Molecular Diagnostics of Prognostic Markers
Molecular Diagnostics of Prognostic Markers
Resistance to chemotherapy and the concept of individualized treatment
The basis of multicomponent cancer therapy is the view that cancer cells resistant
to one drug remain susceptible to other drugs. Clinically, the success of combination
treatments is frequently limited due to the development of broad-spectrum or multidrug
resistance (MDR). Since most established cytostatic drugs lack sufficient tumor specificity,
normal tissues are also affected leading to severe side effects. This prevents the
application of sufficiently high doses to kill resistant tumor cells or resistant
tumor stem cells. Thereby, drug resistance develops causing treatment failure. Novel
strategies to broaden the narrow therapeutic range by separating the effective dose
and toxic dose would be of great benefit for the patients.
Whereas the statistical probability of therapeutic success is well-known for larger
groups of patients from clinical therapy trials, it is, however, not possible to predict
how an individual tumor will respond to chemotherapy. The question arises as to which
particular cytostatic agent and which combination of substances is most suited for
an individual tumor. Although clinicopathological prognostic factors such as tumor
size, lymph node, and far distance metastases are valuable for the determination of
the prognosis of larger cohorts, they are less helpful for individualized cancer treatment.
Further biomarkers are required to predict the likelihood of an individual tumor's
responsiveness or of toxicity in normal organs of each patient.
Clinically, it has been known for many years that the same doses of a medication cause
considerable heterogeneity in efficacy and toxicity across human populations [1], [2]. This heterogeneity can lead to unpredictable life-threatening or even lethal adverse
effects in patients who react hypersensitively [3], [4]. The interindividual variability in drug response cannot satisfactorily be explained
by clinical factors such as renal and liver function, patient's age and comorbidity,
lifestyle, or co-medication and compliance of the patient. Therefore, molecular factors
come into the center of interest.
In the past decades, enormous efforts have been undertaken to predict drug resistance
in vitro [5], [6]. The idea was to determine sensitivity or resistance beforehand to be subsequently
able to choose the clinically most effective treatment for each individual patient.
The methods available at that time, however, were not established for clinical routine
diagnostics. In the 1990s, attempts were made to test the drug response of cancers
by determination of the expression of resistance proteins. However, it was not possible
to define consensus recommendations for the standardized detection of resistance proteins
expressed in low amounts in tumors with low degrees of drug resistance [7], [8], [9]. Hence, this approach was also not realized in clinical routine. Another important
reason is that no single mechanism can sufficiently explain resistance to therapy
[10]. While the concept of individualized therapy itself traces back to the 1950s [11], the advances in cell and molecular biology have only recently opened new avenues
for the characterization of drug-resistant tumors. Still, the transfer of such techniques
from the bench to the bed is an unfulfilled requirement. Nevertheless, current progress
in molecular biology gives reason to believe that molecular approaches will significantly
improve individual tumor therapy.
In the present overview, we focus mainly on our own efforts in this thriving field
of research. We searched for novel markers to predict the responsiveness of tumors
to chemotherapy and the survival chances of patients at the level of DNA, mRNA, and
proteins ([Fig. 1]). Promising candidate markers from these analyses were then taken to identify novel
compounds derived from natural origin, which specifically target these markers. In
a recent survey from the National Cancer Institute (NCI, USA), it was convincingly
shown that the majority of cancer drugs brought on the market during the past half
century is derived from natural sources or is based on principles of action taken
from nature [12], [13]. Therefore, we have concentrated on natural products [14], [15], [16]. For general overviews in the field of personalized cancer medicine, the reader
is also referred to comprehensive reviews in the literature [17], [18], [19], [20], [21], [22].
Fig. 1 Combining molecular diagnostics and natural product research for personalized cancer
medicine. Biomarkers can be identified at the DNA, mRNA, and protein levels. Markers
with prognostic significance for response to treatment and patients' survival can
serve as a target for the development of novel drugs. Natural products with activity
towards cancer cells can be identified by phytochemistry. Functional screening and
molecular docking allow screening for inhibitors of drug targets. The genomic aberrations
in (A) show an amplification of 5q13q14 harboring the DHFR gene in methotrexate-resistant CEM/MTXR3 cells. The deletion represents the loss
of chromosomal locus 9p21 cells in CCRF‐CEM cells, where the MTAP gene and the tumor suppressor genes INK4A, INK4B, and ARF are located [34]. B The microarray hybridization shows the mRNA expression in a breast cancer biopsy
[146]. C Immunohistochemical staining of the oncogene product, c-MYC, in a lung cancer biopsy
[147]. C‐MYC is specifically localized in the nuclei of the cells. D Several yew species, including the Chinese Taxus yunannensis contain paclitaxel, which is already an established anticancer drug. E Functional screening for inhibitors of P-glycoprotein can be done by using fluorescence
microscopy and rhodamine 123 (R123) as a fluorescent probe. The cytotoxic R123 is
extruded by multidrug-resistant cells (black cells). Upon treatment with a P-glycoprotein
inhibitor, cells start to take up R123 and die. Shown are P-glycoprotein-expressing
leukemia cells from a patient, which were treated with R123 and without and with the
P-glycoprotein inhibitor chloropromazine. F Molecular docking of drug molecules to three-dimensional crystal structures allow
the in silico screening for putative inhibitors of therapeutic targets. Shown is the docking of
natural products derived from traditional Chinese medicine to the tyrosine kinase
binding domain of the human epidermal growth factor receptor (top) and the amino acids
located in this binding pocket responsible for binding of erlotinib, a well-known
EGFR-inhibitor. Erlotinib was used as a control drug [40]. Parts of the figure were taken from the above-cited references with permission
of the publishers.
A special feature of our concept is the combination of methods of molecular diagnostics
to identify prognostic markers with natural product research to identify inhibitors
for these markers. Normally, both areas of research are separated from each other.
An integration of both areas in one interdisciplinary approach may open new avenues
for the development of novel treatment options for cancer.
