Main

The importance of exogenously added or endogenously produced soluble signal-transducing ligands in the regulation of cell fate is intensively studied. Although it is widely recognized that a substantial number of these ligands act at interfaces in vivo (that is, constrained to surfaces rather than in solution), studies on the effects of cell exposure to the non–freely diffusible forms of these molecules are just emerging1,2,3,4,5. The design principles associated with bound ligand presentation are fundamental to many tissue-engineering efforts6, and strategies to control the binding of ligands to surfaces and to present immobilized cytokines and growth factors have been an area of active development7. Although these systems have provided evidence that the spatially constrained presentation of proteins can modulate cell responses, only recently has a quantitative analysis of the effects of ligand presentation on signaling activation and cell fate started to be explored8,9. Generally, limitations in the technologies used to date include difficulties in controlling the amount, stability and cellular accessibility of surface-immobilized proteins.

To systematically examine the signaling properties of immobilized cytokines, we developed and tested a method for quantitatively immobilizing functional proteins, with or without added extracellular matrix (ECM), to reactive polymer surfaces. Our method permits straightforward long-term culture of cells, including stem cells, on culture plates with defined signaling-protein concentrations. Our approach relies on generation of thin films (∼5 nm) of alternating maleic acid anhydride co-polymers to which bioactive molecules are immobilized according to different binding schemes10,11,12. Detailed studies on physicochemical properties of the films underpin the use of these substrates to control the function of the attached biomolecules10,13,14. We demonstrated the immobilization strategy for the cytokines LIF and SCF.

As LIF signaling in mESCs is well-characterized, it is an ideal platform to test the robustness and versatility of our technology. Furthermore, the mESC system is demanding in terms of culture requirements and thus puts the described platform to the test. LIF is an interesting molecule to investigate for the cellular effects of immobilized ligand presentation. LIF activates the STAT3 pathway, which supports the self-renewal of mESCs in vitro17,18, as well as MAPK, the activity of which has been shown to induce differentiation in mESCs19,20,21. There is some evidence that the ratio of these pathways may be important in mESC differentiation control22. In vivo, LIF is produced in two forms through expression of alternative transcripts: a diffusible form and a form that associates with the ECM23,24. The expression patterns of diffusible and ECM-associated LIF in the developing embryo are not identical, suggestive of distinct biological roles24. In support of this, forced expression of diffusible and ECM-associated LIF result in different embryological consequences25,26. However, it is unclear whether these alternative modes of LIF presentation have distinct signaling properties, or if expression of the different isoforms serves simply to localize and concentrate LIF to a particular region of the embryo.

Although it has been previously reported that LIF can be retained in a gelatin matrix by photochemical cross-linking, and that this approach can be used to elicit LIF-mediated signaling activation and short-term maintenance of some mESC properties15, this system appears limited in its ability to quantitatively control immobilized ligand presentation. Also, the use of ultraviolet irradiation has been reported to affect the structure and function of proteins16.

We found that immobilized LIF surfaces were capable of long-term (7 d) activation of signaling molecules downstream of LIF, and that the immobilized protein supported mESC pluripotency for at least 6–8 passages (>2 weeks of culture). Immobilized LIF activated STAT3 and MAPK signaling in mESCs in a dose-dependent manner. The mode of ligand presentation did not dramatically affect the differential activation of STAT3 and MAPK, suggesting that in vivo presentation of immobilized versus diffusible LIF may act primarily to localize and concentrate the cytokine. Finally, the immobilization method allows the ligand to be available in combination with ECM coatings, which may facilitate the design of experiments to investigate factor-ECM interactions in a combinatorial manner.

