Long-term culture of human embryonic stem cells in feeder-free conditions



We have demonstrated previously that human embryonic stem (hES) cells possess a characteristic morphologic, antigenic, and molecular profile that can be used to assess the state of ES cells (Carpenter et al., [2004] Dev Dyn 229:243–258). In this manuscript, we have examined the long-term stability of three hES cell lines in feeder-free culture. We demonstrate that the expression of antigens and transcription factors, telomerase activity, telomere length, and karyotype appear stable for all three hES cell lines after continuous culture for over 1 yr. All three lines retained pluripotent differentiation in vitro and in vivo. Although hES cell lines were remarkably stable over the period of analysis, a detailed quantitative analysis of antigen expression by flow cytometry and gene expression by microarray suggested that cell lines show subtle differences in the expression of small subsets of genes upon long-term culture. Developmental Dynamics 229:259–274, 2004. © 2003 Wiley-Liss, Inc.


Human ES (hES) cell lines were first isolated in 1998 and represent a cell population that has indefinite replicative capacity and can differentiate into endoderm, mesoderm, and ectoderm (Thomson et al., 1998; Reubinoff et al., 2000). While fewer than 10 cell lines have been well characterized, the data suggest that different cell lines show similar marker expression and capacity to differentiate (reviewed in Carpenter et al., 2004). Because they are a pluripotent cell population that can be directed to differentiate to specific phenotypes, hES cells may serve as a universal bank of cells from which specific cells may be isolated for cell replacement therapies. Several investigators have now shown that hES cell differentiation can be directed/biased toward a particular phenotype (Carpenter et al., 2001; Kaufman et al., 2001; Xu et al., 2002b; Rambhatla et al., 2003; and others). In addition, cells of a particular phenotype can be mechanically enriched from heterogeneous differentiated hES cell populations (Carpenter et al., 2001; Levenberg et al., 2002; Xu et al., 2002a). Mouse ES (mES) cell derivatives can be transplanted into the adult brain where they will survive, integrate, and provide functional improvement (McDonald et al., 1999; Bjorklund et al., 2002; Kim et al., 2002). Although hES cell derivatives have been engrafted (Zhang et al., 2001; Reubinoff et al., 2001), in vivo function has not yet been demonstrated. However, given the reported success using mES cells, the scientific community is hopeful that the hES-derived cells will also prove effective in animal models of disease. Before using hES cells for therapy, several aspects of hES cell biology must be evaluated. Conditions to propagate hES cells in an undifferentiated state should be identified; predictive markers of the undifferentiated cultures need to be established; feeder- free cultures qualified for human transplants need to be obtained; and, above all, the long-term stability of such cell lines needs to be determined.

hES cells have a doubling time of 35–40 hr and extensive cell expansion will be required to obtain the numbers of cells required for human therapy. Maintaining a large pool of uncommitted stem cells in a physiologically normal state presents a challenge as several factors act to alter the differentiation state of cells in culture. Stem cells may undergo stochastic differentiation, and differentiated cell populations may possess a faster division rate or adapt better to culture conditions and, thus, may be progressively selected in continuous culture. Cells can undergo mutations at a small but measurable rate and transformed cells may possess enhanced growth or proliferation characteristics and may gradually come to represent the dominant population of cells in culture. Therefore, continuous assessment of the cell population over time is critical.

ES cells differ from other tissue specific stem cell populations in their ability to maintain a stable phenotype after prolonged periods in culture. Data from mouse ES (mES) cell cultures have suggested that mES cells differ from other cell types in being relatively stable and possessing an indefinite replicative capacity (Suda et al., 1987). This has been demonstrated most dramatically in the mouse system by the generation of chimeras. In experiments where mES cells maintained in culture for long time periods were injected into mouse blastocysts, it was possible to show that these cultured mES cells contributed to many of the tissues in the resulting chimeric animal (Suda et al., 1987). ES cells contributed to the germ line as well, such that the resulting second-generation offspring were genotypically derived from the original cultured mES cells. Because hES cells were isolated at a similar stage in development by using similar isolation procedures, it is reasonable to assume that hES cells will possess a similar tissue culture stability and self-replicative capacity.

Several lines of evidence argue that this assumption must be tested rigorously. Even murine ES cell lines, which are remarkably stable, appear to undergo functional changes over time in culture (Smith, 2001a, b). Until recently, it has been possible to generate stable mES cell lines from only C57/BL6 and SJ129 strains of mice (Smith, 2001a). However, cell culture conditions have been identified which allow the derivation of mES cell lines from several inbred mouse strains (Schoonjans et al., 2003). These findings suggest that hES cells may also be sensitive to particular culture conditions and the stability of hES cell lines must be examined over long-term culture.

To directly test the stability, self-renewal characteristics, and differentiation ability of hES cell lines, we maintained three hES lines derived in one laboratory (Thomson et al., 1998) in continuous culture for up to two years and periodically tested their growth characteristics, marker expression, pluripotency, and karyotypic stability at multiple passages. We show that all three hES lines can maintain a karyotypically stable phenotype for more than one year. Cells continued to grow at similar rates, expressed markers of undifferentiated cells, maintained high telomerase levels, and were capable of teratoma formation. Some differences in telomere length and small differences in SSEA expression and gene expression were noted, and a moderate frequency of aneuploidy was seen. Overall, the cells appeared remarkably stable over this time period, suggesting that hES cells can serve as a virtually unlimited source of differentiated cells for therapy.