Predictive DNA markers
Classical cytogenetics: Clear cell renal cell carcinoma is a tumor type with a poor prognosis and low responsiveness
towards chemotherapy. Less than 50 % of patients can be cured by other therapies such
as surgery. To evaluate the prognostic significance of cytogenetic findings in clear
cell renal cell carcinoma, the results of classical cytogenetic staining techniques
(DAPI-staining and G-banding) and the results of 118 primary RCCs were evaluated in
relation to classical indicators of prognosis and overall survival [23]. Losses at the short arm of chromosome 3 (3 p) were most prevalent and included
32 monosomies of chromosomes 3 and 84, structural aberrations involving unbalanced
translocations resulting in duplication of sequences at the long arm of chromosome
5 (5 q). Patients with the gain of band 31 to the end of chromosome 5 (5q31-qter)
had a significantly better outcome than those without this aberration (p = 0.001).
There was no association between the gain of 5 q and any of the well-known variables
for prognosis, including low versus high clinical stage and grade of malignancy. Among
additional chromosomal aberrations, loss of chromosome 9/9 p was associated with distant
metastasis at diagnosis (p = 0.006). The data indicate that the gain of 5 q identifies
a clinically favorable cytogenetic variant of clear cell renal cell carcinoma and
demonstrate the impact of specific chromosome aberrations as additional prognostic
indicators in clear cell RCC. The usefulness of cytogenetics for prognosis of a wide
variety of hematopoietic and solid cancers has been demonstrated during the past two
decades [24], [25], [26], [27], [28], [29], [30], [31], [32], [33].
Comparative genomic hybridization (CGH): CGH is an advanced cytogenetic method to analyze unbalanced genomic aberrations,
i.e., gains and losses of genetic material (amplifications and deletions). Balanced
aberrations (translocations and inversions) cannot be visualized by CGH. In a pilot
project, we analyzed whether it is possible to dissect drug-resistant tumor cells
from sensitive ones by comparative genomic hybridization. Ten T-cell acute lymphoblastic
(T‐ALL) CEM cell lines selected for resistance towards methotrexate, doxorubicin,
vincristine, or hydroxyurea, respectively, and parental drug-sensitive CCRF‐CEM cells
were analyzed [34]. Most genomic imbalances were not specific for drug resistance, as they were found
in both parental and drug-resistant lines. We were concerned with those imbalances,
which were specifically present in drug-resistant, but not in drug-sensitive cells.
All methotrexate-resistant cell lines were characterized by an enhancement or an amplification
of 5q13. The methotrexate resistance-conferring dihydrofolate reductase (DHFR) gene is located at this locus. Gain of DHFR was verified by PCR analyses. Some but not all methotrexate-resistant cell lines
showed enh(14q21qter) and amp(5p13p15.2). These two loci harbor the methylenetetrahydrofolate
dehydrogenase (MTHFD1) and 5′-methyltetrahdrofolate-homocysteine methyltransferase reductase (MTRR) genes, both of which are involved in folate metabolism. Their gain indicates a role
in methotrexate resistance. A loss of 4q35 was found in two methotrexate-resistant
sublines and in the doxorubicin-resistant cells, where the proapoptotic caspase-3
gene is located. The thioredoxin (TXN) locus 9q31 was also enhanced in the doxorubicin-resistant cell line. Furthermore,
2p22pter was increased in hydroxyurea-resistant CEM cells. Ribonucleotide reductase
polypeptide M2 (RRM2), which confers resistance to hydroxyurea, resides at this locus.
Furthermore, genomic imbalances were investigated in 15 T-cell acute lymphoblastic
leukemia cell lines using CGH [35]. In addition, the in vitro response to the cytostatic drug doxorubicin was evaluated by means of a growth inhibition
assay. The most frequent genomic imbalance (gain of 6q23) was shared by 9 of the 15
cell lines. This chromosomal locus harbors the C‐MYB oncogene. A significant loss of 18q23 was observed in eight lines. Seven of the cell
lines were characterized by a loss of the entire short arm of chromosome 9 or parts
of it with 9p21 as a minimal band of overlap. This locus contains the tumor suppressor
genes INK4A, INK4B, ARF and the MTAP gene. Interestingly, cell lines with a 9p21 deletion exhibited twice the number of
gains and 1.6 times the number of losses per line as compared with the cell lines
without this deletion. Based on the dose-response curves of the cell lines for doxorubicin,
eight doxorubicin-sensitive cell lines had an inhibition concentration 50 % (IC50) < 10 nM (CCRF-CEM2, JURKAT, KE-37, MOLT-3, MOLT-4, P12-Ichikawa, PEER, and RPMI-8402)
and seven doxorubicin-resistant cell lines had an IC50 > 10 nM (BE-13, CCRF-CEM1, HUT-78, J-Jhan, Karpas-45, MOLT-17, and PF-382). The average
number of copy number alterations (CNAs) per cell line was higher in the sensitive
than in the resistant group.
Eight cell lines newly established from glioblastoma multiforme were also examined
by CGH for their patterns of genomic imbalance [36]. The total number of CNAs varied between 15 and 24 indicating a distinctly progressed
karyotypic evolution. The most frequent CNAs were gains of the entire chromosome 6
or, at least, parts of it, and of 7p22. Other changes were gains of 3q26qter and the
entire chromosome 7 and losses of segments on chromosome 4 q and of the short arm
of chromosome 10. Enh(3q21q25), dim(4q22q33) and dim(4qter), dim(9p21), dim(13q22),
enh(15q14), and enh(18q22q23) were also frequently observed. Using a hierarchical
cluster analysis, the specific patterns of genomic imbalance allowed the separation
into two main groups indicating different karyotypic evolutions.
As a next step, we applied CGH to clinical biopsies of tumors to determine whether
meaningful data can be obtained which are predictive for response to therapy and survival
time. We and others applied CGH on oral squamous cell carcinoma [37]. Gain of 11q13 was involved in advanced stages of malignancy in oral squamous cell
carcinoma. In addition, the proportion of patients deceased within one year after
diagnosis was higher in the group whose tumors showed an increased 11q13 copy number
as compared to the group without this increase. This could point to an association
of gain in 11q13 and tumor aggressiveness [38]. Several cancer-related genes reside to this chromosomal locus making the association
with tumor aggressiveness and an enhancement of band 13 at the long arm of chromosome
11 reasonable. Among these genes are the oncogenes EMS1, EMSY, FGF3, FGF4, the cell-cycle regulator CCND1 and the TAOS1 gene, which is involved in tumor progression. This locus also harbors the ORAOV1 gene (oral cancer overexpressed gene 1). Patients with tumors characterized by the
gain of 3q26-qter plus 5p14–p15 died earlier (i.e., less than 15 months) after excision
of the tumor compared to the group without these imbalances [39]. Nineteen of 35 tumors showed a gain of chromosome band 7p12 [239], where the gene
for the epidermal growth factor receptor (EGFR) is located. A highly complex but strikingly consistent pattern of other genomic
imbalances (average, 32 CNAs per tumor) was associated with the 7p12 alteration. Average
disease-free survival of tumors without a 7 p gain clearly exceeded that of tumors
with a gain of 7 p (36.8 vs. 21.3). Relapse occurred in 63 % in the 7p12-positive
vs. 25 % in the negative group. Average disease-free survival of tumors without the
7 p gain clearly exceeded that of tumors with the gain of 7 p (36.8 vs. 21.3). Then,
genomic imbalances were investigated by hierarchical cluster analysis and clustered
image mapping to investigate whether genomic profiles correlate with clinical data.