Results

Quantification and stability testing of immobilized LIF

We examined three approaches for LIF immobilization (Fig. 1a–c): covalent attachment to poly(octadecene-alt-maleic anhydride) (POMA), covalent attachment to flexible PEG7 spacer arms tethered to POMA (POMA-PEG7) and noncovalent binding to ECM precoatings deposited on top of hydrolyzed POMA (POMA-matrix). To determine whether LIF could be immobilized at different surface densities in these three configurations we used different solution concentrations of cold LIF that was spiked with 2–6% of 125I-labeled cytokine. Both covalent attachment and matrix physisorption increased with increasing solution concentration of LIF (Fig. 1d). For covalently immobilized LIF on POMA-PEG7, saturation was reached at solution concentrations above 5 μg/ml (achieving a maximum surface density of ∼92 ± 4 ng/cm2). In contrast, covalent immobilization directly to POMA yielded a greater amount of immobilized LIF and saturation was not achieved at the solution concentrations tested. PEG7 decoration generates a more hydrophilic, dynamic interface with a lower density of binding sites as compared to unmodified POMA11 and would therefore be expected to have diminished cytokine binding. For POMA-matrix, LIF physisorbed to matrix protein precoatings in quantities comparable to those for covalent attachment directly to POMA (we did not observe saturation at the concentrations tested). Quantification of immobilized LIF demonstrated that on all three substrates similar amounts of cytokine could be immobilized, at nominal grafting densities of 2, 8 or 75 ng/cm2. However, for each substrate it was necessary to use different overlying solution concentrations to achieve those nominal grafting densities (Supplementary Table 1 online).

Figure 1: Long-term stability and ligand accessibility for different LIF immobilization strategies.
figure 1

(a–c) LIF was immobilized by covalent attachment to POMA (a), covalent attachment to flexible PEG7 spacer arms tethered to POMA (b) and noncovalent binding to ECM coating deposited on top of hydrolyzed POMA (c). Red circles indicate covalent bond. Note that POMA was covalently bound to amino-functionalized glass substrates and is the key component of the immobilization platform. POMA and POMA-PEG7 surfaces were coated with matrix for cell-culture experiments (omitted from the figure for clarity). Only one covalent bond between the protein and the surface is shown, although more bonds are possible. Chemical structures depict the POMA layer and the immobilization mode of LIF. Scheme is not drawn to scale. (d) Quantification and stability of immobilized [125I]LIF after immobilization (day 0) on the three surfaces and after a 3-d incubation in cell culture medium (n = 2, *P < 0.05, **P < 0.01). (e,f) Assay of antibody-accessible LIF on matrix-overlaid slides prepared using the indicated amounts of LIF in solution (e; n = 6) or of immobilized LIF (f; n = 6, *P < 0.05, **P < 0.01, POMA versus POMA-PEG7 or POMA-matrix at respective concentration). Samples were incubated for 3 d in cell culture medium. Total amount of LIF was determined by radioassay of [125I]LIF and antibody-accessible LIF by an immunological assay. All error bars indicate s.d.

Full size image

To study the stability of immobilized LIF under cell culture conditions, we incubated the surfaces with tissue culture medium for 1–6 d, changing medium every 2 d. At day 3, the amount of surface immobilized LIF was reduced for all samples (Fig. 1d). The total amount of LIF released was greater from surfaces generated with higher solution concentrations, though the relative amount lost from each substrate was similar at lower nominal grafting densities. The average protein retention over 3 d was 90 ± 8% (POMA), 79 ± 4% (POMA-PEG7) and 75 ± 15% (POMA-matrix).

Adherent culture of mESCs requires the presence of ECM components for cell attachment; this is relevant for the POMA and POMA-PEG7 configurations for which matrix deposition on top of immobilized LIF may mask cytokine presentation to mESCs. To estimate the surface availability of immobilized LIF in the presence of ECM components, we used an antibody to LIF (Fig. 1e). We immobilized LIF at 0.25 μg/ml and 2.5 μg/ml (solution concentrations), incubated the surfaces for 3 d and washed them to remove any unbound LIF, and then measured the percentage of antibody-detectible immobilized LIF. Although radioisotope- and antibody-based measurements correlated well for POMA-PEG7 and POMA-matrix configurations at the lower concentration of immobilized LIF (76 ± 24 and 71 ± 30% detected, respectively), at the higher concentration, antibody to LIF detected a considerably smaller fraction of immobilized LIF in comparison to the radioisotope-based estimate (9.0 ± 1.5%, 20.3 ± 3.5% and 19.3 ± 3.5% for POMA, POMA-PEG7 and POMA-matrix, respectively). We also determined the amount of antibody-accessible immobilized LIF on the three surfaces generated with two different nominal grafting densities of immobilized LIF (2 and 8 ng LIF/cm2) after conditioning in cell culture medium for 3 d (Fig. 1f). Using the antibody, we detected a considerable fraction (up to 55%) of immobilized LIF, with better accessibility of LIF on POMA-PEG7 and POMA-matrix than on POMA.

