Phenotype plasticity rather than repopulation from CD90/CK14+ cancer stem cells leads to cisplatin resistance of urothelial carcinoma cell lines

UCCs are heterogeneous for cytokeratin expression and proportions of differentiation
states

First, we characterized a representative panel of 11 UCCs for the expression of markers
associated with cellular morphology (mesenchymal-like vs. epithelial-like), cytokeratins
(CKs) CK14, CK5, and CK20 and cluster-of-differentiation (CD) surface markers CD90,
CD44, and CD49f. According to the recently published hierarchical ‘differentiation
state model’ for UC the correlated expression of both CK and CD markers may be used
to assign cellular subpopulations to three defined differentiation states (Fig. 1a) 10]. Differences in morphology among UCCs were also reflected distinctively on the molecular
level. As expected, epithelial markers (E-cadherin, miR200) were almost exclusively expressed in cell lines with epithelial morphology (VM-CUB-1,
BFTC-905, HT-1376, RT-112, 5637). In contrast, mesenchymal markers (Vimentin, ZEB1) were almost undetectable in UCCs with an epithelial phenotype, but strongly expressed
in UCCs with a mesenchymal phenotype (253 J, 639 V, T24, UM-UC-3, SW-1710, J82; Fig. 1b). Surprisingly, expression analysis of cytokeratins revealed that mesenchymal phenotype
UCCs lacked RNA-expression of CK14, CK5, and CK20 (Fig. 1c). In addition, cytokeratin expression differed also among epithelial UCCs. Whereas
the CK14 expression level in VM-CUB-1 cells resembled that of normal cultured uroepithelial
cells from ureters (data not shown), expression was 14-fold higher in HT-1376 cells
and 5-fold lower in RT-112. CK5 was robustly expressed in all epithelial-like UCCs,
with the exception of HT-1376, and CK20 was detectable in BFTC-905, HT-1376, and RT-112.

As expected, we observed a heterogeneous distribution of the cell surface markers
CD90, CD44, and CD49f by flow cytometry among UCCs. In contrast to cytokeratins, surface
markers were generally expressed independently of the morphological phenotype (Fig. 1d, Additional file 2: Figure S2). In many UCCs only a small number of cells were positive for CD90, whereas
other lines contained significant numbers of CD90-positive cells, so that their abundance
varied between 0.75 and 99.5 % across the panel. For example, the 639 V cell line
appeared to contain exclusively triple-positive (i.e. CD90
⁺
CD44
⁺
CD49f
⁺
) cells. Intriguingly, this cell line did not express CK14 at detectable levels. With
the exception of HT-1376, all cell lines contained a high fraction of CD44
⁺
cells. Similarly, almost all UCCs were CD49f
⁺
, only the cell lines SW-1710 and J82 contained few CD49f
⁺
cells (Fig. 1d, Additional file 2: Figure S2). Taken together, this characterization demonstrated that UCCs collectively
reflect the heterogeneity observed among primary UCs, but also indicated that the
proposed correlation between CK and CD markers does not apply to UCCs, especially
in mesenchymal cell lines lacking cytokeratins.

Enrichment of CD90
⁺
cells does not enrich for CK14 expression in UCCs

We further investigated the postulated correlation between CD90 and CK14 by enrichment
of CD90-positive cells and subsequent expression analysis of cytokeratins. To this
end, we selected three cell lines representative for different epithelial-like or
mesenchymal-like phenotypes, with variable expression of CD90 and CK14 (RT-112: epithelial,
CD90
^low
, CK14
^intermediate
; J82: mesenchymal, CD90
^low
, CK14
^low
; HT-1376: epithelial, CD90
^low
, CK14
^high
). The efficiency of immunomagnetic enrichment of CD90
⁺
cells was monitored by triple staining (CD90, CD44, CD49f) via flow cytometry. After
enrichment the abundance of CD90-positive cells was significantly increased up to
50 % (Fig. 3a, Additional file 3: Figure S1a). Subsequently, mRNA expression of CK14, CK5, and CK7, which is known to be robustly expressed in all uroepithelial cells, was quantified
by qRT-PCR (Fig. 2a). As expected, the pan-urothelial cytokeratin CK7 did not differ significantly between the populations. However, CK14 expression, too, was not significantly differentially expressed between CD90
⁺
and CD90
^?
cell fractions. CK5, a marker of basal and intermediate differentiation states in UC (Fig. 2a), was invariable in RT-112, but decreased in the CD90
^?
fraction of HT-1376. Concurring with the analysis above (Fig. 1c), CK5 expression was below the limit of detection in J82 cells (Fig. 2a). Accordingly, immunofluorescence staining for CD90 and CK14 demonstrated that both
markers were heterogeneously expressed in the cell lines, but not generally co-expressed,
as illustrated for the HT-1376 cell line in Fig. 2b.