Morphologic Assessment

hES cells maintained in feeder-free culture grow as colonies of undifferentiated cells surrounded by differentiated, stroma-like cells (Xu et al., 2001). As the hES cells were maintained in long-term culture, we have observed that the undifferentiated colonies represent a larger portion of the culture. In some cases, the long-term cultures appear completely devoid of differentiated cells, and the undifferentiated cells appear as a tightly packed monolayer (data not shown).

Marker Expression Over Long-Term Culture

hES cells express several characteristic surface markers such as SSEA-4, TRA-1-60, and TRA-1-81. We have reported previously that different hES cell lines have similar expression of these markers (Carpenter et al., 2003a). We evaluated marker expression in three cell lines (H1, H7, and H9) by assessing many separate cultures of each cell line over long-term culture (Figs. 1, 2). In general, all of the cell lines maintained expression of the characteristic markers. Figure 1 shows SSEA-4 immunostaining of H9 colonies at p38 (Fig. 1A,B) and p114 (Fig. 1C,D). In addition, flow cytometric analysis showed high expression of SSEA-4, TRA-1-60, and TRA-1-81 at early and late passages (H9 p33 and H9 p81, Fig. 1E).

Figure 1.

Expression of surface markers on high- and low-passage human embryonic stem (hES) cell lines. A–D: H9 hES cells were maintained in MEF-conditioned medium supplemented with basic fibroblast growth factor on growth factor reduced (GFR) matrigel (passaged at 7-day intervals) and analyzed for SSEA-4, Tra-1-60, Tra-1-81 expression. SSEA-4 expression in H9 p38 (A,B) and H9 p114 (C,D). E: Flow cytometric detection of SSEA-4, Tra-1-60, and Tra-1-81 on H9 hES cells maintained in feeder-free conditions for 33 and 81 passages. Nonviable cells were identified by using propidium iodide and excluded during analysis. The horizontal marker in each plot is positioned to represent Tra-1-60 or Tra-1-81 high population; the vertical marker is positioned to represent SSEA-4 high population. Scale bars = 200 μm in A,C, 20 μM in B,D.[Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 2.

A: Maintenance of surface marker expression in extended cultures of H1 (circles), H7 (triangles), and H9 (squares) human embryonic stem (hES) cell lines. Each symbol represents measurement of a separate culture at a different point in time. Passages in culture are indicated on the horizontal axis. Cultures that were measured earlier than passage 29 are shown in the first column of the graph. Measurements of cultures at passage 30–39, 40–49, etc., are presented in the remaining columns, as indicated. Cells were maintained in MEF-conditioned medium supplemented with basic fibroblast growth factor (bFGF) on growth factor reduced (GFR) matrigel for over 70 passages. Cultures were harvested for flow cytometric analysis upon reaching confluence (day 7). Only viable cells were included in this analysis; positive events were defined as having greater fluorescence than 99% of the mean of the appropriate isotype matched control. Passages in culture are indicated on the horizontal axis. B: Maintenance of SSEA-4high, Tra-1-60high, and Tra-1-81high populations in extended cultures of H1 (circles), H7 (triangles) and H9 (squares) hES cell lines.

Assessment of individual markers did not show large changes in the percentage of cells expressing SSEA-4, SSEA-1, TRA-1-60, TRA-1-81, CD9, or CD133 over extended culture. Figure 2 shows the percentage of cells in many separate cultures expressing each marker at early and late passages and Table 1 displays the mean ± SEM and of the percentage of cells in each cell line expressing the different markers over time in culture. The percentage of cells expressing SSEA-4, TRA-1-60, or TRA-1-81 showed a slight increase (Fig. 2A). Although some of these increases were statistically significant, the magnitude of the changes was relatively small (70–90% in early passages to >90% in late passages).

Table 1. Comparison of Surface Marker Expression on H1, H7, and H9 hES Cell Lines at Different Passagesa
  • a

    Values indicate mean ± SEM; sample size is in parentheses. hES, human embryonic stem.

  • *, * (4)

    Significantly different than <passage 29, P<0.05.

  • **

    Significantly different than P30-39, P<0.05.

  • ***

    Significantly different than P40-49, P<0.05.

