Green and catalyst-free preparation of triazinyl polyimide for the efficient adsorption of glycoproteins

Abstract

Triazinyl polyimide is prepared via a solvent-free and catalyst-free imidization reaction by using melamine and pyromellitic dianhydride as the precursors. The characterization by means of FT-IR, X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), surface charge analysis and thermogravimetric analysis (TGA) indicates that the obtained triazinyl polyimide is highly crystalline with excellent thermal-stability. The strong hydrogen-bonding interactions between the glycan groups in the glycoproteins and the carbonyl groups in polyimide give the as-prepared triazinyl polyimide excellent adsorption selectivity and an ultra-high adsorption capacity towards glycoproteins, an adsorption capacity of 1666.7 mg g⁻¹ is presented for ovalbumin. This offers the possibility for the removal of glycoprotein species of high abundance from complex sample matrices by using the triazinyl polyimide as a powerful adsorbent. The practical feasibility is well demonstrated through the efficient removal of ovalbumin and conalbumin from egg white, as confirmed using an SDS-PAGE assay.

---

1. Introduction

Nowadays proteomics analysis has become a very hot research area for the design and optimization of bioprocesses, and the acquisition of high purity protein species is the primary basis for the successful and comprehensive understanding of protein function and regulation. In most physiological environments, the protein species of interest always coexists with complex sample matrices, e.g., mineral salt, nucleic acid, polysaccharide, lipid and other proteins. In practice, the analysis of low-abundant proteins is usually disrupted by the matrix interference and the signal suppression from high-abundance protein species.¹ In this respect, efficient sample pretreatment procedures are generally in high demand for the purpose of either eliminating the sample matrix including the high abundance proteins or enhancing the concentration of protein species by means of preconcentration.

In the last decade, a range of materials have been utilized for protein adsorption, including carbon-based materials,^2–4 ionic liquids and their derivatives,^5,6 inorganic nanoparticles,⁷ covalent organic frameworks⁸ and coordination organic polymers.^9,10 By taking advantage of the various binding interactions, e.g., hydrophobic interaction, electrostatic interaction and chelation between the adsorbing materials and the protein species, efficient adsorption of proteins of interest is usually achieved. Glycoprotein is one of the most common post-translational modified proteins and it plays a key role in various physiological processes such as tumor immunology and inflammation.¹¹ It has been reported that more than half of the proteins in mammalian cells belong to glycoproteins.¹² Recently, the areas of hydrazide chemistry and boronic acid chemistry have been demonstrated to be the most commonly adopted methodologies for performing selective adsorption of glycoproteins, in which covalent interactions are among the main driving force for facilitating protein adsorption. In practical applications, a longer conjugation time is needed to retain glycoproteins onto the solid phase¹³ and alkaline conditions might cause the degradation of some unstable glycoproteins¹⁴ in hydrazine and boronic acid chemistry. Lectin affinity is an efficient strategy for the adsorption of glycoproteins based on hydrogen bonding along with hydrophobic interactions and van der Waals forces.^15,16 While the immobilization of lectin onto a supporting matrix generally needs to undergo a series of complex operation processes. Polyimide is a kind of polymer containing imide rings on the main chains. Since the synthesis of the first polyimide in 1908,¹⁷ polyimides have been widely applied in the fields of aerospace, coatings, separations, microelectronics and the optoelectronic industry, due to their high surface area, low density, favorable thermal stability, chemical resistance and mechanical strength.¹⁸ Generally, polyimides are prepared via the condensation of appropriate amines and anhydrides.^19,20 The high content of imide groups and unique structure provide polyimides outstanding capability for binding with other materials via host–guest recognition and hydrogen-bonding interactions.²¹ Thus, polyimide might be a potential suitable replacement for lectin for the purpose of protein adsorption.

In this work, triazinyl polyimide is prepared via a facile green route by using melamine and pyromellitic dianhydride as the precursors without using any catalyst. The obtained polyimide exhibits an ultra-high adsorption capacity and favorable selectivity toward glycoproteins mainly due to hydrogen-bonding interactions between the glycan groups in the glycoproteins and the carbonyl groups in polyimide. In practical applications, high abundance glycoproteins, e.g., ovalbumin and conalbumin in egg white, are efficiently adsorbed and collected or removed, providing a useful approach for the elimination of high abundance glycoproteins to free the non-glycosylated proteins, i.e., lysozyme in this particular case. The removal efficiency is well confirmed using a sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) assay.