Fig. 2. Enrichment of CD90
⁺
cells does not enrich for CK14 expression in UCCs. a Relative RNA expression of CK7, CK14, and CK5 in RT-112, J82, and HT-1376 cells magnetically enriched for CD90. Expression levels
of the CD90
⁺
cell fraction were set as 1. nd not detectable. b Representative immunofluorescence stainings of CD90 (red) and CK14 (green) for cell lines RT-112, J82, and HT-1376; DAPI was used for nuclear staining. Scale
bars, 50 ?m. Values are expressed as the mean?±?SD of triplicates

CD90
⁺
UCCs do not exhibit a distinct stem cell-like phenotype

Next, we investigated whether CD90
⁺
cells isolated from UCCs exhibit CSC properties by measuring clonogenicity, self-renewal
and differentiation capacity or increased resistance to cisplatin. Following magnetically
CD90 enrichment or depletion, CD90
⁺
and CD90
^?
subpopulations of RT-112, J82 and, HT-1376 were recultured and followed over time.
After reculturing, the number of CD90
⁺
cells in enriched cultures regressed to baseline expression. Interestingly, in the
CD90 depleted subpopulation, the amount of CD90
⁺
cells likewise increased to the baseline level (Fig. 3a, Additional file 3: Figure S1b). Similarly, we did not observe any difference in colony forming potential
between CD90 enriched and CD90 depleted cell fractions subsequent to immunomagnetic
sorting (Fig. 3b; MACS panel). Since immunomagnetic sorting only allows enrichment but not complete
separation of cell fractions, RT-112 and HT-1376 cell lines were additionally sorted
by flow cytometry. Colony formation assays of highly purified cell fractions revealed
a slight advantage for CD90
⁺
cells (Fig. 3b; FACS panel). We also analysed self-renewal capacity subsequent to FACS sorting by
seeding single cells in 96-well-plates. Although spheres from CD90
⁺
cells appeared to grow slightly faster (Fig. 3b; right panel) we did not observe a significant difference in the number of colonies
originating from single CD90
⁺
or CD90
^?
RT-112 cells. Thus, RT-112 CD90
^?
cells as well exhibited self-renewal capacity. HT-1376 cells did not form colonies
after single cell isolation (data not shown).

Fig. 3. CD90
⁺
UCCs do not exhibit a distinct stem cell-like phenotype. a CD90
⁺
fraction in unsorted (grey bars), CD90 enriched (dark grey bars), CD90 enriched cells after reculturing for about 7–8 population doublings (dark grey bars, shaded), CD90 depleted (white bars), and CD90 depleted cells after reculturing (white bars, shaded) in RT-112, J82, and HT-1376 as measured by flow cytometry. b Clonogenic potential in magnetically and FACS sorted CD90
⁺
and CD90
^?
populations from RT-112, J82, and HT-1376 cell lines shown by Giemsa staining. Colony-forming
potential of single cells positive or negative for CD90 from RT-112 cells after FACS
sorting. Phase-contrast microscopy, scale bars, 100 ?m. c Cisplatin sensitivity was measured in unsorted and MACS sorted CD90
⁺
and CD90
^?
fractions of RT-112, J82, and HT-1376 by MTT assay after 72 h treatment. Untreated
cells were set as 100. d Relative cisplatin sensitivity of FACS sorted CD90
⁺
and CD90
^?
cells from RT-112 and HT-1376 cell lines as measured by CellTiter-Glo Luminescent
Cell Viability Assay after 72 h cisplatin treatment with IC
₅₀
concentrations (see Fig. 4a). Untreated cells were set as 1. Values represent the mean?±?SD of quadruplicates.
*P 0.05; **P 0.001

Additionally, CD90
⁺
magnetically enriched and FACS sorted cell fractions were examined for their cisplatin
sensitivity. CD90
⁺
enriched cell fractions from RT-112, J82, and HT-1376 cell lines were no more resistant
to cisplatin than the corresponding CD90 depleted fractions (Fig. 3c). However, highly purified FACS sorted CD90
⁺
cells were less sensitive to cisplatin treatment than CD90
^?
cells from RT-112 and J82 cell lines (Fig. 3d).