Immobilized LIF activates mESC signaling

To determine whether our approaches present LIF in a biologically active form, we examined the signaling responses of mESCs to immobilized LIF by quantitatively examining STAT3 and MAPK phosphorylation using immunocytochemistry and single-cell image analysis26 (Fig. 2a). Both in the presence and the absence of PEG7 spacer, immobilized LIF induced a dose-dependent activation of STAT3 signaling (Fig. 2b). Although the amount of antibody-accessible immobilized LIF was greater when the cytokine was immobilized through PEG7 spacer (Fig. 1f), the presence of the spacer did not appear to alter signaling properties and was also not necessary to enable signaling through an ECM-overlayed matrix. This result, in combination with the observation that both the amount (Fig. 1d) and efficiency (Supplementary Table 1) of immobilization was greater on POMA without PEG7, motivated our in-depth analysis of STAT3 and MAPK signaling on this surface. LIF immobilized on this surface resulted in dose-dependent STAT3 activation (Fig. 2c). For LIF immobilized on the same surface we observed a similar dose-dependent increase in the signaling activation of phosphorylated MAPK (Fig. 2d). To compare the mode of LIF presentation on the activation ratio of these two signaling pathways, we normalized the values for phosphorylated STAT3 and MAPK to the maximum responses and plotted the ratio (Fig. 2e). Our results suggest that the mode of LIF presentation (diffusible versus immobilized) does not differentially affect the STAT3 and MAPK activation ratio.

Figure 2: Immobilized LIF activates STAT3 and MAPK in a dose-dependent manner.
figure 2

(a) Immunocytochemistry and single-cell image analysis of mESCs cultured on 75 ng/cm2 LIF immobilized covalently to POMA. Hoechst nuclear staining with (top left) and without (top right) image algorithm cell mask. Blue masks represent objects fitting selection criteria and orange represents objects excluded from analysis. Nuclear masks were applied to a second channel (bottom left) to quantify phosphorylated STAT3 fluorescence (bottom right). Scale bar, ∼20 μm. (b) STAT3 activation in mESCs cultured for 12 h on surfaces of LIF immobilized to POMA or POMA-PEG7. Diffusible LIF was used as control on hydrolyzed surfaces (3 × 103–1 × 104 single-cell measurements n = 4 ± s.e.m., *P < 0.05, ** P < 0.01; diffusible LIF served as reference for Dunnett's post-hoc test). (c) STAT3 activation in mESCs cultured for 72 h in the presence of of POMA-immobilized LIF and diffusible LIF. (d) Activated (phosphorylated) MAPK was measured for the same cells and conditions as in c. (e) The ratio of phosphorylated MAPK to phosphorylated STAT3. Error bars, s.d. (c–e)