 SSEA-489.3 ± 1.3 (6)94.9 ± 1.2* (10)95.6 ± 0.5* (4)88.22 ± 5.1 (5)93.0 ± 4.4* (4) 
 TRA-1-6087.4 ± 2.4 (4)92.5 ± 1.3 (8)91.5 ± 3.3 (3)91.2 ± 2.7 (5)96.5 ± 1.7* (4) 
 TRA-1-8179.7 ± 2.0 (4)94.6 ± 1.8 (6)91.6 ± 2.9* (4)91.5 ± 5.9* (3)92.0 ± 4.4* (4) 
 CD974.0 ± 7.5 (6)83.0 ± 5.6 (7)87.2 ± 5.6 (2)78.5 ± 14.1 (3)74.2 ± 11.6 (4) 
 AC13354.9 ± 2.1 (3)61.9 ± 3.7 (7)58.7 (1)44.0 ± 26.3 (2)61.6 ± 1.8 (2) 
 SSEA-11.5 ± 1.7 (6)4.1 ± 1.7 (8)5.7 ± 1.7 (3)2.9 ± 1.5 (3)4.9 ± 2.8 (4) 
 SSEA-4high67.5 ± 1.7 (6)88.4 ± 2.2* (10)82.2 ± 3.2* (4)79.4 ± 6.9 (5)87.2 ± 5.3* (4) 
 SSEA-4high/TRA-1-60high54.4 ± 2.3 (4)63.3 ± 3.5 (7)64.8 ± 3.4* (3)62.2 ± 8.6 (5)69.2 ± 8.0 (4) 
 SSEA-4high TRA-1-81high51.9 ± 5.7 (4)84.8 ± 3.6* (5)61.2 ± 11.3 (4)74.9 ± 13.1 (3)76.9 ± 7.1* (4) 
 SSEA-495.4 ± 1.0 (4)95.0 ± 1.6 (6)96.3 ± 0.7 (7)88.6 ± 8.0 (3)97.3 ± 2.2 (2)98.1 ± 0.3 (4)
 TRA-1-6087.2 ± 2.6 (4)93.1 ± 1.5 (6)95.1 ± 1.9 (3)93.4 ± 1.8 (3)98.7 ± 0.5* (2)98.1 ± 0.8* (4)
 TRA-1-8177.9 ± 3.7 (4)93.9 ± 1.9* (4)96.4 ± 0.7* (7)89.0 ± 9.7 (2)98.7 ± 0.6* (2)99.0 ± 0.1* (4)
 CD951.6 ± 20.2 (2)88.9 ± 6.1* (6)92.9 ± 1.4* (5)87.5 ± 7.8 (3)96.2 ± 0.1 (2)96.1 ± 1.9* (4)
 AC133 62.6 ± 4.7 (5)68.4 ± 5.8 (5)81.6 ± 1.1 (2)44.3 ± 7.4 (2)53.0 ± 3.9 (4)
 SSEA-118.6 ± 7.0 (4)6.3 ± 1.5 (6)4.9 ± 3.0 (3)1.5 ± 0.3 (2)0.9 ± 0.5 (2)1.6 ± 0.4 (4)
 SSEA-4high73.7 ± 0.7 (2)84.4 ± 3.9 (5)96.3 ± 0.8*,**(6)83.6 ± 6.6 (4)98.1 ± 0.1* (2)96.5 ± 1.1* (4)
 SSEA-4high/TRA-1-60high62.1 ± 1.3 (3)73.9 ± 6.6 (5)88.3 ± 3.8 (4)73.1 ± 5.4 (4)95.3 ± 0.1* (2)86.9 ± 7.7* (4)
 SSEA-4high TRA-1-81high36.0 ± 6.1 (2)83.7 ± 8.4 (3)93.4 ± 1.6 (6)75.1 ± 11.0 (3)97.9 ± 1.5 (2)97.0 ± 0.4 (4)
 SSEA-439.3 ± 11.3 (8)90.2 ± 1.5* (14)90.5 ± 0.9 (2)94.8 ± 2.8* (5)97.4 ± 0.7*,***(2)97.7 ± 0.7*,**,***(9)
 TRA-1-6089.6 ± 2.1 (8)91.3 ± 2.2 (13)96.3 (1)97.0 ± 1.1* (5)94.6 ± 1.7 (2)96.5 ± 1.3* (9)
 TRA-1-8173.4 ± 6.8 (8)90.5 ± 1.8 (11) 96.9 ± 1.1 (4)95.2 ± 0.1 (2)98.4 ± 0.3 (9)
 CD984.3 ± 3.1 (8)90.0 ± 3.0 (11)96.9 (1)93.2 ± 2.9 (5)75.3 ± 1.5 (2)98.3 ± 0.5 (9)
 AC13373.3 ± 2.6 (3)68.0 ± 2.9 (10)72.9 (1)60.7 ± 9.7 (4)60.6 ± 3.9 (2)59.4 ± 4.5 (9)
 SSEA-14.7 ± 1.5 (8)2.0 ± 0.6 (12)5.2 (1)1.4 ± 0.7 (5)0.9 ± 0.8 (2)88.3 ± 9.0 (9)
 SSEA-4high38.8 ± 9.1 (12)74.0 ± 6.9 (8)83.6 (1)79.6 ± 12.9 (5)88.1 ± 3.0 (2)88.3 ± 9.0 (9)
 SSEA-4high/TRA-1-60high68.9 ± 4.6 (9)74.8 ± 4.7 (11) 89.5 ± 1.9 (5)80.9 ± 5.2 (2)81.8 ± 6.7 (9)
 SSEA-4high TRA-1-81high21.3–9.3 (8)58.5 ± 7.2 (10) 70.0 ± 14.7 (4)85.5 ± 1.7 (2)86.1 ± 8.5 (9)

Closer analysis of the SSEA-4 expression showed that a portion of the cells express SSEA-4 at high levels (SSEA-4high) and a smaller portion express SSEA-4 at low levels (SSEA-4low; described in Carpenter et al., 2003). TRA-1-60 and TRA-1-81 had similar distribution patterns to SSEA-4, resulting in high- and low-expressing populations. This allowed us to determine that hES cell cultures contained a population of cells coexpressing SSEA-4 and TRA-1-60 or TRA-1-81 at high levels (SSEA-4high/TRA-1-60high or TRA-1-81high populations). This population represented the majority of the cells in undifferentiated cultures and correlated well with the undifferentiated morphology.

Evaluation of TRA-1-60 and TRA-1-81high/SSEA-4high cells revealed that in the H7 and H9 cell lines these subpopulations showed a statistically significant increase between early and late passages (H7 <p30 vs. >p70 P = 0.04, H9 <p30 vs. >p60–69 P = 0.04; Table 1). This finding appeared to correlate with the observed changes in morphology of the hES cell cultures. As mentioned above, as the cells are maintained in long-term culture, the amount of differentiated cells diminishes while the undifferentiated cells represent the majority of the culture.

Evaluation of fibroblast growth factor receptor (FGFR), CD117 (c-kit), CD135 (Flt3), and CD130 (gp130) demonstrated no significant differences in overall expression of these growth factor receptors (Fig. 3). Expression of these growth factor receptors on the SSEA-4high population did not change over long-term culture, while expression on the SSEA-4low/neg population was difficult to evaluate due to the diminishing number of these cells in later passages.