UCCs sensitivity towards short-term treatment with cisplatin is not correlated with
abundance of CD90
⁺
cells

To investigate the relation between the abundance of CD90
⁺
cells and cisplatin sensitivity, we sought to identify the appropriate doses and time
schedule for cisplatin treatment. Thus, we determined IC
₅₀
concentrations for cisplatin after 48 and 72 h and also checked for changes in the
abundance of CD90
⁺
cells from 24 to 96 h by flow cytometry (Fig. 4a). Based on the results, the following experiments were performed within a period
of 72 h, IC
₅₀
values ranged between 1.07 and 12.5 ?M (Fig. 4a). Across the cell lines, no correlation was obvious between the abundance of CD90
⁺
cells (Fig. 1d) and sensitivity to cisplatin (Fig. 4a). For instance, the 639 V cell line comprising the biggest fraction of CD90
⁺
cells was highly sensitive to cisplatin, whereas RT-112 containing a small fraction
of CD90
⁺
cells was the most resistant cell line. However, following short-term treatment (STT)
with cisplatin at IC
₅₀
doses for 72 h, the abundance of CD90
⁺
cells increased significantly in 6/11 UCCs, particularly in cell lines with originally
low abundance of CD90
⁺
cells (Fig. 4b, Additional file 2: Figure S2). The fraction of CD44
⁺
and CD49f
⁺
cells was augmented only in UCCs with low abundance of these markers, namely HT-1376
(CD44
^low
) and J82 (CD49f
^low
).

Fig. 4. UCCs sensitivity towards short-term treatment with cisplatin is not correlated with
abundance of CD90
⁺
cells. a Cell viability was measured 72 h after cisplatin treatment by MTT assay in 11 UCCs,
categorized into UCCs with epithelial (dark grey bars) and mesenchymal (light grey bars) phenotypes. b Subsequent to short-term treatment with cisplatin (STT, 72 h) most cell lines displayed
increased numbers of CD90+ cells. Mean percentages of CD90
⁺
cells in untreated (dark grey bars) and STT (dark grey bars, shades), CD44
⁺
cells in untreated (light grey bars) and STT (light grey bars, shades), and CD49f
⁺
cells in untreated (grey bars) and STT (grey bars, shades) UCCs as measured by flow cytometry. c) Relative expression of CK14, CK5, and CK20 was measured by qRT-PCR in a panel of 11 untreated and STT-UCCs. Expression in the
respective untreated control cells was set as 1. Data represent the mean?±?SD of three
independent experiments. SDHA was used as reference gene and relative expression calculated by using the 2
^???CT
method. Untr Ctrl untreated control, STT short-term cisplatin treatment. *P 0.05; **P 0.001

We additionally analysed the cell lines with an increased fraction of CD90
⁺
cells after cisplatin treatment for cytokeratin RNA expression. Again, cytokeratins
generally expressed in uroepithelial cells, like CK7 and CK13, remained mostly unchanged
and no increase in CK14 expression was observed (Fig. 4c).

