, we identified the first five ones that were differentially expressed in more than 50% of the samples and limit further explanation to these with special reference to apoptotic function in sarcoma or previous reports concerning doxorubicin treatment, although the understanding of their exact functions is limited. The up-regulated genes beyond the mentioned five are summarized in table 5.

Up-regulated genes
As previously shown, doxorubicin could overcome TRAIL resistance in a variety of sarcoma cell lines. An up-regulation of the TRAIL-R2, increasing the susceptibility to apoptosis inducing agents such as TRAIL, may be a possible explanation [62][63][64][65], as we found the Trail receptor 2 gene expression increased in 15 probes.
The second most frequently up-regulated gene was cyclindependent kinase inhibiton 1A (CDKN1A), a downstream target of p53, has has already been shown to be involved in cell cycle arrest and apoptosis induction by doxorubicin in sarcoma cells. CDKN1A can act as apositive regulator of senescence-like terminal proliferation arrest, but its function seems neither sufficient nor absolutely required for a treatment response to doxorubicin in tumor cells, especially soft tissue sarcoma [66][67][68][69][70].
GADD45A, a potent inhibitor of the c-Jun N-terminal kinase (JNK) cascade and NFKBIA, inhibits transcription factors associated with tumor growth [71][72][73][74] and was upregulated by doxorubicin in 12 probes. In a variety of soft tissue sarcoma cell lines GADD45A was found to increase cell cycle arrest and apoptosis {Zhu, 2008 #165}. For rhabdomyosarcoma, increased GADD45A has previously been associated with less aggressive tumor behaviour [75].
FAS, a member of the tumor necrosis factor receptor family, seems to be an important mediator in doxorubicin induced apoptosis. Its effects were shown to be dependant on metalloproteinases in soft tissue sarcoma and Ewing sarcoma. These metalloproteinases have been associated with aggressive tumor behaviour and may promote invasiveness and the occurrence metastasis of malignant cells [76][77][78].
Another member of the TNF receptor superfamily, CD40, was also up-regulated by doxubicin in our experiments.
To the authors' knowledge, CD40 has not yet been associated with doxorubicin treatment in liposarcoma, but its expression in soft tissue sarcoma was associated with an unfavourable outcome [79], whereas Lodge et al. could show a beneficial effect of antibody mediated CD40 activation in elimination of fibrosarcoma in nude mice [80].

Down-regulated genes
Among the down-regulated genes related to apoptotic pathways CD14, a receptor marking apoptotic cells was found in 8 probes. Increased CD14 has been associated with apoptosis induction and cell clearance, especially mediated by macrophages, but is expressed by a variety of other cells too [81][82][83]. Its role in terms of cell death mediation in sarcoma cells has not yet been examined, therefore interpretation of this finding remains difficult.
TIA-1, which encodes an RNA-binding protein with translation-regulatory functions has already been reported to be up-regulated in tumor specimens post-treatment with TNF alpha in soft tissue sarcomas. It was further hypothesized that TIA-1 could mediate death receptor mediated apoptosis in soft tissue sarcoma and that its overexpres- sion might sensitize endothelial cells to proapoptotic stimuli present in the tumor microenvironment and enhance NK cell cytotoxic activity against cancer cells [84].

Heatmap of total gene expression changes and cluster analysis after incubation with doxorubicin for 24 h
YWHAH, or 14-3-3 eta, is a member of the dimeric 14-3-3 family of signal transduction proteins that specifically binds to phosphorylated serine on a variety of signalling molecules, such as Bcl-2, MDMX, and Bax, thereby promoting cell survival and acting antiapoptotic in several tumor cells [85][86][87]. On the other hand, it is supposed to be associated with tumorigenesis through its binding interaction with gremlin1 [88]. Therefore the issue of further studies should be awaited before interpreting this finding.
PPP3CA (CCN1/Cyr61) is susceptible to various growth factors and promotes cell proliferation, adhesion, and differentiation and plays important roles in angiogenesis. Additionally, PPP3CA has been associated with tumorigenesis. It was reported that PPP3CA exerts its functions via interacting with integrins as well as heparan sulfate proteoglycan. By activating NF-kappaB and tyrosine kinase signalling pathways, PPP3CA is not only able to control cell growth, but also induce or suppress apoptosis in a cell type-specific manner [89,90]. To the authors' knowledge, it has so far not been reported in context with liposarcoma or doxorubicin treatment.
Integrin B2 (ITGB2) is known to play a role in mediating apoptosis [91] and chemotherapy resistance. Although it is widely attributed to white blood cells, it is also expressed in a variety of other benign and malignant cells and seems to play a major role in cell invasion and migration [92].