Figure 3.

A: CD130, CD117, CD135, and fibroblast growth factor receptor (FGFR) expression on low- and high-passage human embryonic stem (hES) cells. Cultures of H1 cells maintained in MEF-conditioned medium with basic fibroblast growth factor on growth factor reduced (GFR) matrigel for 27 and 68 passages were harvested for staining at confluence. Nonviable cells were identified by using propidium iodide and excluded during analysis. In top plots, open histograms represent IgG3 isotype control, and black histograms represent SSEA-4 staining. In other plots, gray histograms represent appropriate isotype staining and open histograms represent corresponding antibody staining. B: Table of means for CD130, CD117, CD135, and FGFR expression on SSEA-4high populations at different passages. Values indicate mean ± SEM, sample size is in parentheses.

Transcription factors associated with pluripotent stem cells were also evaluated in early and late cultures. Expression of OCT4, Cripto, and human telomerase reverse transcriptase (hTERT) were quantitatively assessed using TaqMan analysis. RNA samples from H1, H7, and H9 cells at early (p30–40) and late (p60–140) passages showed remarkably similar expression patterns (Fig. 4D). Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of OCT-4, SOX2, Rex-1, BCRP also showed that cells at late passage (p60–80) express these markers (Fig. 4C). None of the stem cell markers measured appeared to change over time in cultures (Fig. 4C,D).

Figure 4.

Maintenance of telomerase activity and molecular marker expression in extended cultures of human embryonic stem (hES) cells. A: Telomeric repeat amplification protocol (TRAP) analysis was performed with 1,000 or 5,000 cells or heat-inactivated (HI) total cellular protein. The telomerase-expressing tumor line H1299 was used as a positive control. B: Southern Blot analysis of terminal restriction fragments (TRF) of hES cells. The mean TRF length was calculated based on the densitometric readings in reference to the standard on the gel. C: Reverse transcriptase-polymerase chain reaction (PCR) analysis of OCT 4, SOX-2, Rex-1, and BCRP in H1, H7, and H9 late-passage cultures. D: TaqMan analysis of human telomerase reverse transcriptase (hTERT), OCT 4, and Cripto. The relative fold difference in the detected signal is expressed on the vertical axis and in the table. Results for each cell line were normalized to the average signal obtained from low-passage cells. Bars indicate mean ± SD for three replicate PCR tests from individual biological samples.

Telomerase Activity and TERT Expression in Early- and Late-Passage Cultures

hES cell cultures show high levels of telomerase activity in early (∼p30–40) and late passages (∼p60–130; Fig. 4A) as determined by telomeric repeat amplification protocol (TRAP) assays. This finding is consistent with the long telomere lengths found in the same passage samples in the TRF assay. The mean TRF length for early-passage H9 cells (p37) was approximately 11 kb. This culture was evaluated at late passage (p77–82) and showed a TRF length of approximately 8–9 kb. In contrast, the H7 cell line showed shortened TRF length at early passages (p36–41), which actually increased to 10–11 kb by passage 84–89 (Fig. 4B). These data indicate that, although the telomere length may be fluctuating over time, the hES cells are not showing signs of senescence with continuous culture.

Telomerase-mediated telomere elongation plays a key role in determining cellular replicative capacity and senescence. The mechanisms regulating the production of an active telomerase enzyme are still unknown, although roles for transcriptional control of hTERT, alternative-splicing of hTERT transcripts, and posttranslational phosphorylation of hTERT protein have been described (Kilian et al., 1997; Cerezo et al., 2002; Ding et al., 2002). More recently, it has been shown that alternative splicing occurs in specific patterns in different tissue types during human development and that splicing likely participates in the regulation of telomerase activity (Ulaner et al., 1998, 2001; Brenner et al., 1999; Yokoyama et al., 2001). We, therefore, examined the expression of TERT and its splice variants in undifferentiated hES cell cultures using primers that spanned splice domains. As can be seen in Figure 5, splicing at both sites occurred in hES cells and multiple isoforms of TERT were present. The pattern of expression was similar at early and late passages in all three lines tested and the relative proportion of the isoforms appeared identical. These data are consistent with the relatively constant level of telomerase expression over long-term culture of undifferentiated hES cells. It will be of interest to conduct similar analyses with differentiated populations of cells.

Figure 5.

Alternative splicing of human telomerase reverse transcriptase (hTERT) in early- and late-passage cell lines. mRNA from H1, H7, and H9 was harvested at the passages indicated and assessed for hTERT splicing patterns by using the primers indicated by arrows. The polymerase chain reaction (PCR) products from different primers are shown in the bottom panels.

Pluripotency of Early and Late Passage Cultures

To assess the pluripotency of late-passage hES cells, we compared the capacity of early- and late-passage hES cells to differentiate in vitro and generate teratomas after injection into immunocompromised mice. All of the cell lines examined generated AFP-, beta-tubulin III-, and muscle-specific actin-positive cells as well as spontaneously contracting cTnI-positive cells after formation of EBs from late-passage cells (Fig. 6A). In addition, all cell lines produced complex teratomas containing representatives of all three germ layers as identified by histologic analysis (Fig. 6B). We did not observe any differences in the frequency of teratoma formation or in the gross appearance of the tumors in between cell lines or over long-term passage.

Figure 6.