There was indeed a significant relationship: patients suffering from tumors without
enh(7p12) lived significantly longer than patients with tumors that harbor enh(7 p)
(p = 0.024) [39], [40]. The data clearly show that several genomic imbalances may affect the clinical outcome
in human oral squamous cell carcinoma.
DNA methylation: The genome-wide simultaneous methylation status of CpG islands in colorectal carcinoma
was investigated by means of a microarray-based technique [31]. Amplicons from tumor and control samples were pools of differentially methylated
CpG island fragments hybridized to a panel of approximately 8000 CpG island tags.
Data analysis identified 694 CpG island loci hypermethylated in a group of 14 colorectal
tumors. The Stanford hierarchical cluster algorithm segregated the tumors into two
subgroups, one of which exhibited a high level of concurrent hypermethylation while
the other had little or no methylation. This is in agreement with observations of
a CpG island methylation phenotype present in colorectal tumors [42]. The present study demonstrates that this microarray-based technique is useful in
classifying tumors according to their methylation profiles.
Single nucleotide polymophisms: An important result of the human genome project is the high DNA variability. Statistically
a genetic variation (polymorphism) occurs once in 1200 bases. In most cases, polymorphic
gene variants lead to a diminished protein function; in some cases, however, increased
activities have been reported [1]. In contrast to somatic mutations, i.e., in cancer, polymorphisms in germ line cells
are stable and heritable. Polymorphisms include single nucleotide polymorphisms (SNPs),
and length differences in micro- and minisatellites. An SNP represents a single base
exchange that may or may not cause an amino acid exchange in the encoded protein.
The frequency of SNPs is greater than 1 % in a population and accounts for over 90 %
of genetic variation in the human genome. The number of SNPs has been estimated in
a range from 1 to 10 million [43], [44], [45]. Between 50 000 and 250 000 SNPs are distributed in and around coding genes [46].
In our research, we focused on SNPs in drug transporter genes. The ATP-binding cassette
(ABC) transporter family consists of 49 members. More than 10 of them are implicated
in drug resistance to cancer chemotherapy [47–49]. They are important determinants
of drug absorption, tissue targeting, and drug elimination. ABC transporters confer
drug resistance by lowering the intracellular drug concentrations down to sublethal
levels. Cancer cells, which express P-glycoprotein (ABCB1, MDR1) reveal a multidrug resistance phenotype to a broad range of structurally and functionally
different drugs, including anthracyclines, anthracenediones, Vinca alkaloids, taxanes, epipodophyllotoxins, and others. P-glycoprotein is also expressed
in various normal organs, such as brain vessels, adrenal gland, kidney, liver, and
gastrointestinal tract. P-glycoprotein contributes to the blood-brain barrier, translocates
hormones, and detoxifies xenobiotics taken up along with nutrients. Although the clinical
therapy failure of tumors is multifaceted, its role for drug resistance is evident,
and the prognostic significance of P-glycoprotein as an indicator for failure of chemotherapy
and poorer outcome has been demonstrated in a number of studies [50], [51], [52], [53], [54], [55]. As of yet, 29 polymorphisms have been identified in the ABCB1 gene, with considerable differences in their frequencies among ethnic groups [56], [57], [58], [59]. Of them, the G2677T/A SNP in exon 21 and the C3435T SNP in exon 26 have been most
intensively studied, because they diminish the expression and function of P-glycoprotein
[58], [60]. Whereas the C2677T/A polymorphism causes an A893S/T amino acid substitution, the
functional relevance of the C3435T variant is unknown, because this is a silent SNP.
It is possible that specific haplotypes of the ABCB1 (MDR1) gene might determine the efficacy and toxicity of drugs. The effect of the G2677T
and C3435T SNPs on the pharmacokinetics and pharmacodynamics of drugs is still a matter
of controversy, since contrasting results have been provided [61], [62], [63], [64], [65], [66], [67], [68]. The expression of P-glycoprotein in normal tissues is thought to play an important
role for the pharmacokinetics and pharmacodynamics of many drugs of different drug
classes. A definitive answer on the role of ABCB1 SNPs requires further studies.
Predictive mRNA markers
We have developed a low-density DNA microarray which contains 38 genes of the ATP-binding
cassette (ABC) transporter gene family [69]. This tool has been validated with three different multidrug-resistant sublines
(CEM/ADR5000, HL60/AR, and MCF7/CH1000) known to overexpress either the ABCB1 (MDR1), ABCC1 (MRP1), or ABCG2 (MXR, BCRP) genes. When compared with their drug-sensitive parental lines, we observed not only
the overexpression of these genes in the multidrug-resistant cell lines but also of
other ABC transporter genes, pointing to their possible role in multidrug resistance.
These results were corroborated by quantitative real-time reverse transcription-PCR
[69].
As a next step, we applied this microarray to detect drug resistance in clinical samples.
We identified four new ABC transporters, which were overexpressed in many samples
of patients with acute myeloid leukemia (AML) compared with healthy bone marrow: ABCA2, ABCA3, ABCB2, and ABCC10 [70]. The overexpression of these four genes was verified by real-time PCR in 42 samples
from children with AML and 18 samples of healthy bone marrow. The median expression
of ABCA3 was three times higher in 21 patients who had failed to achieve remission after the
first course of chemotherapy than in a well-matched group of 21 patients who had achieved
remission at this stage (p = 0.023). Incubation of cell lines with a number of different
cytostatic drugs induced an up-regulation of ABCA3. Downregulation of ABCA3 by small interfering RNA sensitized cells to doxorubicin. These results show that
ABCA2, ABCA3, ABCB2, and ABCC10 might play a role for AML. ABCA3 is the most likely transporter to cause drug resistance.