Conclusion
In summary, pro-and antiapoptotic genes were found upas well as down-regulated with a dominance of up-regulation of proapoptotic genes. The heterogeneous expression profiles reflect the heterogeneous reaction of liposarcomas to doxorubicin therapy. A lot of genes we found differentially expressed have not yet been associated with apoptosis in liposarcoma or doxorubicin treatment. Therefore it is not possible to relate our findings to other studies. Our study shows that the low clinical response rates of highly differentiated liposarcoma correlate to minimal changes in the expression patterns and that only high grade tumors, especially myxoid/roundcell and pleomorphic tumors, respond to doxorubicin on gene expression level. In most cases, this response seems to be  Heatmaps of expression changes of selected genes associated with apoptotic pathways after incubation with doxorubicin for 24 h ordered by responder group (5 a,b), grading (6 a,b), and histologic subtype (7 a/b) Figure 4 Heatmaps of expression changes of selected genes associated with apoptotic pathways after incubation with doxorubicin for 24 h ordered by responder group (5 a,b), grading (6 a,b), and histologic subtype (7 a/b). Expression changes as determined by comparison analysis were considered only if the probeset showed at least one P detection in untreated/treated sample pairs. Excluded expression changes are shown in grey. Only probesets with expression changes in 7 or more samples are given. Horizontal rows represent individual probesets. Vertical columns represent individual samples (left to right: low responders to high responders); gene symbols are indicated to the right. Gene expression changes are indicated by a continuous scale; the brightest red indicating ≥ 4-fold up-regulation; and, brightest green ≥ 4-fold down-regulated gene expression relative to the untreated control sample. Table 5: Summary of genes beyond the ones already mentioned in the text that were found to be up-regulated by doxorubicin treatment in more than 50% of the probes.

Gene symbol
Gene name Probes upregulated Additional information Apoptotic function PAWR PRKC (protein kinase C) apoptosis WT1 (Wilms tumor gene) regulator protein 11 STS with high WT1 mRNA expression levels have poorer outcome than those with low levels [89,90]. Ecteinascidin that has been shown to be effective against STS also increases expression of PAWR [92]. PAWR inhibits the PKC (atypical protein kinase)-NF-(kappa)B (nuclear factor-(kappa)B)-XIAP pathway [91].
antiapoptotic NFKBIA nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor alpha 10 The doxorubicin analogon DA-125 reduces proliferation in HT1080 fibrosarcoma cells through a NFKB dependent pathway [96,97]. MDM2 (upregulated in 9 samples in our series) is overexpressed in several liposarcoma subtypes [58] and increases NFKB activity in a p53 dependent manner and thereby leads to doxorubicin resistance [98].
proapoptotic/antiapoptotic IER3 immediate early response 3 10 IER3 function is increased by p53, that is induced by doxorubicin. IER3 is involved in cell cycle arrest and programmed cell death [99,100].
proapoptotic PEA15 phosphoprotein enriched in astrocytes 15 10 RPEA15 is regulating caspase-3 function in epidermal cells [104], but has not yet been associated with doxorubicin treatment or apoptosis in sarcoma cells.
antiapoptotic CD44 cell surface glycoprotein CD44 10 CD44 is a proapoptotic factor in FAS mediated apoptosis in sarcoma cells [112], but is also connected to cancer drug resistance [113]. CD44 has successfully been used as a target for liposomal encapsuled doxorubicin [114].
proapoptotic/antiapoptotic based on an increase of the extrinsic pathway such as TRAIL Receptor 2 and FAS but also members of the intrinsic pathway such as BCL2A1 were found to be differentially expressed. Interestingly several factors (NFKBIA, GADD45A, RIPK2, and PAWR) point to the NFKB transcriptional factor as possible mediator of doxorubicin effects.
Our results indicate that gene expression profiling may be a promising approach to improve the understanding of the diverse modes of programmed cell death in liposarcoma following doxorubicin treatment and can provide a molecular basis for new chemotherapeutic strategies.