Full size image

mESC maintenance on immobilized LIF

We next examined the ability of LIF to support self-renewal of mESCs, using different surface densities of LIF covalently immobilized (POMA) or absorbed to POMA (POMA-matrix) and overlaid with ECM to allow cell attachment. We used histograms of mESC expression of the pluripotency marker Oct-4 (ref. 27) to calculate the percentage of Oct-4+ and Oct-4− cells under each condition (Fig. 3a). We determined the percentage Oct-4− cells generated after 3 d of culture in cells exposed to diffusible LIF, no LIF or different concentrations of immobilized LIF. As expected from the STAT3 activation studies, immobilized LIF inhibited the generation of Oct-4− cells over 72 h in a dose-dependent fashion. In contrast, matrix-physisorbed LIF was incapable of maintaining Oct-4 expression, and differentiation ensued similar to that in no-LIF controls (Fig. 3b). We also tested the effectiveness of our surfaces to support mESC pluripotency. First, we seeded cells at low cell densities and examined Oct-4 expression for 7 d (Fig. 3c). As predicted from our signaling analysis, only the higher concentrations of immobilized LIF supported Oct-4 expression over this time period, with even the intermediate immobilized LIF concentration yielding low numbers of Oct-4+ cells by day 7. Although Oct-4 activation is associated with undifferentiated mESCs, we wanted to test the ability of immobilized LIF to support mESC pluripotency over many passages. Thus, we compared mESCs passaged on POMA-immobilized LIF surfaces in media devoid of diffusible LIF for greater than 2 weeks (6–8 passages) to mESCs passaged in media containing diffusible LIF, in chimera formation assays using diploid embryos. Note that mESC passaged without any LIF rapidly differentiate (after 2–3 passages) and cannot be extensively propagated. This experiment demonstrates that mESC propagated on covalently immobilized LIF can contribute to embryonic development and support development to term (adulthood) (Fig. 3d and Supplementary Table 2 online).

Figure 3: Immobilized LIF promotes Oct-4 expression and chimera formation in a dose-dependent manner.
figure 3

(a) Representative histograms showing Oct-4 expression for cells cultured for 72 h on 75 ng/cm2 immobilized LIF on POMA or POMA-matrix. Red lines on each plot represent Gaussian fits of Oct-4+ (right peak) and Oct-4− (left peak) subpopulations. The blue lines are the overall fit for both populations. (b) Concentration-dependent expression of Oct-4 in mESCs cultured for 72 h on POMA-immobilized LIF, or matrix-physisorbed LIF. Diffusible LIF was replaced during medium change every 24 h (results obtained from 5 × 103–1 × 104 single-cell measurements; error bars, s.e.m.; *P < 0.05, Student's t-test for respective conditions). (c) Oct-4 expression in cells cultured in the absence of LIF, the presence of indicated amount of LIF immobilized covalently to POMA and diffusible LIF over time. Error bars, s.d. (d) Chimeric mice generated from cells cultured and passaged 6–8 times on POMA substrates with covalently attached LIF (8 ng/cm2), hydrolyzed POMA with diffusible LIF and tissue culture plastic with diffusible LIF (10 ng/ml each). All substrates were overlaid with gelatin to facilitate cell attachment.

Full size image

Our immobilization approach is broadly applicable

To test whether our approach is applicable to cytokines other than LIF, we immobilized defined amounts of bioactive SCF on POMA and POMA-PEG7 (Fig. 4a). Comparable to the trends observed for LIF, POMA substrates allowed greater surface immobilization of SCF compared to covalent binding of SCF through a PEG spacer (Fig. 4b) and both immobilized SCF systems were biologically active and induced the expansion of the erythroleukemia cell line Mo7E (a model system for blood progenitor cell growth) in a concentration-dependent manner (Fig. 4c, d). Considering the geometrically restricted accessibility of the immobilized cytokine for Mo7E cells, which are typically suspension-grown cells and therefore attach only temporarily or loosely to the matrix, our data suggest high signaling efficacy of the immobilized SCF.

Figure 4: Functional immobilization of SCF.
figure 4

(a) Amount of covalently bound SCF after a 1-d incubation under cell culture conditions. Graph indicates the relationship between the solution concentration and the amount of covalently immobilized [125I]SCF on POMA and POMA-PEG7. (b) Longer-term behavior of immobilized SCF on surfaces prepared using 2.5 μg/ml SCF in solution. (c) Analysis of proliferation induction of Mo7E cells by immobilized SCF. Mo7E cells were cultured without medium change on POMA or POMA-PEG7 with immobilized SCF at 3.5–74 ng/well and counted on day 6. Numbers next to plot symbols correspond to the respective surface concentration in ng/cm2. Samples were used after a 1-d preincubation as described in a. Dashed line corresponds to samples with diffusible SCF at 2 ng/ml, dotted line to Mo7E proliferation without SCF. Graphs show typical datasets generated from triplicate samples ± s.d. (d) Phase contrast image of Mo7E cells cultured on polymeric substrate at day 5. Both intimately associated Mo7E cells (open arrow) and suspended cells (closed arrow) are present in culture. Scale bar, 50 μm.