Furthermore, we observed a consistent overexpression of ABCA2/ABCA3 in clinical samples of T-cell acute lymphoblastic leukemia (T‐ALL) [71]. Therefore, we analyzed the association of these two genes with drug resistance.
Treatment of CCRF‐CEM and Jurkat cells with methotrexate, vinblastine, or doxorubicin
led to an induction of ABCA3 expression, whereas a significant increase of ABCA2 expression was only observed in Jurkat cells. To study the causal relationship of
ABCA2/A3 overexpression with drug resistance, we applied RNA interference (RNAi) technology.
RNAi specific for ABCA2 or ABCA3 led to a partial decrease of expression in these two ABC transporters. Upon cotreatment
of RNAi for ABCA2 with methotrexate and vinblastine, a partial decrease of ABCA2 expression as well as a simultaneous increase of ABCA3 expression was observed. Vice versa, ABCA3 RNAi plus drugs decreased ABCA3 and increased ABCA2 expression. This indicates that downregulation of one ABC transporter was compensated
by the upregulation of the other. Application of RNAi for both ABCA2 and ABCA3 resulted in a more efficient reduction of the expression of both transporters. As
a consequence, a significant sensitization of cells to cytostatic drugs was achieved.
In conclusion, ABCA2 and ABCA3 are expressed in many T‐ALL and contribute to drug resistance.
Predictive protein markers
So far, our data obtained from CGH and DNA methylation analysis indicate that no single
mechanism can explain the resistance to chemotherapy. The multifactorial nature of
drug resistance implies that the analysis of comprising expression profiles may predict
drug resistance with higher accuracy than single gene or protein expression studies.
Therefore, 40 cellular parameters (drug resistance proteins, proliferative, apoptotic,
and angiogenic factors, products of proto-oncogenes, and suppressor genes) were evaluated
mainly by immunohistochemistry in specimens of primary non-small cell lung carcinoma
(NSCLC) of 94 patients and compared with the response of the tumors to doxorubicin
in vitro [72]. The protein expression profile of NSCLC was determined by hierarchical cluster
analysis and clustered image mapping. The cluster analysis revealed three different
resistance profiles. The frequency of each profile was different. The resistance proteins
P-glycoprotein (MDR1, ABCB1), thymidylate synthetase, glutathione-S-transferase-π, metallothionein, O
6-methylguanine-DNA-methyltransferase and major vault protein/lung resistance-related
protein were most frequently upregulated in one of the three clusters, while microvessel
density, the angiogenic factor vascular endothelial growth factor and its receptor
FLT1, and ECGF1 as well were downregulated. In addition, the proliferative factors
proliferating cell nuclear antigen and cyclin A were reduced compared to the sensitive
NSCLC. In this resistance profile, FOS was upregulated and NM23 downregulated. In
the second profile, only three resistance proteins were increased (glutathione-S-transferase-π, O
6-methylguanine-DNA methyltransferase, major vault protein/lung resistance-related
protein). The angiogenic factors were reduced. In the third profile, only five of
the resistance factors were increased (P-glycoprotein, thymidylate synthetase, glutathione-S-transferase-π, O
6-methylguanine-DNA-methyltransferase, major vault protein/lung resistance-related
protein).
In a second analysis, we analyzed the expression of resistance-related proteins for
survival times of patients. NSCLC is usually associated with a poor survival prognosis
[73]. Some patients survive their disease, and the underlying molecular mechanisms are
still poorly understood. Therefore, we have evaluated the expression of 21 gene products
(oncogene and tumor suppressor products and proliferative, apoptotic, and angiogenic
factors) in paraffin-embedded primary NSCLCs from 216 patients and correlated the
data with the survival times of the patients (survival of more or less than five years).
The protein expression of FOS, P53, RAS, ERBB1, JUN, PCNA, cyclin A, FAS/CD95, and
HIF-1β revealed a correlation to survival by means of the χ2 test. In a second step, these nine parameters were further analyzed by hierarchical
cluster analyses of all patients, of stage III patients, and of patients with squamous
cell lung carcinomas. We identified clusters with significantly more long-term survivors.
The expression of FOS, JUN, ERBB1, and cyclin A or PCNA were decreased in carcinomas
of patients with long-term survival. The expression profile of these factors predicts
a significantly better long-term outcome of NSCLC patients. This may have implications
for the development of individualized therapy options in the future.
Survival time of patients is mostly influenced by metastasis. Therefore, we were also
interested to investigate the role of these proteins for metastasis and to evaluate
whether different protein expression patterns exist in primary squamous cell lung
carcinomas of patients with and without lymph node involvement. Formalin-fixed, paraffin-embedded
specimens from 130 patients with squamous cell lung carcinomas were analyzed by immunohistochemistry
[74]. In a first step, proteins were selected which showed a relationship to lymph node
involvement. The expression of JUN, ERBB2, MYC, cyclin D, PCNA, bFGF, VEGF and Hsp70
proteins revealed a positive correlation to lymph node involvement. In contrast, caspase-3,
Fas ligand, Fas/CD95, and PAI showed an inverse correlation to lymph node involvement.
In a second step, these parameters were further analyzed by hierarchical cluster analyses.
The resulting clusters were correlated to patients with or without lymph node involvement.
The data show that different protein expression patterns exist between primary squamous
cell lung carcinomas with and without lymph node involvement and within carcinomas
with lymph node involvement. The data suggest that various metastasis profiles exist.
This has also been shown for other tumor types by other authors [75], [76], [77], [78], [79], [80], [81], [82], [83].
Molecular Targeted Therapy with Natural Products
Molecular Targeted Therapy with Natural Products
A chemical “treasure box” from medicinal plants
Given the fact that a considerable portion of all drugs used nowadays in oncology
as well as in general pharmacology are from natural origin [12], [13], natural products represent a valuable source for drug development. Surprisingly,
this potential is frequently underestimated in pharmacology, biology, and medicine.
