Aldehyde Dehydrogenase Discriminates the CD133 Liver Cancer Stem Cell Populations

In the last several years, accumulating evidence has been lending support to the idea that human cancer can be considered a stem cell disorder. According to the hierarchical “stem cell model of carcinogenesis,” tumors are not to be viewed as simple monoclonal expansions of transformed cells, but rather as complex tissues where abnormal growth is driven by a rare cancer stem cell (CSC) pool that has not only acquired tumor-like features such as uncontrolled growth but also maintains its innate capacity to self-renew and differentiate into a phenotypically heterogeneous progeny (1). CSCs, like normal stem cells, are also generally believed to possess unique survival mechanisms. And in view of their stem like properties, these cells are believed to be responsible for the maintenance and propagation of the entire tumor. Therefore, to completely eradicate a tumor and prevent recurrence, it is imperative that CSCs be specifically targeted. To date, CSCs have been identified as the driving force behind tumor formation in several tumor types including those of leukemia (2), cancers of the breast (3), brain (4), and prostate (5), and more recently, cancers of the colon (6) and liver (7-10).

Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide, affecting 1 million individuals annually. Despite extensive research efforts, there is little understanding of the characterization of the specific group of cells that are the target of tumorigenic transformation. Recently, we and others have identified, characterized, and isolated a CD133⁺ CSC-like subpopulation from human HCC cell lines and xenograft tumors. Studies in our laboratory found that CD133⁺ HCC cells not only possess a greater colony-forming efficiency, higher proliferative output, and greater ability to form tumor in vivo but are also endowed with characteristics similar to those of stem/progenitor cells including the preferential expression of “stemness” genes, the ability to self-renew, and the ability to differentiate into non–hepatocyte-like, angiomyogenic-like lineages. CD133 expression is found to represent only a minority of the tumor cell population in human HCC specimens (8-10). In addition, CD133⁺ cells are also more resistant to conventional chemotherapeutic drugs such as doxorubicin and 5-fluorouracil, the underlying mechanism of which required the preferential activation of the Akt/protein kinase B and Bcl-2 survival pathways (11).

Our recent findings were based on experiments done primarily on HCC cell lines Huh7 and PLC8024 or xenograft tumors derived from CD133⁺ (isolated from Huh7) cells, where expression of CD133 was detected at levels as high as 65%. Yet, based on the CSC theory, it is believed that CSCs only represent a minority number of the entire tumor mass (<5% of the total tumor mass), and if true, may be indicative that the CD133⁺ HCC CSC population is still heterogeneous, consisting of perhaps subsets of cells with differing tumorigenic potential. In fact, CSCs in other cancer types have also been characterized by more than one marker [e.g., CD44⁺CD24^−/low in breast (3), CD44⁺α₂β₁^hiCD133⁺ in prostate (5), and CD133⁺CD166⁺ in colon (6)]. For this reason, identification of other markers, expressed along with CD133, with the goal of better characterizing the CSC population in HCC, is much needed and was our chosen focus in this present study. Using a proteomic approach, differential profiles of CD133⁺ and CD133⁻ cells isolated from HCC cell line Huh7 were analyzed by two-dimensional PAGE. Candidate protein spots were selected for identification by mass spectrometry (MS) analysis followed by the study of their potential correlation to CD133⁺ cells and their prospective role as liver CSC markers. Our results reveal, for the first time, that aldehyde dehydrogenase (ALDH), expressed along CD133, can more specifically characterize the liver CSC population.

To identify candidate genes differentially expressed in CD133⁺ and CD133⁻ subpopulations, two-dimensional PAGE analysis was done to compare protein expression profiles between the two groups of cells sorted from the HCC cell line Huh7. Nearly 300 protein spots were resolved in each two-dimensional gel, with the majority of the spots matching between the two gels of CD133⁺ and its corresponding CD133⁻ counterparts (data not shown). Those spots that were significantly differentially expressed were selected for further identification by MS analysis. A total of 10 spots, all found to be up-regulated in the CD133⁺ subpopulation, were selected (Fig. 1; Table 1). Following MS/MS analysis, the matching score of all 10 selected proteins were found to be statistically significant, indicating identity or extensive homology (P < 0.05) to their proposed protein identity. Of these, aldehyde dehydrogenase 1A1 (ALDH1A1) was chosen for further study.

To confirm the preferential expression of ALDH1A1 found predominantly in the CD133⁺ subpopulation as detected by two-dimensional PAGE analysis, both quantitative PCR (Fig. 2A, first two column sets) and Western blot analyses (Fig. 2B) were carried out to confirm the results at both the genomic and proteomic levels, respectively. In accordance with our two-dimensional PAGE results, ALDH1A1 was found predominantly in the CD133⁺ subpopulation in cells isolated from both PLC8024 and Huh7 HCC cell lines.