2. Experimental

2.1 Materials and reagents

Melamine and pyromellitic dianhydride are purchased from Sinopharm Chemical Reagent Co. Ltd (Shanghai, China) and employed as received. Other reagents including NaCl, Na₃PO₄, Na₂HPO₄ and NaH₂PO₄ are supplied by Tianjin Damao Chemical Reagent Factory (Tianjin, China) and are at least of analytical reagent grade. Deionized (DI) water is used throughout.

Ovalbumin (Ova, A5503, 98%, Mr: 44.3 kDa), conalbumin (ConA, C7786, 98%, Mr: 77.7 kDa), lysozyme from chicken egg white (Lys, L6876, 90%, Mr: 14.3 kDa), immune globulin from human serum (IgG, 14 506, 95%, Mr: 150.0 kDa), lactoferrin from bovine milk (bLf, L9507, 85%, Mr: 86.0 kDa) and bovine serum albumin (BSA, A3311, >98%, Mr: 66.4 kDa) are purchased from Sigma-Aldrich (St. Louis, USA). These proteins are used without further purification. The protein molecular weight marker (broad, D532A, Takara Biotechnology Company, Dalian, China) is a mixture of nine purified proteins (Mr in kDa: myosin, 200; β-galactosidase, 116; phosphorylase B, 97.2; serum albumin, 66.4; ovalbumin, 44.3; carbonic anhydrase, 29; trypsin inhibitor, 20.1; lysozyme, 14.3; aprotinin, 6.5).

2.2 Preparation and characterization of triazinyl polyimide

Triazinyl polyimide is prepared via a modified route according to the reported procedure.²² Melamine (MA, 0.631 g, 5 mmol) and pyromellitic dianhydride (PMDA, 1.636 g, 0.75 mmol) are mixed and ground in an agate mortar. The mixture is transferred into a porcelain ark, followed by calcination in a tube furnace at 285 °C for 4 h with a heating rate of 7 °C min⁻¹ at ambient atmosphere. The obtained yellowish product is washed with DI water at 60 °C to remove the unreacted MA and PMDA. Thereafter, the product is dried at 70 °C in a vacuum oven overnight and the final product is obtained (shortly termed as MA-PI).

Fourier transform infrared spectra (FT-IR) of MA, PMDA and MA-PI are obtained on a Nicolet 6700 spectrometer (Thermo Electron, USA) from 400 to 4000 cm⁻¹. An X-ray photoelectron spectroscopy (XPS) scanning curve for MA-PI is recorded on an ESCALAB 250 surface analysis system (Thermo Electron, England). The powder XRD patterns of the materials are obtained by using an X’Pert Pro MPD X-ray diffractometer (PW3040/60, PANalytical BV, Holland) with Cu Kα irradiation at room temperature. The surface charge property is recorded on a ZEN3600 Nano Zetasizer (Malvern, UK) by dispersing MA-PI into DI water at a mass ratio of 1%. The thermo-gravimetric analysis (TGA) is performed using a TGA/DSC 1 STAR^e System (Mettler-Toledo, Switzerland) from 30 to 800 °C with a heating rate of 10 °C min⁻¹ under a nitrogen atmosphere. The surface morphology of MA-PI is recorded on an SSX-550 scanning electron microscope (Shimadzu, Japan). Nitrogen adsorption–desorption of MA-PI is measured at 77 K using a Micromeritics Tristar 3000 analyzer (USA) to obtain its specific surface area.

2.3 Protein adsorption with MA-PI as the adsorbent

In the present work, Ova (pI 4.7), ConA (pI 6.8–7.1), Lys (pI 10.5–11) and BSA (pI 4.7) are employed as model proteins to investigate the protein adsorption behaviors onto the surface of MA-PI. The pH value of the medium is adjusted using 0.1 mol L⁻¹ HCl and 0.1 mol L⁻¹ NaOH.

Generally, 1.0 mg of MA-PI is added into 1.0 mL of protein solution and the mixture is shaken vigorously for 30 min to facilitate the adsorption of proteins onto the surface of MA-PI. The supernatant is collected after centrifugation at 8000 rpm for 8 min and the concentration of the proteins in the aqueous phase is obtained by measuring their characteristic absorption at 595 nm after the proteins have been stained with Coomassie Brilliant Blue dye. The adsorption efficiency is calculated based on the following equation, with E, C₀ and C₁ as the adsorption efficiency, the protein concentration in the original solution and in the supernatant, respectively.