In Europe, medicinal herbs gradually lost importance in the course of chemistry's
progress in industrialized countries during the 20th century. From the perspective
of the pharmaceutical industry, chemical compounds from natural origin may pose more
problems concerning intellectual property and patenting issues. Furthermore, plant
extracts can hardly be subjected to high-throughput screening technologies. For pharmaceutical
chemists, it might be more attractive to fiddle about huge chemical libraries of synthetic
compounds obtained by combinatorial chemistry than to isolate compounds from plants,
which can be painful and time-consuming. This might at least in part explain the decline
of natural products in drug development during the 20th century. Fortunately, the
potential of natural products for chemotherapy seems to be rediscovered very recently,
and the current thriving revival of phytotherapy is followed by an increasing scientific
interest in bioactive compounds as lead drugs for semi-synthetic modification.
Our own interest to identify novel cytotoxic compounds with activity against tumor
cells derived from medicinal plants used in traditional Chinese medicine (TCM) was
raised in the 1990s [84]. The rationale behind this approach is that TCM looks back on a tradition of more
than 5000 years. Hence, it can be expected that many ineffective medicinal herbs have
vanished over the centuries. Indeed, a number of clinical studies were conducted on
TCM providing convincing evidence to gain credibility and reputation outside China.
As recently reviewed, clinical trials with TCM remedies focus on three major fields
in cancer research: (i) improvement of poor treatment response rates towards standard
chemo- and radiotherapy, (ii) reduction of severe adverse effects of standard cancer
therapy, and (iii) unwanted interactions of standard therapy with herbal medicines
[85].
Apart from the approved drugs artemisinin and its semisynthetic derivative artesunate
[86], [87], we have analyzed cellular and molecular mechanisms of several other chemically
characterized natural products derived from TCM. Among them were known compounds with
still insufficiently defined modes of action, which were investigated by us using
molecular biological and pharmacogenomic approaches, i.e., arsenic trioxide, ascaridol,
berberine, cantharidin, cephalotaxine, curcumin, homoharringtonine, luteolin, isoscopoletin,
scopoletin, vitexin, isovitexin, and others [88], [89], [90], [91], [92], [93], [94], [95], [96], [97], [98], [99], [100], [101], [102], [103], [104], [105], [106], [107], [108], [109]. Furthermore, several novel bioactive compounds were described and analyzed in the
course of our investigations, i.e., tetracentronsine A, tetracentronsides A, B and
C, the two novel α-tetralone derivatives, berchemiasides A and B, as well as the novel
flavonoid quercetin-3-O-(2-acetyl-α-L-arabinofuranoside) [110], [111], [112]. Artemisinin also reveals profound antimalarial activity [14]. During the past years, several derivatives (artesunate, artemether, arteether,
artelinate) have been synthesized to improve the antimalarial activity.
Furthermore, we started the analysis of antiviral effects of natural products. We
were the first to demonstrate that artesunate inhibits NF-κB activity, leading to
the inhibition of viral replication. NF-κB is involved in the transcriptional regulation
of immediate early, early, and late proteins of human cytomegalovirus (HCMV) necessary
for viral replication [113]. Artesunate also acts against cytomegaloviruses in vivo [114]. The antiviral activity of artesunate is not limited to HCMV. We showed that herpes
simplex virus 1, hepatitis B and C viruses and others are also efficiently inhibited
by artemisinin and artesunate [115], [116].
Summarizing the results of our investigations in search of molecular markers, it turned
out that patient survival and response to chemotherapy is multifactorial and that
no single factor sufficiently explains treatment failure and death. We conclude that
resistance markers may also be valuable targets for strategies to develop targeted
therapies for individual cancer patients. The idea is to identify small molecules,
which inhibit proteins that are essential for therapy response and worse prognosis.
The focus of our own research is on small molecules from natural origin. We have selected
three targets to prove the validity of this concept: (i) the chromosomal locus 9p21
harbors several genes, including MTAP, INK4A, INK4B, and ARF. MTAP offers the opportunity for a chemoselective treatment which affects tumor cells with
9p21 deletion but spares normal tissues; (ii) P-glycoprotein is an efflux transporter
that extrudes many anticancer drugs out of tumor cells rendering them resistant to
chemotherapy. We have searched for compounds that are not recognized by P-glycoprotein
and which, hence, kill P-glycoprotein-expressing multidrug-resistant tumor cells with
a similar efficacy than P-glycoprotein-negative drug-sensitive cells. Alternatively,
the inhibition of this drug transporter by small molecules may overcome multidrug
resistance; (iii) EGFR represents an important signal transducing molecule regulating
tumor growth, apoptosis, differentiation and other key processes. The recognition
of its value as a target for novel drugs resulted in the development of therapeutic
antibodies and small molecules. However, tumor cells can also exert resistance to
these novel therapeutics and novel EGFR inhibitors are required.
Chemoselective treatment of tumor cells with 9p21 deletion by L-alanosine
Unfortunately, most established anticancer drugs not only kill tumor cells but also
affect normal tissues [117]. It would, thus, be desirable to have treatment targets which allow the distinction
between normal and cancerous tissue.
A deletion of the short arm of chromosome 9 at band 21 (9p21) is a frequent chromosomal
aberration in many tumor types including acute lymphoblastic leukemia [118], [119]. This locus harbors the tumor suppressor genes p16INK4A, its alternative splice product p14ARF, and p15INK4B. Their gene products regulate the progression from G1 to S phase of the cell cycle
via the RB1 or p53 pathways [120]. In addition, the methylthioadenosine phosphorylase (MTAP) gene is also localized at this chromosomal region and p16INK4A and MTAP genes are frequently codeleted in tumors [121]. MTAP converts methylthioadenosine into adenine and 5-methylthioribose 1-phosphate
by phosphorolysis. Adenosine is used to recruit adenine nucleotide pools for DNA synthesis.
Methylthioadenosine can, therefore, serve as an alternative purine source, if the
de novo purine biosynthesis is inhibited by antimetabolites, i.e., methotrexate [122]. Tumor cells with a deletion of the MTAP gene at chromosome 9p21 cannot use this salvage pathway and die upon methotrexate
challenge. As all normal tissues have MTAP activity [122], they do not die at methotrexate concentrations lethal for MTAP-deficient cancer
cells. The MTAP salvage pathway may, thus, offer a unique opportunity for a selective
tumor therapy which spares normal tissues. This treatment advantage may, however,
vanish in MTAP-deficient cells, which developed resistance towards MTX, i.e., by amplification
of the dihydrofolate reductase gene.
Therefore, we analyzed the role of MTAP for chemoselectivity of the antimetabolites
trimetrexate and L-alanosine and whether cross-resistance to methotrexate hampers their effectiveness
[123], [124]. Trimextrexate is a derivative of the established anti-cancer drug methotrexate.