To determine whether ALDH1A1 is implicated in hepatocarcinogenesis or whether it is correlated with CD133 expression, we proceeded with examining the expression of ALDH1A1 in a panel of liver cell lines, including hepatoblastoma cell line HepG2 and HCC cell lines Huh7 (12), H2P, H2M (13), PLC8024, Hep3B, 7701, 7703, and 7402, by quantitative PCR analysis. The trend of ALDH1A1 expression correlated with expression of CD133 in that those cell lines that expressed little CD133, like HepG2 and H2P, correspondingly expressed fewer ALDH1A1, and those cell lines that expressed high levels of CD133, such as Huh7, PLC8024, and Hep3B, also correspondingly expressed high levels of ALDH1A1 (Fig. 2C). Subsequent studies for ALDH activity in liver cell lines by the novel ALDEFLUOR technique were consistent with our quantitative PCR data, showing a direct correlation between ALDH activity and CD133 expression (Fig. 2D; Table 2). Because the substrate used for the ALDEFLUOR assay is known to be processed by ALDH1A1, as well as by other ALDH genes in the superfamily, we also went on to examine by quantitative PCR whether other ALDH isoforms (ALDH1A2, ALDH1L1, ALDH2, ALDH3A1, and ALDH9A1) were also differentially expressed in CD133⁺ and CD133⁻ cells sorted from HCC cell lines Huh7 and PLC8024 (Fig. 2A). Selection of which ALDH isoforms to test was based on its relevance to cancer, drug resistance, or its expression in human or in the liver. Results indicate that isoforms ALDH1A2 and ALDH3A1 also seem to be preferentially expressed in the CD133⁺ subfractions. Because we cannot rule out that cells sorted by labeling with the ALDEFLUOR assay exclusively contain ALDH1A1^+/− subpopulations only, we chose to address this population as the more general ALDH, rather than ALDH1A1.

To determine the immunophenotype of ALDH⁺ HCC cells in liver cell lines, we conducted dual-color flow cytometry analysis for both CD133 expression and ALDH activity. As evident in Fig. 3, the majority of ALDH⁺ cells gated in R2 (the entire ALDH⁺ population) were CD133⁺. It is also interesting to note that a subfraction of CD133⁺ cells were indeed ALDH⁻, which indicated that ALDH activity may possibly be used to further enrich the CD133⁺ subpopulation and can potentially be used to better characterize the representative liver CSC population. Table 3 summarizes the expression of CD133⁺ALDH⁺, CD133⁺ALDH⁻, CD133⁻ALDH⁺, and CD133⁻ALDH⁻ in liver cell lines that were found to express ALDH, including HepG2, Huh7, PLC8024, Hep3B, and H2M, when dually stained with CD133-phycoerythrin and for ALDH activity.

To determine the tumorigenic potential of different subpopulations of cells dually stained for CD133 expression and ALDH activity, CD133⁺ALDH⁺, CD133⁺ALDH⁻, or CD133⁻ALDH⁻ cells were isolated from HCC cell line PLC8024 (Fig. 4A) and subsequently subjected to both clonogenic and proliferative assays in vitro (Fig. 4B and C) and tumorigenic experiments in vivo (Fig. 4D; Table 4). Unsorted PLC8024 and CD133⁺ cells isolated from PLC8024 were incorporated as controls. CD133⁺ALDH⁺ cells were able to proliferate significantly faster, induce bigger and greater number of tumor colonies, and induce tumor formation with fewer numbers of cells and in a significantly lesser time than their respective CD133⁺ALDH⁻ and CD133⁻ALDH⁻ counterparts or CD133⁺ cells alone. As few as 500 CD133⁺ALDH⁺ cells were sufficient for consistent tumor development in severe combined immunodeficient (SCID) mice, whereas at least 1,000 CD133⁺ cells or 100,000 CD133⁺ALDH⁻ cells were necessary to consistently generate a tumor in the same model, and not to mention, requiring a significantly longer period of time. Tumor nodules that formed subcutaneously in the mice were confirmed by pathologic study (data not shown).

The growing understanding of stem cell and cancer biology has led to the clear concept that stem cells play a critical role in the development of cancer. According to the CSC hypothesis, not all cells in a tumor are equal and that tumors are hierarchically organized and are composed of a heterogeneous population of cells where a rare population called CSCs sit at the top. These cells possess unique survival mechanisms and distinctive stem cell properties, including the ability to self-renew and differentiate, and also exhibit a marked ability to proliferate following a prolonged period of quiescence.