Full size image

Discussion

Strict ligand dosage control is a necessity given the threshold nature of cell fate signals28. In the case of LIF, specific amounts of ligand and consequently of STAT3 activation have been shown to be necessary to support mESC self-renewal29. Our measurements of STAT3 and MAPK activation in response to diffusible and immobilized LIF demonstrate that as one moves from differentiation-permissive to differentiation-inhibitory concentrations of LIF, the ratio of two activated downstream signaling molecules (STAT3 and MAPK) remains relatively constant. Even in the absence of exogenous LIF (wherein locally produced STAT3 activation is significant22), this ratio is not substantially different from that measured at high concentrations of diffusible or immobilized LIF. Together, these data support the hypothesis that it is the extent of activation (the signaling threshold) of the signaling intermediates, not their relative activation level, which determines stem cell fate.

Surfaces with covalently immobilized LIF could be coated with an ECM, and we did not detect temporal or dose-dependent changes in ligand accessibility owing to ECM remodeling. ECM itself is important in modulating cell attachment and fate30, and the possibility of independently controlling ECM and immobilized cytokine exposure to cells should enable screening studies into the interactive effects31 of these two important cell-signaling components. Notably, we observed differences in antibody accessibility between POMA and POMA-PEG7 immobilized LIF, but functional responses (signaling activation) on these surfaces were similar. These results suggest that although matrix may interfere with antibody access to immobilized LIF, cells efficiently accessed immobilized protein through the matrix. In contrast to the condition where LIF was simply physisorbed, the cells did not appreciably degrade immobilized LIF over the period studied (that is, immobilized LIF was active for extended periods of time).

The approach we report here sets the groundwork for an in-depth comparison of other diffusible and immobilized ligands such as activin and nodal (known to be important morphogens in development32) or Notch ligands33 on cell signaling and fate. The approach may be generally applicable to address the importance of exogenous cues in stem cell niches, as indicated by our observations with SCF. Notably, the polymer thin film coating can be applied to many amine-bearing carrier materials, for example, glass or silicone wafers (after pre-treatment with aminosilanes) as well as different bulk polymer materials such as polystyrene or polydimethylsiloxane (after ammonia plasma treatment). This allows adaptation of the coatings to standard multiwell substrates and combination with common techniques of microstructuring and upscaling for surface modification of three-dimensional objects. The versatility and robustness of the substrate functionalization scheme can be expected to facilitate multifactorial cell fate screening experiments.

Methods

LIF immobilization.

We dissolved LIF (recombinant, mouse; Chemicon) in Dulbecco's PBS (pH 7.4), and covalently immobilized it to POMA films directly or via PEG7 spacers (POMA-PEG7). We prepared POMA films as described in Supplementary Methods online. For LIF immobilization to plain POMA surfaces, we incubated freshly annealed polymer films with LIF solutions (0.25–10 μg/ml) overnight at room temperature (18–22 °C) and rinsed them twice with PBS before use. LIF conjugation to POMA-PEG7 was achieved with 50 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) (Sigma-Aldrich) and 25 mM N-hydroxysulfosuccinimide (sulfo-NHS) (Fluka) in phosphate buffer (67 mM, pH 7.4) for 4 h at room temperature. Then we rinsed the surfaces twice with purified water. For noncovalent LIF immobilization to ECM (POMA-matrix), we incubated the POMA-coated coverslips overnight in PBS to hydrolyze the anhydride moieties. Subsequently, we coated the polymer films overnight at 37 °C with a solution of native collagen type I (0.2 μg/ml, Vitrogen; Cohesion Technologies) and fibronectin (12.5 μg/ml, purified from adult human plasma) in PBS34. Then we gently rinsed POMA-matrix samples with PBS and physisorbed LIF as described above. For cell culture experiments, we used gelatin (type B, from lime-cured tissues; Sigma) instead of collagen type I. For all cell culture experiments, we applied gelatin to all other studied surface preparations and their respective controls to facilitate cell attachment. Additional details are available in Supplementary Methods.

Preparation of 96-well plates with POMA substrates.