L-Alanosine is an amino acid analogue and antibiotic derived from the bacterium Streptomyces alanosinicus. The analyses were performed with CCRF‐CEM cells in which the 9p21 deletion was found
by CGH and fluorescence in situ hybridization. T-cell acute lymphoblastic leukemia (T‐ALL) cells (CCRF‐CEM) were
transfected with an MTAP expression vector. A green fluorescent protein (GFP) plasmid was co-transfected to monitor the transfection efficacy by flow cytometry.
The response of MTAP-transfected cells to the antimetabolites methotrexate, trimetrexate, and L-alanosine was decreased compared to mock control transfectants using growth inhibition
assays. The activity of doxorubicin which is not involved in DNA biosynthesis was
not changed in MTAP transfectants. As the p16INK4A tumor suppressor gene resides also at 9p21, we transfected CCRF‐CEM cells with a
p16INK4A expression vector. These transfectant cells were more resistant to all four drugs
indicating that p16INK4A did not specifically affect antimetabolites. The chemoselective effect of antimetabolites
in MTAP-deleted tumor cells may, however, be compensated by the development of drug resistance.
To prove this possibility, we analyzed a methotrexate-resistant subline, CEM/MTX1500LV,
in which the methotrexate resistance-conferring dihydrofolate reductase (DHFR) gene was amplified. While trimetrexate exhibited considerable cross-resistance in
CEM/MTX1500LV cells, L-alanosine did not. Thus, L-alanosine could exhibit chemoselectivity in 9p21/MTAP-deleted cells, even if DHFR amplification occurs. We conclude that L-alanosine may be more suitable than methotrexate or trimetrexate for MTAP-mediated chemoselective treatment of T‐ALL. Pre-therapeutic detection of 9p21 and
MTAP deletion may be helpful in developing a predictive molecular chemosensitivity test
for T‐ALL.
Natural products that bypass or modulate P-glycoprotein-mediated multidrug resistance
Non-cross-resistant natural products: In a recent investigation, we analyzed the cross-resistance profile of clinical
samples of 59 tumors of different origins and 38 lung tumors in vitro [125]. Cytostatic drugs from different classes were used (anthracyclines, antibiotics,
Vinca alkaloids, epipodophyllotoxins, antimetabolites, and alkylating agents). Tumors exert
broad resistance profiles. Tumors resistant to one drug also tend to be resistant
to other drugs, while sensitive tumors reveal sensitivity towards many drugs. Expression
of P-glycoprotein and the proliferative activity of tumors were identified as underlying
mechanisms of broad spectrum resistance.
As a second step, the cross-resistance profiles were analyzed in sensitive and resistant
cell lines from different tumor types to study underlying mechanisms. In an effort
to find new treatment possibilities, novel cytotoxic compounds with activity against
otherwise drug-resistant tumor cells were investigated [125]. We used the CEM/ADR5000 leukemia cell line overexpressing P-glycoprotein/MDR1 and its parental cell line, CCRF‐CEM to investigate cross-resistance profiles. CEM/ADR5000
cells were more than 1000-fold resistant to the selecting agent, doxorubicin. They
were also highly cross-resistant to the anthracycline epirubicin (484-fold) but less
cross-resistant to idarubicin (6.9-fold). We also tested the anthracycline metabolites
doxorubicinol, epirubicinol, and idarubicinol, which revealed cross-resistance to
a lesser degree than the non-metabolized parental drugs (range from 1.6- to 382-fold).
Cross-resistance to the Vinca alkaloids vincristine, vinblastine, vindesine, and vinorelbine was in a range of
14- to 613-fold and to the taxanes paclitaxel and docetaxel in a range of 200- to
438-fold. The degrees of resistance to the epipodophyllotoxines etoposide and etoposide
phosphate were lower (18- and 11-fold, respectively). In an effort to identify novel
compounds with activity against otherwise drug-resistant tumor cells, we analyzed
natural products derived from medicinal plants used in traditional Chinese medicine
(TCM) in CCRF‐CEM and CEM/ADR5000 cells. Interestingly the multidrug-resistant cells
revealed either low degree of cross-resistance (cephalotaxine, berberine, homoharringtonine,
maesopsin), no clear cross-resistance (E)-3-(4-hydroxyphenyl)-[2-(4-hydroxyphenyl)-ethyl]-prop-2-enamide, N-p-coumaryl tyramine, maslinic acid) or even enhanced sensitivity towards the natural
products from TCM [cantharidin, tetracentronside, 3-(2-hydroxyethyl)-1H-indole-5-O-β-D-glucopyranoside, kaempferol, artesunate].
In order to gain a systematic approach in identifying natural products from TCM with
inhibitory activity against multidrug-resistant tumor cells, we compiled 531 cytotoxic
natural products and derivatives thereof in a database [126]. These compounds were tested in the drug screening program of NCI (www.dtp.nci.nih.gov).
In combination with microarray data, the generation of hypotheses regarding their
modes of action is a starting point for further mechanistic studies. We correlated
the IC50 values of the 60 NCI tumor cell lines for these 531 natural products with the accumulation
data for rhodamine 123 (R123) as a functional assay for P-glycoprotein and with the
cell doubling times as a parameter of proliferation. While the IC50 values for only 18 compounds correlated with R123 accumulation (3 %), 162 natural
products were significantly associated with the cell doubling times of the cell lines
(31 %), indicating that natural products might be a rich source for novel drug candidates
with activity to bypass P-glycoprotein- and proliferation-associated drug resistance.
We will systematically exploit the chances to develop novel cancer drugs from TCM
with improved features.
Modulators of P-glycoprotein function: Another concept was the blockage of P-glycoprotein by specific inhibitors. Despite
huge efforts in academia and industry, no P-glycoprotein inhibitor has clinically
showed satisfying results and reached the pharmaceutical market yet [127], [128]. Most of these resistance-modifying agents (RMA) are too toxic at the required doses.
Therefore, the development of novel RMAs to overcome MDR represents a major challenge
to modern cancer chemotherapy. In this context, we analyzed natural products for their
ability to inhibit P-glycoprotein function.