Our work (10, 11), along with the work of other research groups (8, 9), suggests that human HCC is very likely to be also hierarchically organized and originates from a primitive stem/progenitor group of cells for which CD133⁺ precursors constitute one of the most immature stage. Results from our previous studies were mainly derived from experiments done on HCC cell lines or xenograft tumors where CD133 expression was as high as 48% to 65% (It was recently observed that HCC cell lines Huh7 and PLC8024 express varying levels of CD133 expression, ranging from 48% to 60% for PLC8024 and from 49% to 65% for Huh7. The authors speculate that the cause for this variance in CD133 expression may partially be dependent on the confluence of cell culture conditions; but the reason remains elusive). Based on the CSC theory, it is believed that CSCs should only represent a minority number of the entire tumor mass; if this is true, it is unlikely that all 65% of the CD133⁺ cells in PLC8024 and Huh7 cell lines are CSCs and is also indicative that the CD133⁺ HCC CSC population is still heterogeneous, consisting of perhaps subsets of cells with differing tumorigenic potential. In fact, CSCs in other cancer types have also been characterized by more than one marker. Keeping in mind that the CD133⁺ HCC CSC population is heterogeneous, we seek to further dissect tumorigenic HCC subsets in the current study. Here, we extended our study to identify the differentially expressed protein profiles of CD133⁺ and CD133⁻ cells sorted from the HCC cell line Huh7 by two-dimensional PAGE technique. From the list of identified proteins, ALDH1A1 was chosen for further study. The potential correlations of other identified differentially expressed proteins with CD133 expression and their possible role as liver CSC markers are also currently being investigated in our laboratory.

ALDH is a cytosolic enzyme that is responsible for the oxidation of intracellular aldehydes. To date, more than 17 human ALDH isoforms have been identified to belong in the ALDH superfamily, of which most are highly conserved across a variety of species (14). Recent studies have shown that ALDH is also highly expressed in human and murine hematopoietic stem/progenitor cells (HSC; refs. 15, 16). In particular, a functional role of ALDH was recently shown, in which diethylaminobenzaldehyde, a specific inhibitor of ALDH, deregulated human HSC self-renewal by interfering with endogenous retinoic acid biosynthesis (17). These data suggested that ALDH may regulate HSC function and is a potential therapeutic target. More importantly, ALDH has also recently been shown to play a pivotal role in mediating chemoresistance in CSCs where it acts to detoxify harmful substances like aldehydes that would otherwise insult these crucial stem/progenitor cells. This is based on the theory that a number of chemotherapy drugs work through conversion in the body to an aldehyde that attacks cancer cells. These therapies lose potency over time because the relevant ALDH increases in concentration in cancer, deactivating the aldehyde more quickly, thus rendering these therapies ineffective. In particular, ALDH bright cells from human umbilical cord blood and peripheral blood leukemic stem cells have been shown to contain all the nonobese diabetic/SCID engrafting cells in CD34⁺ cells (18, 19). In addition, recent research efforts have shown that ALDH is highly expressed in non–small-cell lung, colon, and pancreatic cancers (20-22), and that higher ALDH activity levels can be used to better identify the colon CSC population with EpCAM^highCD44⁺ phenotype (6), as well as the glioblastoma (23), retinoblastoma (24), and breast (25) CSC populations. In view of these recent advances, it was both reasonable and logical to speculate that ALDH1A1 or the more general ALDH may possibly play an important role in the better characterization of CD133⁺ HCC CSCs.

Subsequent studies in the present work found ALDH to be differentially expressed in the CD133⁺ subfraction of both HCC cells lines Huh7 and PLC8024. Analysis of ALDH expression and enzymatic activity in a panel of liver cell lines found ALDH to be positively correlated with CD133 expression. Further analysis with dual-color flow cytometry for both CD133 expression and ALDH activity found the majority of ALDH⁺ cells in cultured HCC cell lines to be CD133⁺. Most importantly, not all CD133⁺ HCC cells were ALDH⁺, indicating that ALDH can possibly be used to further enrich the CD133⁺ subpopulation and thus, in turn, better characterize the representative liver CSC population. In addition, purified CD133⁺ALDH⁺ cells were found to be significantly more tumorigenic than their respective CD133⁺ALDH⁻ or CD133⁻ALDH⁻ counterparts, both in vitro and in vivo, in that they were able not only to proliferate significantly faster but also to induce bigger and greater number of tumor colonies or nodules with less number of cells in a significantly shorter period of time.