In vitro and in vivo multilineage differentiation from high-passage human embryonic stem (hES) cells. A: EB differentiation from late-passage (H9p70) hES cells. β-tubulin III, muscle-specific actin and AFP-positive cells were observed as well as contracting cells (not shown). Antibody staining is in green, DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride) in blue. B: The 5 × 106 cells from high and low passages of hES cells were injected intramuscularly into SCID/beige mice. Low-power views of teratomas generated by H1 hES cells at passage 29 and 70 are shown. The arrows indicate specific differentiated structures. Each mouse received a single injection of cells, and the table shows the efficiency of teratoma generation in SCID/beige mice with low- and high-passage cells (no. of mice with teratomas/no. of mice injected).

Karyotypic Stability

Cytogenetic analysis of all three cell lines was carried out at multiple times over continuous culture. In this assessment, separate cultures maintained by several individuals were evaluated. Analysis included cultures maintained continuously by single individuals as well as cultures exchanged between individuals. Cultures derived from a single-parent culture but maintained by different individuals were considered to be separate cultures. Cytogenetic analysis included g-banding of at least 20 cells in each culture. Cultures were considered to be aneuploid if one or more cells assessed showed mutations. Overall, 20% of the cultures contained some aneuploid cells. The percentage of aneuploid cells in the cultures ranged from 2 to 100% of the cells within the cultures. As shown in Figure 7, the frequency of aneuploidy did not appear to increase with continuous subculturing of the cells. By using the Students' t-test, we did not find statistically significant differences in the rate of aneuploidy between early- and late-passage cultures or between the different cell lines. In some cases, cultures which contained aneuploid cells were maintained to determine whether the percentage of cells increased or decreased over time; both trends were identified. Of interest is the finding that the most frequent mutation identified was trisomy 20. This finding did not appear to correlate with changes in proliferation rate, but this phenomenon is still under investigation.

Figure 7.

Karyotype stability of human embryonic stem (hES) cells in extended cultures. Metaphase spread of the H9 hES cell line at passage 91 Representation of aneuploidy occurrences as a function of extended culture for H1 (B), H7 (C), and H9 (D) hES cell lines.

Microarray Analysis of Gene Expression

Microarray analysis was used to look for significant differences in gene expression that would distinguish between late- and early-passage hES cultures. The microarray used in this analysis was designed to randomly survey named human genes and contained 4,224 cDNA clones. Five sets of RNA samples were used in this analysis. Three sets were from early passage (p28–p37), and two sets were from late passage (>p48). Each set contained samples from four consecutive subcultures, except in the case of H1 for which samples were collected every other subculture due to cell availability. Only one sample was collected for H9 at late passage.

Two microarrays were analyzed for each RNA sample by using dye reversal. Experimental samples were labeled with Cy3 or Cy5 and a common reference consisting of a pool of experimental samples was labeled with the other dye. Log transformed, normalized expression ratios were averaged for the two arrays. A strict filtering scheme, by which expression of a gene had to be detected in all 17 RNA samples for data to be considered, was applied to the data to eliminate genes with low expression levels or marginal data (see Experimental Procedures section). A total of 2,798 cDNA clones, or 66%, passed the filtering criteria, and this set was used in subsequent analysis.

Remarkably few genes showed differential expression. Only 109 genes (<3%) showed differential expression of threefold or greater when the highest expressing sample was compared with the lowest expressing sample. A total of 38 genes (1.4%) showed 5-fold or greater (Table 2), and 11 genes (0.4%) showed 10-fold or greater differential expression when using the same criteria.

Table 2. Fold Expression Relative to Reference Sample for Genes Showing at Least Five-fold Differential Expression in Microarray Analysis

Hierarchical clustering was performed on the set of 109 genes showing threefold or greater differential expression (Fig. 8A). The most distinct difference among hES samples was between early and late passage. In a dendrogram of hES samples, the highest branch point separates the 5 late-passage samples from the 12 early-passage samples (Fig. 8B). This difference was driven by two classes of genes; those that showed increased expression levels in late passage cultures (Fig. 8C) and those that showed decreased expression levels in late-passage cultures (Fig. 8D). Genes that show up-regulation in late passage include nodal, osteopontin, melanoma antigen A4, lefty, and an expressed sequence tag (EST). Genes that are down-regulated in late passage include four collagen genes, fibronectin 1, STAT4, a lectin, and two genes involved in transforming growth factor-β signaling.

Figure 8.

Cluster analysis of microarray data. A: A set of 109 genes that showed a maximum differential expression of threefold or greater between the highest and lowest expressing sample in 17 samples of human embryonic stem (hES) cells were used in two-dimensional hierarchical cluster analysis. B: A dendrogram showing the relationship among the 17 hES cultures analyzed is shown. C,D: Specific clusters of genes showing progression from low to high expression (C) and high to low expression (D) in later-passage cultures are shown. The color scheme progresses from green to black to red representing low expression, no differential expression, and high expression, respectively.


We have examined three cell lines, i.e., H1, H7, and H9, and find that pluripotency, the expression of surface markers, and transcription factors remain consistent for up to two years in continuous culture. However, we observed subtle changes in morphology of the cells as well as changes in factors that were identified using microarray technology.

The expression of surface markers characteristic of undifferentiated hES cells was quite similar at early and late passages in all three lines examined. The percentage of cells that expressed SSEA-4, TRA-1-60, TRA-1-81, and CD9 appeared to increase slightly over time. This trend correlated with the changes in morphology of the cells in which the differentiated cells (stroma) are generally lost in long-term culture. CD133 expression was more variable, when measured individually or as coexpressed with SSEA-4. This finding may indicate that, although CD133 appears to be a useful marker for the identification of neural stem cells or hematopoietic stem cells, the relevance of its expression by the undifferentiated hES cells will need to be investigated further. Quantitative assessment of the transcription factors OCT-4, hTERT, and Cripto showed remarkably stable expression over time. Furthermore, transcription factors such as Rex-1 and SOX2 were expressed by early- and late-passage cells, indicating that quantitation of these markers may be useful tools in further assessment of multiple hES derivations.