Long-term cisplatin treated UCCs are not enriched for CD90
⁺
/CK14
⁺
cells

Since some enrichment of CD90
⁺
cells was observed after short-term treatment with cisplatin, we wondered whether
long-term cisplatin treatment further selects for this cell fraction. RT-112, J82,
and HT-1376 cells underwent long-term cisplatin treatment (LTT) over several months
with escalating doses, following a protocol similar to the Resistant Cancer Cell Line
(RCCL) collection 24]. Ultimately, resistant RT-112 and J82 cells could be maintained as proliferating
cultures with 50 and 1.6 ?M cisplatin, respectively, added after each passage. No
long-term surviving cells could be obtained from HT-1376. IC
_50,
determined after 72 h treatment of LTT cell lines, increased significantly compared
to the corresponding parental cells. In RT-112-LTT IC
₅₀
for cisplatin was 50 ?M compared to 12.5?±?2.7 ?M in the parental cells and IC
₅₀
for J82-LTT increased from 1.5?±?0.48 ?M to 9.2?±?4.2 ?M. The RT-112-LTT cells grew
much slower than the parental line, with population doubling times of 32.7 and 25.3 h,
respectively. Similar results were found in J82-LTT and untreated parental J82 with
doubling times of 33.1 and 24.5 h, respectively (Fig. 5a). Moreover, the LTT-UCCs were much less clonogenic than the parental cells in the
absence of cisplatin (Fig. 5b). However, upon cisplatin treatment, only LTT-UCCs were capable of clonogenic growth
compared to the parental controls indicating their growth advantage through acquired
cisplatin resistance (Fig. 5b). Subsequently, CD90-, CD44- and, CD49f-positive cell fractions as well as mRNA expression
of CK7, CK13, and CK14 were determined in theRT-112-LTT and J82-LTT. In contrast to
the STT, CD90
⁺
cells were not enriched in LTT sublines (Fig. 5c, Additional file 4: Figure S3). The fraction of CD44 and CD49 positive cells also remained largely unchanged
as compared to the parental cell lines. To corroborate our results on resistant RT-112
and J82 cells, further resistant UCCs were generated (Fig. 5c, Additional file 4: Figure S3; VM-CUB-1-LTT, 5637-LTT, 253 J-LTT, T24-LTT, SW-1710-LTT). In these cell
lines, too, CD90
⁺
cells were not enriched compared to their parental UCC. On the contrary, a significant
decrease of CD90
⁺
cells was observed in 5637-LTT and T24-LTT. Expression of CK14 was not increased in
any LTT-cell line, rather expression of cytokeratins decreased generally (Fig. 5d). Fitting the decreased expression of cytokeratins, RT-112-LTT and J82-LTT cells
underwent morphological changes. While parental RT-112 cells form epithelial colonies
with strong cell-to-cell-adhesion, RT-112-LTT colonies were less compact and cells
often assumed a more mesenchymal phenotype (Fig. 6a). Compared to the untreated controls, RT-112-LTT and J82-LTT increased in size. In
summary, morphological changes pointed towards an epithelial to mesenchymal transition
(EMT) in UCCs upon LTT.

Fig. 5. Long-term cisplatin treated UCCs are not enriched for CD90
⁺
/CK14
⁺
cells. a Relative cell number in RT-112-LTT and J82-LTT and parental cell lines was measured
by MTT after 24, 48, 72 and 96 h. Population doubling time was calculated based on
raw absorbance data. b Clonogenic potential of RT-112-LTT and J82-LTT and parental cell lines without or
with cisplatin, Giemsa staining. c Mean percentages of CD90
⁺
cells in untreated (dark grey bars) and LTT (dark grey bars, shades), CD44
⁺
cells in untreated (light grey bars) and LTT (light grey bars, shaded), and CD49f
⁺
cells in untreated (grey bars) and LTT (grey bars, shaded) UCCs as measured by flow cytometry. d Relative expression of CK7, CK13, and CK14 in untreated and LTT-UCCs. Expression levels in the untreated control were set as
1. SDHA was used as a reference gene and relative expression was calculated by the 2
^???CT
method. Values represent the mean?±?SD of biological triplicates. Untr Ctrl untreated control, LTT long-term cisplatin treatment, PDT population doubling time, na not available. *P 0.05; **P 0.001

Fig. 6. Activation of WNT-signalling may contribute to survival of UCCs upon long-term cisplatin
treatment. a Morphology of RT-112-LTT and J82-LTT and their parental cell lines. Scale bars, 100 ?m.
qRT-PCR demonstrated relative expression levels of E-Cadherin, Vimentin, Twist1, ZEB1 (b) and CLDN3 and CLDN4 (c) in RT-112-LTT and J82-LTT and their parental cell lines. d Immunofluorescence stainings for E-cadherin and Vimentin (d) and ?-Catenin (e) in RT-112-LTT and J82-LTT and their parental cell lines. DAPI staining (blue) was used to visualize nuclei. Scale bars, 50 ?m. f Relative RNA expression levels of ?-Catenin, AXIN-2, CCDN1, c-MYC, and PITX2 in untreated and RT-112-LTT and J82-LTT. Expression levels in the untreated control
were set as 1. g Basal and inducible activity of a TCF/?-Catenin-dependent promotor. Mean?±?SD of
duplicates of TopFlash/FopFlash and TopFlash+S33Y/TopFlash ratio are shown in RT-112-LTT
and J82-LTT and their parental cell lines. 22RV1 and HepG2 cell lines were used as
controls. h Relative cell viability was measured 72 h after treatment with cisplatin, niclosamide,
or combination of both by MTT assay in RT-112-LTT and J82-LTT. *P 0.05; **P 0.001