The antimycobacterial quinolones 1-methyl-2-undecyl-4-quinolone, dihydroevocarpine
and evocarpine as well as the indoloquinazoline alkaloids rutaecarpine and evodiamine
– all from the Chinese medicinal herb Evodia rutaecarpa – were tested in two in vitro assays, for cytotoxicity and interaction with P-glycoprotein. Cytotoxicity was measured
in a cell proliferation assay against sensitive CCRF‐CEM and multidrug-resistant CEM/ADR5000
cells. An assay monitoring the P-glycoprotein-dependent accumulation of the dye calcein
in porcine brain capillary endothelial cells (PBCECs) was used to study interactions
of the test substances with this efflux pump. Rutaecarpine and evodiamine showed quite
a high toxicity with IC50 values from 2.64 to 4.53 µM and were weak modulators of P-glycoprotein activity.
The degrees of resistance in CEM/ADR5000 towards the saturated quinolones 1-methyl-2-undecyl-4-quinolone
and dihydroevocarpine were between 3 and 4. In the calcein assay, these two quinolones
were shown to be modulators of P-glycoprotein activity. Evocarpine, on the other side,
is not transported by P-glycoprotein, and showed only slight toxicity at the highest
test concentration of 30 µM [129]. Furthermore, four antimycobacterial geranylated furocoumarins, from the fruits
of Tetradium daniellii (Rutaceae) showed considerable cell proliferation inhibition with IC50 values ranging from 1.72 to 11.02 µM against CCRF‐CEM and 2.09 to 13.56 µM against
CEM/ADR5000, respectively. The calcein assay to monitor P-glycoprotein function showed
that all four compounds are modulators of P-glycoprotein [130].
In another project, we described the effect of oxalyl-bis(N-phenyl)hydroxamic acid (OBPHA) and copper N-(2-hydroxyacetophenone)glycinate (CuNG) on multidrug-resistant P-glycoprotein-expressing
CEM/ADR5000 T-cell acute lymphoblastic leukemia cells [131]. CuNG, a known depleting agent for glutathione (GSH) and inhibitor of glutathione
S-transferase (GST) and multidrug resistance-related protein 1 (MRP1), also inhibited
P-glycoprotein-mediated doxorubicin accumulation and retention. The resistance-modifying
effects of OBPHA were stronger than that of CuNG. Both novel RMAs overcame drug resistance
more efficiently than verapamil, a well-known P-glycoprotein inhibitor. OBPHA and
CuNG exposure resulted in an increased doxorubicin accumulation after 1–3 h incubation
by downregulation of P-glycoprotein expression after 24 h incubation. This is a clue
that different mechanisms may contribute to modulation of P-glycoprotein-mediated
drug resistance by these compounds.
We investigated the effect of CuNG on reactive oxygen species (ROS) generation and
antioxidant enzymes in normal and doxorubicin-resistant Ehrlich ascites carcinoma
(EAC/Dox)-bearing Swiss albino mice [132]. The effect of CuNG has been studied on ROS generation, multidrug resistance-associated
protein1 (MRP1) expression and on activities of superoxide dismutase (SOD), catalase
(CAT) and glutathione peroxidase (GPx). CuNG increased ROS generation and reduced
MRP1 expression in EAC/Dox cells while only temporarily depleted glutathione (GSH)
within 2 h in heart, kidney, liver and lung of EAC/Dox-bearing mice, which were restored
within 24 h. The level of liver copper was observed to be inversely proportional to
the level of GSH. Moreover, CuNG modulated SOD, CAT and GPx in different organs and
thereby reduced oxidative stress. Hence, nontoxic doses of CuNG may be utilized to
reduce MRP1 expression and thus sensitize EAC/Dox cells to standard chemotherapy.
Moreover, CuNG modulated SOD, CAT and GPx activities to reduce oxidative stress in
some vital organs of EAC/Dox-bearing mice. CuNG treatment also helped to recover liver
and renal function in EAC/Dox-bearing mice.
Furthermore, we have determined the efficacy of CuNG in overcoming multidrug-resistant
cancer using drug-resistant murine and human cancer cell lines [133]. The action of CuNG following single i.m. administration (5 mg/kg body weight) was tested in vivo on doxorubicin-resistant Ehrlich ascites carcinoma (EAC/Dox)-bearing mice and doxorubicin-resistant
sarcoma 180-bearing mice. Tumor size, ascitic load, and survival rates were monitored
at regular intervals. Apoptosis of cancer cells was determined by cell cycle analysis,
confocal microscopy, annexin V binding, and terminal deoxynucleotidyl transferase-mediated
dUTP nick end labeling assay ex vivo. IFN-γ and tumor necrosis factor-α were assayed in the culture supernatants of in vivo and in vitro CuNG-treated splenic mononuclear cells from EAC/Dox-bearing mice and their apoptogenic
effect was determined. Sources of IFN-γ and changes in the number of T regulatory
marker-bearing cells in the tumor site following CuNG treatment were investigated
by flow cytometry. Supernatants of in vitro CuNG-treated cultures of peripheral blood mononuclear cells from different drug-insensitive
cancer patients were tested for presence of the apoptogenic cytokine IFN-γ and its
involvement in induction of apoptosis of doxorubicin-resistant CEM/ADR5000 cells.
CuNG treatment could resolve drug-resistant cancers through induction of apoptogenic
cytokines, such as IFN-γ and/or tumor necrosis factor-α, from splenic mononuclear
cells or patient peripheral blood mononuclear cells and reduced the number of T regulatory
marker-bearing cells while increasing infiltration of IFN-γ-producing T cells in the
ascetic tumor site. Our results show the potential usefulness of CuNG in immunotherapy
of drug-resistant cancers irrespective of multidrug resistance phenotype.
Besides P-glycoprotein, other drug transporters have also been described in recent
years to confer drug resistance. Among them is the breast cancer resistance protein
(BCRP, ABCG2). We have focused on this protein in the context of the highly active
antiretroviral therapy (HAART), whose safety and effectiveness is frequently challenged
by viral resistance to antiretrovirals and the frequent occurrence of drug interactions
which may limit the access of these drugs to the target sites. In particular, drug
distribution and elimination may be modified by BCRP. Therefore, we investigated the
influence of all important anti-HIV drugs on BCRP activity in vitro in one assay to allow unrestricted comparison of the results [134]. BCRP inhibition was assessed by an increase in pheophorbide A accumulation in expressing
MDCKII cells and compared with the corresponding parental cell line MDCKII lacking
human BCRP. According to the IC50 estimation, the rank order for BCRP inhibition was lopinavir > nelfinavir > delavirdine
> efavirenz > saquinavir > atazanavir > amprenavir > abacavir. Whereas nevirapine
and zidovudine exerted weak inhibition, the inhibitory potency for ritonavir and tipranavir
could not be estimated due to their low solubility and all other tested compounds
(indinavir, didanosine, emtricitabine, lamivudine, stavudine, tenofovir and zalcitabine)
were devoid of an effect. This study demonstrated a significant inhibition of BCRP
by many anti-HIV drugs. This suggests that inhibition of BCRP might contribute to
drug-drug interactions observed during HAART in vivo and possibly also the superior effectiveness of combination antiretroviral therapy.