Previous work from our lab has shown that late-passage hES cells (H9, H9.1, and H9.2) maintained telomere length (Amit et al., 2000). The work described here demonstrates the maintenance of telomere length in the H7 line as well. The abundance of telomerase expression is consistent with the maintenance of telomere length. These data coupled with the extensive proliferative capacity of these cells indicates that these populations are functionally immortal.

The karyotype analysis demonstrated that approximately 20% of individual cultures showed some degree of aneuploidy. In most cases, however, less than 50% of the cells within an “abnormal” culture showed an abnormal karyotype. Recent reports have demonstrated that neuroblasts isolated from embryonic and adult mouse cortex show approximately 33% aneuploidy (Rehen et al., 2001). It is unclear whether the presence of aneuploid cells in the hES cell cultures will alter the fundamental characteristics of the cells. Surprisingly, mouse ES cells exhibiting 38% or more normal karyotype can generate germ-line chimeras (Suzuki et al., 1997). It is unknown whether the types of aneuploidy seen in the hES cultures will effect growth rate, cell cycle regulation or the capacity to differentiate. However, the occasional identification of atypical cells, albeit at a low frequency, nevertheless highlights the importance of careful monitoring.

We used microarrays with approximately 4,500 genes to assess the overall stability of the hES cell lines. The arrays contain commercially available, sequence verified cDNA clones. In addition, housekeeping and spike-in controls were included to verify the quality of the data. In some cases, duplicate clones from a single gene were included on the microarray. Data from these duplicate spots were quite consistent across the 17 RNA samples examined in this study, as shown in Table 2 for duplicate spots of COL1A2.

Less than 1.4% of all genes expressed showed major changes (fivefold or greater) in expression levels when the highest expressing sample was compared with the lowest expressing sample. It should be pointed out that these effects were seen in the H7 cell line, and it is possible, therefore, that these changes are cell line-specific. However, the vast majority of genes show no differential expression among three cell lines and over 27 passages. The similarity of gene expression in the subset of early-passage samples presented here is further discussed in Carpenter et al. (2004). Most of the differentially expressed genes distinguish between early- and late-passage cultures. Of the 38 genes showing fivefold or greater differential expression, all but two, RAR2b and adenylate cyclase 8, show the greatest difference between late and early passage. RAR2b and adenylate cyclase 8 each have a single early-passage sample with elevated expression levels; therefore, the differential expression most likely reflects either an experimental artifact or a random fluctuation in gene expression rather than a consistent difference among cell lines or between late- and early-passage cultures. The 36 genes that show the largest difference between late- and early-passage cultures are consistent with the loss of stroma-like cells in later-passage cultures.

Of 15 collagen genes detected on the array, six were differentially expressed. The largest difference observed was that all six are expressed at lower levels in late-passage H7 cultures. With the exception of COL6A1, all show the same downward trend in H9p49. This correlates with the loss of stroma-like cells, because the five late-passage samples came from cultures lacking stroma, while the 12 low-passage samples came from cells with significant stroma. Thus, differences that we observe appear to reflect the adaptation of cells to culture and do not appear to change critical features of the ES cells in terms of self-renewal characteristics, telomere length, telomerase activity, or the ability to contribute to all three germ layers. Nevertheless, these results highlight the importance of careful monitoring and a comprehensive analysis of multiple parameters.

These data suggest that the panel of markers currently used to assess the hES cell lines from the same laboratory over time in culture are not sensitive enough to elucidate differences between the cell lines. The three lines analyzed were all derived from the same laboratory and were maintained by the same group in feeder free conditions over the course of the experiments. It should be noted that differences in culture conditions such as maintaining the cells on feeders may alter the signature profile of any cell line. Therefore, we believe that we cannot, at present, extrapolate from the current analysis to other hES cell lines maintained under different conditions. A similar detailed analysis must be considered to profile all existing lines. In addition, hES cell cultures represent a heterogeneous population of cells, containing undifferentiated cells as well as some spontaneously differentiated cells. Furthermore, these cells are in different states of cell cycle and, therefore, may be expressing different markers, at different levels. Our analysis, however, highlights the parameters that should be tested and predicts that comparison across a common set of comprehensive markers may allow us to predict the behavior of additional lines in the future.

Overall, our results show that, like mES cells, hES cells appear quite stable, do not appear to senesce, continue to maintain their telomere integrity and exhibit high levels of telomerase activity, and maintain markers characteristic of undifferentiated cells up to two years in culture. We have observed more subtle changes in long-term cultures such as morphology and changes in extracellular matrix, which may represent significant changes or may merely reflect an adaptation to these culture conditions. It will be important, therefore, to generate characterized cell banks for the use of hES cells in cell therapy. These results support the therapeutic potential of hES cells but also offer a note of caution in demonstrating that even stable cells can change over time in culture, and these changes must be monitored.


Human embryonic stem cell lines H1, H7, and H9 (Thomson et al., 1998) were maintained and passaged under feeder-free conditions (MEF-CM + 8 ng/mL bFGF) as described (Xu et al., 2001). Cell lines were maintained on feeders until passage 13–16 before switching to feeder-free conditions. Passage numbers indicated in the manuscript include this initial culture period. Cultures were passaged once each week, and all passage numbers reflect number of passages in continuous culture.