Phenotypic plasticity of UCCs facilitates evasion from long-term treatment with cisplatin

To verify that UCCs undergo EMT upon long-term treatment with cisplatin, we compared
the mRNA levels of E-cadherin, Vimentin, ZEB1, and Twist1 between paternal and LTT-cells (Fig. 6b). In accord with the morphological changes, expression of the epithelial marker E-cadherin decreased significantly in RT-112-LTT and J82-LTT whereas expression of the mesenchymal
marker Vimentin increased significantly. Likewise, the expression of transcription factors inducing
EMT, such as ZEB1 and Twist1, was increased in the LTT-lines, albeit not significantly in RT-112-LTT. As we observed
less tight cell-cell-contacts after cisplatin treatment, we also quantified the expression
of CLDN3 and CLDN4, encoding structural molecules of tight junctions. Indeed, CLDN3 and CLDN4 expression was decreased in RT-112-LTT (Fig. 6c), derived from an epithelial cell line, but not in in J82-LTT, derived from a cell
line with mesenchymal phenotype. IF staining for E-cadherin in RT-112-LTT and Vimentin
in J82-LTT confirmed the mRNA results (Fig. 6b, d). Further, IF staining for ?-catenin revealed diminished membrane localisation, but
no major shift into the nucleus (Fig. 6e).

Activation of WNT-signalling may contribute to survival of UCCs upon long-term cisplatin
treatment

As some results pointed towards an activation of canonical WNT-signalling in LTT-cells,
we measured downstream targets of the pathway, namely AXIN-2, CCDN1, c-MYC, and PITX2. Indeed, the mRNA expression of these genes was increased in LTT-UCCs compared to
their parental cell lines (Fig. 6f). We further confirmed augmented WNT-pathway activity by the TOP-FOPflash-assay indicating
that the pathway could be activated by transfection of mutant ?-catenin-S33Y in RT-112-LTT,
but not in parental cells (Fig. 6g) 25]. Positive controls for endogenous and inducible ?-catenin activity in HepG2 and 22Rv1
cell lines, respectively, demonstrated the functionality of the assay (Fig. 6g). Next, we investigated whether treatment of LTT sublines with the WNT-inhibitor
niclosamide might revert cisplatin resistance of the cells. To this end we performed
a dose response curve for niclosamide after 72 h treatment (Additional file 5: Figure S4a). We used the colorectal carcinoma cell line HCT-116 as a control for
a cell line with significant WNT-pathway activity 26]. HCT-116 cells were more sensitive towards niclosamide treatment than RT-112-LTT
and J82-LTT, as well as their parental cell lines (Additional file 5: Figure S4a, b). As expected, treatment of HCT-116 cells with niclosamide for 72 h
at IC
₅₀
resulted in downregulation of the WNT/?-catenin targets ?-catenin, Axin-2, and CyclinD1
(Additional file 5: Figure S4c upper panel; p 0.05). We further demonstrated that WNT-signalling activity could be significantly
inhibited in HCT-116 cells by the applied doses of the compound in the TOP-FOPflash-reporter
assay (Additional file 5: Figure S4d). Thus, the compound is active at the applied concentrations in cell
lines with canonical activation of the WNT/?-catenin signalling. However, we did not
observe synergistic effects on cell viability of resistant LTT cell lines by combined
treatment with cisplatin and niclosamide (Fig. 6h) indicating that cisplatin resistance could not be reverted by this inhibitor. Moreover,
treatment with niclosamide did not result in significant expression changes of WNT-
target genes UC cells (Additional file 5: Figure S4c, lower panel).