Natural products that inhibit the epidermal growth factor receptor
The epidermal growth factor receptor (EGFR) has an extraordinary relevance in tumor
biology. In our CGH analyses with oral squamous cell carcinoma, the amplification
of the locus 7p12 also points in this direction [39]. A number of novel treatment options for tumors overexpressing EGFR have been launched
to the market in recent years, e.g., therapeutic antibodies and small molecules such
as gefitinib and erlotinib. These two compounds compete with ATP for binding to the
ATP-binding domain of EGFR and, thereby, kill cancer cells. Despite considerable success
with EGFR inhibitors, resistance against these new treatment modalities can also occur
[135] and novel EGFR tyrosine kinase inhibitors are urgently required.
Using the above-described database with 531 compounds derived from TCM [126]; 35 candidate compounds were identified by correlation analyses of their IC50 values and the microarray-based mRNA expression values of the EGFR gene in 60 cell
lines of the NCI [40]. Among the 35 natural products, two have already been reported to be associated
with EGFR function [136]. Then, molecular docking studies (or protein-ligand interaction studies) were carried
out with these 35 compounds using the crystal structure of the EGFR tyrosine kinase
domain as the docking template. As a control, two crystal structures of EFGR tyrosine-kinase
domain with inhibitors (erlotinib and lapatinib, respectively) bound in the ATP-binding
site were taken from the Protein Data Bank (www.rcsb.org). Visual inspection of erlotinib
control docking studies showed that the predicted erlotinib binding site and its orientation
in binding site agreed very well as observed in the crystal structure. The predicted
Ki (pKi) values in our approach also agreed well with the reported values in the literature
for erlotinib (Ki = 0.7 ± 0.1 nM) and gefitinib (Ki = 0.4 ± 0.1 nM) [137].
The 35 selected TCM compounds were individually docked into the crystal structure
of EGFR tyrosine-kinase domain (EGFR‐TK) for the appropriate conformational search.
Eighteen out of 35 natural products were docked in the erlotinib binding site with
docked energy values in the range of − 6.6 to − 10.2 kcal/mol. Among them, neo-oxyberberine,
dicentrine, piceatannol and organol scarlet exhibit similar binding features to that
of erlotinib in the tyrosine-kinase active site of EGFR [40].
Binding site analysis showed that interaction of dicentrine, organol scarlet and erlotinib
with residues Thr766 and Met769 was a conserved feature. Thr766 and Met769 formed
a hydrogen bond with the quinazoline nitrogen atom of erlotinib, while they were involved
in hydrogen bond formation with the benzodioxoloquinoline oxygen atom of dicentrine.
In the case of organol scarlet, Thr766 and Met769 formed a hydrogen bond with the
diazenyl nitrogen atom of organol scarlet that is not a part of any cyclic system
as found in erlotinib [40]. Further investigations are underway to analyze these findings in more detail.
Conclusions and Perspectives
Conclusions and Perspectives
Our data indicate that a great diversity of molecular mechanisms is operative in clinical
drug resistance. We showed that different resistant profiles exist within tumors of
homogeneous histology. Thus, it is possible to identify novel subgroups of otherwise
homogeneous tumor groups. These results are in accordance with a large body of evidence
in the literature [138], [139], [140], [141], [142], [143], [144]. The systematic investigation of combinations of cellular factors in cancer clearly
yields improved predictive information. Recently developed technologies for genome-,
transcriptome- or proteome-wide analyses facilitates the simultaneous analyses of
thousands of genes or proteins in a single experiment, raising expectations that it
will revolutionize cancer diagnosis.
On the other hand, the results of our group as well as of other authors [145] indicate that a minimal set of about 10 to 50 factors may be sufficient and may
bring more robust results than sets of thousands of factors. Therefore, it is reasonable
to focus on a few relevant prognostic factors which may serve as drug targets to treat
individual patients exhibiting these specific poor prognostic factors. We consider
these results as one step further to the ultimate goal of prediction of drug response
of each individual patient. We have exemplarily done this for three target proteins,
i.e., MTAP, P-glycoprotein and EGFR and showed the principal feasibility of this concept.
In the long run, it has to be seen whether genomics and proteomics along with other
“-omic” technologies will provide real hope or just another hype. It has to be taken
cautiously, bearing in mind that the response to chemotherapy depends not only on
intracellular, e.g., molecular and biological factors, but also on extracellular factors.
The relevance of pharmacokinetics and dynamics must not be underestimated.
Generally, our knowledge on cancer biology and molecular factors with prognostic significance
has exponentially grown over the past two decades delivering a wide array of proteins
that may serve as target structures for drug development. This will for sure lead
to a tremendous increase in novel compounds to be tested in the years to come, and
small molecules will be a major player in this scenario.
Given the fact that a considerable portion of all drugs used nowadays is from natural
origin, natural products represent a valuable source for drug development. Chemical
compounds developed over millions of years during evolution bear a tremendous potential
as novel drugs. The hope is that this potential will be utilized for the sake of the
cancer patients.
We will face a shift from the established, rather unspecific and toxic anticancer
drugs to new generations of targeted drugs with improved pharmacological features
concerning tumor specificity and toxicity. At the moment it cannot be foreseen whether
cancer patients will routinely be curable. Rather, cancer might turn from a life-threatening
into a chronic disease, which needs life-long therapy. This raises concerns, and one
might ask “who should pay?” The only way to reduce costs is to avoid overtreatment
of patients. Personalized medicine will provide sophisticated diagnostic tools to
adjust the right treatment for the right patient. As a consequence, it is to be hoped
that treatment failure and unwanted side effects will more and more disappear and
treatment efficacy will raise improving quality of life.