Thus, in summary, our work provides concrete experimental evidence showing that ALDH, when coexpressed with CD133, can more specifically characterize the liver CSC cell population. The work presented is novel in that not only is it the first study showing the use of ALDH in the identification of HCC CSCs but it is also the first study showing the preferential ALDH enzymatic activity in the CD133⁺ subpopulation. The more specific identification of the tumorigenic liver CSC population will shed new light not only on the cellular origin of HCC but also on the underlying biological mechanism of this specific CSC population and, thus, will ultimately lead to the development of more specific therapeutic agents for the treatment of this deadly cancer.

Hep3B was obtained from the American Type Culture Collection. Huh7 was provided by Dr. H. Nakabayashi (Hokkaido University School of Medicine, Sapporo, Japan; ref. 12). QGY-7701, QGY-7703, BEL7402, HepG2, and PLC8024 were obtained from the Institute of Virology, Chinese Academy of Medical Sciences (Beijing, China). H2P and H2M were previously established in our laboratory (13).

ALDEFLUOR reagent (Stem Cell Technologies, Inc.) was used for the immunofluorescent detection of intracellular ALDH enzyme activity. In some experiments, cells were costained with phycoerythrin-conjugated antihuman CD133/1 antibody (Miltenyi Biotec). In the latter case, isotype-matched mouse immunoglobulins were incorporated as controls. All samples were analyzed and sorted with BD FACSVantage SE cell sorter and CellQuest software (BD Biosciences). Aliquots of CD133⁺ALDH⁺, CD133⁺ALDH⁻, and CD133⁻ALDH⁻ sorted cells were evaluated for purity with a FACSCalibur (BD Biosciences) using the ALDEFLUOR Kit (Stem Cell Technologies) and phycoerythrin-conjugated antihuman CD133/2 antibody (Miltenyi Biotec).

For the first dimension, 150-μg proteins were separated on a precast immobilized pH gradient strip (pH 4-10, 18 cm; Amersham Biosciences). Separation in the second dimension was done on 12.5% polyacrylamide/bis gel. Gels were fixed overnight and stained with PlusOne Silver Staining Kit (Amersham Biosciences) according to the manufacturer's protocol. Stained gels were scanned with an ImageScanner and analyzed and quantified with ImageMaster 2D Platinum software (Amersham Biosciences). For comparison of protein levels among gels, normalization was done against the total intensity of all spots present in the gel. Only those significantly different spots (2-fold increase or decrease) were selected for analysis by MS.

Proteins of interest were excised from gels and digested with trypsin. MS analyses were done on an ABI4700 Voyage-DE STR matrix-assisted laser desorption/ionization-time of flight MS (Applied Biosystems). Generated data were searched via the Mascot search engine to obtain protein identity by undertaking the peptide mass fingerprinting approach.

Quantified protein lysates were resolved on SDS-PAGE gel, transferred onto a polyvinylidene difluoride membrane (Millipore), and immunoblotted with mouse anti-human CD133/1 (Miltenyi Biotec), goat anti-human ALDH1A1 (Santa Cruz Biotechnology), or goat anti-human β-actin (Santa Cruz Biotechnology), followed by incubation with secondary horseradish peroxidase. Blots were visualized by enhanced chemiluminescence (Amersham Biosciences).

Total RNA was isolated with TRIzol reagent (Invitrogen), and complementary DNA (cDNA) was synthesized with an Advantage RT-for-PCR Kit (Clontech Laboratories) according to the manufacturer's instructions. cDNA was subjected to quantitative PCR with a SYBR Green PCR Kit (Applied Biosystems) using primers listed in Supplementary Table S1. Amplification protocol consisted of incubations at 95°C for 15s, 60°C for 1 min, and 72°C for 1 min for 40 cycles. Quantification was done using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems). All quantitative PCR reactions were done in triplicate. The C_T of each gene of interest [C_T(gene of interest) test] in each sample was normalized with control β-actin [C_T(β-actin) test] for RNA amount variation and calibrator for plate-to-plate variation using the following formula: ΔC_T(test) = C_T(gene of interest) test − C_T(β-actin) test. Relative expression level was presented as the relative fold change and calculated using the following formula: 2^−ΔΔC_T = [C_T(gene of interest) − C_T(β-actin)]test − [C_T(gene of interest) − C_T(β-actin)]calibrator.

Experiments were done as previously described (10).

Experiments were done as previously described (10).

The study protocol was done in accordance with the Committee of the Use of Live Animals in Teaching and Research at The University of Hong Kong. SCID mice, ages between 4 and 8 wk, were used to test the tumorigenicity potential of sorted cells. Observation for growth of tumors was done as previously described (10).

Statistical analysis was done by applying the independent t test using Microsoft Office Excel software. Statistical significance was declared if P < 0.05.

No potential conflicts of interest were disclosed.

We thank Fai Ng of the Department of Microbiology of the University of Hong Kong for his expert help with cell sorting analysis.