Flow Cytometry

Flow cytometric analysis was performed as described previously (Carpenter et al., 2004). Briefly, confluent cultures of hES cells were harvested for analysis by using 0.5 mM ethylenediaminetetraacetic acid (Sigma, St. Louis, MO) in phosphate buffered saline (PBS). After blocking with heat-inactivated rabbit serum, cells were incubated with appropriate primary antibodies, phycoerythrin (PE) -conjugated antibodies, or appropriate isotype matched controls (Southern Biotechnology Associates, Birmingham, AL, or Sigma or Becton Dickinson, San Jose, CA) for 30 min at 4°C. Primary antibodies used were MC480 (SSEA-1), 1:5; MC813 (SSEA-4), 1:5 (all from Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA); TRA-1-60, 1:12; TRA-1-81, 1:20 (a gift from Dr. Peter Andrews, University of Sheffield, UK); and anti-FGFR, 1:25 (Biogenesis, Brentwood, NH). PE-conjugated antibodies used were CD9-PE, 1:5 (BD PharMingen, San Diego, CA); CD133-PE, 1:10 (Miltenyi Biotec, Auburn, CA); CD135, 1:10 (Caltag Laboratories, Burlingame, CA); CD117, 1:10 (BD PharMingen); and CD130, 1:10 (R&D Systems, Minneapolis, MN). Cells were washed and incubated for 30 min at 4°C with fluorescein isothiocyanate-conjugated goat F(ab′)2 anti-mouse IgG3, 1:100 and PE-conjugated goat F(ab′)2 anti-mouse IgM, 1:100 (Southern Biotechnology Associates). Cells were washed as before and resuspended for analysis in staining buffer containing 1 μg/ml of propidium iodide (PI; Sigma) to identify nonviable cells. Flow cytometric analysis was performed by using a FACSCalibur Flow Cytometer (Becton Dickinson). At least 10,000 PI-negative events were collected for analysis. Acquired data were analyzed using CELLQuest software (Becton Dickinson). Statistical analysis was performed by using a two-tailed Student's t-test with Prism software.

Telomerase Activity

Telomerase activity was measured using the TRAP assay as described (Kim et al., 1994; Weinrich et al., 1997). Terminal restriction fragment (TRF) size was determined using Southern hybridization essentially as described (Harley et al., 1990; Allsopp et al., 1992; Vaziri et al., 1993).

Teratoma Formation

Teratomas were generated by intramuscular injection of 5 million undifferentiated hES cells into SCID/beige mice. Tissues were harvested after 41–77 days and histologic analysis was performed by IDEXX (West Sacramento, CA).

Cytogenetic Analysis

Cytogenetic analysis was performed by the Medical Genetics Cytogenetics Laboratory at Children's Hospital in Oakland, CA. Because samples were not paired, the Student's t-test was used to assess statistical differences between time points. The percentage of euploid cells in each bin (groups of 10 passages) were compared.

hTERT Splicing

RNA was isolated with the chaotropic reagent Trizol (Life Technologies, Inc., Gaithersburg, MD) according to the manufacturer. One microgram of total RNA was reverse-transcribed by using the RNA PCR kit (Perkin Elmer) and subjected to PCR as recommended by the manufacturer. Primer sequences were chosen to amplify a region of the hTERT mRNA containing the motifs for the reverse transcriptase activity. RT-PCR experiments were conducted by using primers (5′ gcctgagctgtactttgtcaa 3′ and 5′ cgcaaacagcttgttctccatgtc) partly encompassing the reverse transcriptase domain of the hTERT mRNA (see Results section). Within this region, two splice sites have been identified. Splicing at the site causes a 36-base deletion (bases 2186 to 2221) in the mRNA, and β splicing results in the loss of 182 bases (bases 2342 to 2524) from the transcript. Nucleotide loss at the splice site causes partial ablation of the conserved reverse transcriptase motif A, whereas splicing at the β site results in loss of the motifs B, C, D, and E. As a positive control, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA was amplified in parallel with the primers GAPDH-S, 5′ acgaccactttgtcaagctcat 3′, and GAPDH-A, 5′ ggtactttattgatggtacatg 3′, for 30 cycles with an annealing temperature of 60°C.


Real-time PCR, TaqMan RT-PCR, was performed on the ABI 7700 under the following conditions: 1× RT-Master Mix (ABI), 300 nM for each primer, and 80nM of probe, and 10 pg to 100 ng of total RNA in nuclease-free water. The reaction was conducted under default RT-PCR conditions of 48°C hold for 30 min, 95°C hold for 10 min, and 40 cycles of 95°C at 15 sec, and 60°C hold for 1 min. RNA was isolated by a guanidinium isothiocyanate method according to manufacturer's (RNeasy kit, Qiagen) instructions and subsequently DNAse treated with the DNAfree kit (Ambion). Gene-specific primers and probes were designed by PrimerExpress software (version 1.5, ABI). Probe oligonucleotides were synthesized with the fluorescent indicators 6-carboxyfluorescein (FAM) and 6-carboxy-tetramethylrhodamine (TAMRA) at the 5′ and 3′ ends, respectively. The primers and probe for Oct 4 (GenBank accession no. NM_0020701) were as follows: forward primer (GAAACCCACACTGCAGCAGA), reverse primer (CACATCCTTCTCGAGCCCA), and probe (FAM-CAGCCACATCGCCCAGCAGC-TAM). The primers and probes Cripto (GenBank accession no. NM_003212) were as follows: forward (TGAGCACGATGTGCGC), reverse (TTCTTGGGCAGCCAGGTG), and probe (6 FAM-AGAGAACTGTGGGTCTGTGCCCCATG-TAM). The primers and probe for hTERT purchased from Applied Biosystems (PDAR, predeveloped assay reagent). Relative quantitation of gene expression between multiple samples was achieved by normalization against endogenous 18S ribosomal RNA (Applied Biosystems) by using the ΔΔCT method of quantitation. Fold changes were calculated as 2-ΔΔCT.