We detected LIF by immunological assay and carried out cell-culture experiments in glass-bottom 96-well plates (Greiner Bio-One). We coated glass-bottom plates with POMA layers as described above followed by fixation of the plates to polystyrene 96-well grids. To regenerate anhydride groups, we annealed POMA-coated plates for 18 h at 90 °C under vacuum. Then we attached PEG7, LIF or ECM as described above.

Immobilization of predefined amounts of immobilized LIF.

We immobilized 2, 8 and 75 ng LIF/cm2 on either POMA, POMA-PEG7 or POMA-matrix as described above. The corresponding solution concentrations were extrapolated from [125I]LIF calibration curves (see Supplementary Table 1 for solution concentrations).

Quantification of immobilized LIF and stability testing.

We quantified LIF using [125I]LIF (custom-ordered from GE Healthcare), which was immobilized on the respective surface in custom-made incubation chambers (2 cm2 sample surface). We made a dilution series of LIF (0.25–10 μg/ml) in 250 μl of PBS that we spiked with 2–6% of 125I-labeled cytokine (50 μCi/mg protein, <2% free 125I). We shook the samples and incubated them at room temperature for 24 h. After the immobilization period, we washed the samples three times with PBS and measured coverslip-associated radioactivity (gamma counter UMo LB 123; Berthold Technologies). To assess the stability of attachment of immobilized LIF in cell culture medium (DMEM with 15% knockout serum replacement and 2 mM NaN3) we repeated the measurements of coverslip-associated radioactivity on day 3. We quantified the amount of LIF using [125I]LIF standards.

Examination of accessibility of LIF.

We used a LIF-specific antibody to assay the accessibility of immobilized LIF. We prepared samples in a 96-well format, rinsed them twice with PBS and blocked them with 50 μl/well of 5% BSA in Tris-buffered saline (TBS, pH 7.3) for 30 min at room temperature. Subsequently we incubated 50 μl/well of 1.5 μg/ml biotinylated goat antibody to LIF (R&D Systems) in TBS for 2 h at room temperature. Then we rinsed the samples with TBS, added 100 μl/well of horseradish peroxidase–conjugated streptavidin (75 ng/ml in TBS; Rockland) and incubated the samples in the dark for 30 min. Next we rinsed the wells, added 100 μl/well of tetramethylbenzidine (TMB) peroxidase enzyme immunological assay substrate (BioRad), and stopped the reaction after 8 min with 100 μl/well of 0.5 M H2SO4. Optical densities were measured using a microplate reader (Tecan) at 450 nm. We applied the matrix required for cell attachment and determined the antibody accessibility of immobilized LIF after 3 d in cell culture medium.

mESC culture, chimera production, immunocytochemistry and single-cell image analysis.

We used mouse R1 mESCs35 for the experiments; cell culture conditions, staining procedures and analysis are described in Supplementary Methods. In vivo studies were conducted in accordance with national regulations and institutional guidelines. For chimera experiments we directly thawed mESCs, cultured and passaged them on three different matrices: (i) covalently attached LIF (8 ng/cm2) to POMA, which we subsequently overlaid with gelatin, (ii) hydrolyzed POMA or (iii) tissue-culture plastic wells, which we overlaid with gelatin followed by diffusible LIF (10 ng/ml). We subcultured the cells at a split ratio of either 1:4 or 1:6 under normal tissue-culture conditions. We performed two chimera experiments after 6 and 8 passages for each condition. We left the mice to term and determined in vitro–cultured mESC contribution at 3 weeks of age. We formed chimeras with diploid imprinting control region host embryos as previously described36,37.

Statistical analysis.

We analyzed statistical differences using GraphPad Instat 3.06 using Student's t-test for the analysis of 2 groups and where indicated analysis of variance (ANOVA) followed by Tukey's or Dunnett's post-hoc test. We assumed the s.d. of the data to be equal and normally distributed.

Additional methods.

Descriptions of SCF immobilization (similar to the approach reported for LIF), characterization of immobilized SCF and proliferation studies with Mo7E cells are available in Supplementary Methods.

Note: Supplementary information is available on the Nature Methods website.