3. Results and discussion

3.1 Preparation and characterisation of MA-PI

The solvent-free procedure for the preparation of MA-PI is illustrated in Scheme 1. The preparation is carried out at a temperature of 285 °C, at which the melted PMDA acts as both the reactant and the reaction medium and thus avoids the use of toxic high boiling-point solvents. During the preparation process, the polycondensation of MA-PI is accompanied by dehydration, and the water molecules produced in the imidization reaction are evaporated rapidly at the reaction temperature, which further promotes the condensation degree between the MA and PMDA. After the completion of the condensation of MA and PMDA, a yellow imide product MA-PI is produced. Due to the densely cross-linked structure the obtained MA-PI exhibits excellent solvent-stability, i.e., it is insoluble in aqueous medium and other common organic solvents including methanol, dimethylformamide, tetrahydrofuran and even in dimethyl sulfoxide.

Scheme 1 Schematic illustration of the solvent-free preparation of MA-PI.

FT-IR spectra of MA, PMDA and MA-PI are shown in Fig. 1. MA-PI shares three identical characteristic absorptions at 1658, 1546 and 1446 cm⁻¹ with MA, corresponding to the stretching vibrations of the triazine ring. The attenuation of absorptions of MA-PI at 1860 and 1224 cm⁻¹, assigned to the anhydride groups, demonstrates the consumption of the anhydride groups in the reaction process. The replacement of the oxygen atoms between the two carbonyl groups with nitrogen atoms in the imidization reaction gives rise to an obvious blue-shift of stretching vibration from 1780 to 1710 cm⁻¹ in the spectrum of MA-PI. The shifts of absorptions from 3130 to 3208 cm⁻¹ and 1030 to 1060 cm⁻¹ are contributed to by the change of chemical environment around the terminal N–H bond with respect to that in MA. It should be noted that even though there are primary amino groups existing at the frontier of the MA-PI framework, MA-PI is negatively charged (Fig. S1†). This might be caused by the hydrolysis of unreacted anhydride groups in the aqueous medium.

Fig. 1 FT-IR spectra of melamine (MA), pyromellitic dianhydride (PMDA) and triazinyl polyimide (MA-PI).

The XPS spectrum (Fig. 2A) of MA-PI shows three peaks at 531 eV, 399 eV and 284 eV, that are attributed to O 1s, N 1s and C 1s. In the high-resolution N 1s spectra (Fig. 2B), the appearance of the peak at 399.2 eV, assigned to the imide group, is distinct evidence of the occurrence of the imide reaction. The peak at 398.2 eV is contributed to by the carbon–nitrogen double bond of the triazine ring, indicating that it is kept integrated in the imidization process. The peak at 399.6 eV, related to the primary amino group, demonstrates the existence of –NH₂ at the edge of MA-PI. This observation demonstrates the successful preparation of triazinyl polyimide.

Fig. 2 (A) XPS spectrum of the prepared triazinyl polyimide (MA-PI). (B) XPS spectra of triazinyl polyimide (MA-PI) illustrating the high resolution peaks for N 1s.

Fig. 3A shows the TGA analysis results for MA, PMDA and MA-PI. It is seen that the presence of active amino and anhydride groups causes obvious weight loss for MA and PMDA at a temperature of >283 °C and >232 °C, respectively. MA-PI exhibits a higher thermal stability than PMDA and MA, contributed to by the imide groups formed after the imidization reaction. Two distinct weight loss stages were observed in the TGA curve of MA-PI, the first starts from 310 °C and is attributed to the decomposition of the functional groups on the edge of MA-PI, while the second at 399 °C is most probably due to the degradation of the polymer backbones. Fig. 3B illustrates the XRD patterns of MA, PMDA and MA-PI. It could be seen that two new sharp diffraction peaks at 18.9 and 29.5° appear in the XRD patterns of MA-PI, suggesting that the obtained MA-PI is of high crystallinity. The high degree of crystallinity might originate from the improvement on the chain orientation and molecular ordering of the crystal domains in the polymer.²³ During the thermal annealing process, the hydrogen bond and π–π stacking interaction between the conjugated core units facilitates the crystallization of the early-formed oligomers.

Fig. 3 TGA curves (A) and XRD patterns (B) of melamine (MA), pyromellitic dianhydride (PMDA) and triazinyl polyimide (MA-PI).