Staining was carried out as described (Xu et al., 2001). Briefly, cells were fixed for 20 min at room temperature in 4% paraformaldehyde and permeabilized for 2 min in 100% ethanol. Monoclonal antibodies against AFP (C3) and β-tubulin (SDL-3D10) were purchased from Sigma Aldrich (St. Louis, MO), and the monoclonal antibody against muscle-specific actin was purchased from LabVision (Fremont, CA).

Microarray Analysis

Libraries of human cDNAs were obtained from Research Genetics (Huntsville, AL). Purified plasmid DNA served as a template for PCR in 100-μl reactions using vector primers. After checking PCR products on agarose gels and reformatting to remove failed reactions, they were purified by using the PCR Purification Kit (Telechem, Sunnyvale, CA) as described by the manufacturer. Dried PCR products were resuspended in 50% dimethyl sulfoxide and arrayed onto GAPS amino silane-coated slides (Corning, Inc., Life Sciences, Acton, MA) by using a GMS 417 Arrayer (Affymetrix, Santa Clara, CA). To ensure uniform spot intensity, arrays were moistened over a boiling water bath then snap-dried and ultraviolet cross-linked (150–300 mJ). To reduce background, slides were processed as described by the slide manufacturer (Corning). Briefly, blocking solution was prepared by dissolving 5.5 g of succinic anhydride in 335 ml of pyrrolidine then adding15 ml 1 M sodium borate. Arrays were immediately plunged into the blocking solution and incubated at room temperature for 15 min. Blocked arrays were then placed in a boiling water bath for 90 sec, briefly rinsed in isopropanol, and dried by centrifugation.

RNA samples were prepared from confluent cultures of hES cells grown for 6 days after subculture using feeder-free conditions as described (Xu et al., 2001). All samples were collected over a two-month period. Media was removed and cells were rinsed with warm PBS. Cells were then harvested in situ by overlaying with RLT lysis buffer (Qiagen, Valencia, CA). RNA was prepared using RNeasy Midi kits as described by the manufacturer (Qiagen). To remove contaminants, RNA was precipitated by adding an equal volume of LiCl Precipitation Solution (Ambion, Austin, TX), washed with 70% ethanol and resuspended in water. The reference sample used throughout this data set was a pool of undifferentiated hES RNA.

Probes were prepared from 10 μg of total RNA by direct incorporation of Cy3 or Cy5 dCTP (Amersham Biosciences, Piscataway, NJ) into oligo dT primed first-strand cDNA using SuperscriptII reverse transcriptase, as described by the dye manufacturer. RNA was removed by alkaline hydrolysis, then neutralized probes were purified by using Qiaquick columns as described by the manufacturer (Qiagen).

Purified probes were dried and resuspended in 20 μl of ArrayHyb Low Temp Hybridization Buffer (Sigma-Aldrich, St. Louis, MO) with the addition of blocking solution containing 15 μg of CotI DNA, 15 μg of polydA, and 20 μg of tRNA. Slides were prehybridized in 5× standard saline citrate (SSC), 0.1% sodium dodecyl sulfate (SDS), 1% bovine serum albumin, 50% formamide containing blocking solution for 30 min at 42°C. Slides were washed in water, rinsed in isopropanol, dried by centrifugation, and hybridized overnight at 42°C under coverslips sealed with contact cement. After the coverslips were removed, the slides were rinsed in 2× SSC, washed two times in 0.1× SSC, 0.1% SDS for 5 min each, two times in 0.1× SSC for 5 min each, and dried by centrifugation.

Arrays were scanned by using a GenePix 4000A microarray scanner (Axon Instruments, Fremont, CA) with 100% scan power and PMT voltage set for each array to minimize saturated pixels and approximately equalize signal intensities in the two channels. PMT voltages ranged from 710 to 850 V for scans at 635 nm and 610 to 680 V for scans at 532 nm. Image processing and data extraction were performed by using GenePix Pro 3.0.6 (Axon Instruments).

Quality check calculations were preformed to eliminate signals outside of the linear detection range for each of the 68 individual scans. The intensity value for each spot was assigned a flag value of 1 if either less than 80% of measured pixels were above two times the standard deviation of the measured local background or more than 5% of the pixels were saturated; otherwise the flag value was 0.

Expression ratios were calculated for a spot on an array if a signal was detected in either the Cy3 or Cy5 channel. The ratio was calculated by subtracting the median local background from the mean pixel intensity of each feature then dividing the sample value by the reference value. The expression ratio was then balanced by multiplying the calculated expression ratio by the ratio of means for all data points on the same array. Log transformed, balanced ratios of background-subtracted intensities were used for subsequent calculations.

For each experimental RNA sample, duplicate arrays were run by using dye reversal. Average log expression ratios were calculated from dye-reversed replicates when at least two of four measurements were not flagged; otherwise, the expression ratio was left blank. Data from the 17 experimental samples were then pooled and filtered to remove cDNAs in which any of the 17 expression ratios was blank or the sum of the 68 possible flags was greater than 9. Hierarchical clustering of log-transformed normalized expression ratios using Ward's method to calculate distances was carried out using JMP 5.0 (SAS Institute, Cary, NC).


We thank Michael Mok for technical assistance. We thank Drs. Lebkowski and Harley for their comments on this manuscript. We gratefully acknowledge the input of all members of our laboratory provided through discussions and constructive criticisms. Mahendra S. Rao was a consultant for Geron on this project and work constitutes approved outside activity. SSEA-1 and -4 antibodies were developed by D. Solter and B.B. Knowles and were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA, 52242.