The layered-sheet morphology as illustrated in the SEM image of MA-PI (Fig. S2†) is due to the strong π–π stacking interaction between polymer layers, which is consistent with the XRD pattern of MA-PI. The surface area of MA-PI is measured using a N₂ adsorption/desorption experiment (Fig. S3†). The BET surface area of MA-PI is calculated to be 7.322 m² g⁻¹.

3.2 Protein adsorption behavior

In the present study, non-glycoproteins, BSA and Lys, and glycoproteins, Ova and ConA, are used as models to investigate their adsorption behaviors onto the MA-PI surface within a pH range 3–12, as illustrated in Fig. 4. It is obvious that for the non-glycoprotein species a maximum adsorption efficiency is achieved at pH values close to their isoelectric points, i.e., pH 5 for BSA and pH 11 for Lys. At pH values of the sample solution departing from their pIs, a decline in the adsorption efficiency is obviously observed for BSA and Lys. Considering the fact that the surface of MA-PI is always negatively charged within the pH range tested, it is safely concluded that electrostatic interactions are not among the driving forces for the adsorption of non-glycoproteins. It has been demonstrated that the inner hydrophobic regions of the proteins are prone to be exposed at pH conditions close to their isoelectric point,^24,25 thus hydrophobic interactions between the protein species and the MA-PI surface might contribute greatly to the protein adsorption.

Fig. 4 The pH-dependent adsorption behaviors of Ova, ConA, Lys and BSA onto the surface of MA-PI. Concentration/volume of protein solution: 70 mg L⁻¹/1.0 mL; amount of adsorbent: 1.0 mg; adsorption time: 30 min.

It is known that Ova shares a very similar isoelectric point to that of BSA, while it is interesting to see that the adsorption behavior of Ova is very much different from BSA. A much higher adsorption efficiency is recorded for Ova with respect to that for BSA within the whole pH range studied, with a maximum adsorption efficiency of >88%. Ova is a representative glycoprotein and it possesses a solvent-exposed carbohydrate moiety consisting of 4–6 mannose residues and 2–4 N-acetyl-b-D-glucosamine residues.²⁶ It has been demonstrated that hydrogen bonds play a vital role in lectin–carbohydrate interactions.¹⁶ Our previous study confirmed that hydrogen-bonds could be formed between the hydroxyl groups on the carbohydrate chains of glycoproteins and the oxygen atoms of solid supports.²⁷ Considering the abundant carbonyl groups and N heteroatoms on the surface of MA-PI, it is quite reasonable to form hydrogen bonds between the O/N atoms of MA-PI and the glycan groups of the glycoprotein, and the hydrogen-bonding interactions become another important driving force to promote the adsorption of Ova. For the case of BSA, there are no glycan groups in its framework structure, thus no hydrogen-bonds would be formed between MA-PI and BSA and therefore the adsorption of BSA is much lower than that of Ova. On the other hand, it should be noted that the protonation of MA-PI in acidic medium and the competition from free OH⁻ in alkaline medium pose the potential for deterioration on the adsorption of Ova, and thus render a decrease of the adsorption efficiency when the pH departs from the maximum adsorption point on both sides. The above observations clearly demonstrate the feasibility of performing selective adsorption of glycoproteins from other non-glycoproteins.

To further demonstrate the effect of hydrogen-bonding interactions on the adsorption of glycoproteins onto MA-PI, the adsorption behavior of ConA is investigated. ConA is a glycoprotein coexisting with Lys and Ova in egg white. It is found that the adsorption behavior of ConA within the pH range tested is quite similar to that of Ova, and a maximum adsorption is achieved in a neutral medium. At the same time, it is seen that the adsorption efficiency of ConA is obviously higher than that of BSA. These observations confirm the contribution of hydrogen-bonding interactions between MA-PI and the glycan groups of the glycoprotein. It should be noted that the adsorption efficiency of ConA is much lower than that of Ova. The reason might lie in the fact that ConA contains a lower carbohydrate content and its sugar chain is masked and buried under the protein surface,²⁸ thus the inaccessibility of the glycosidic bond would hinder the formation of hydrogen bonds. To further clarify the universal effect of the hydrogen-bond interactions between the glycoprotein and MA-PI, the adsorption behaviors of another two glycoproteins, IgG and bLf, are investigated at the same conditions and the results are shown in Fig. S5.† A similar adsorption behavior with ConA is obtained even though they have very different isoelectric points (pH 6.8 for ConA, pH 8 for IgG, pH 8.2 for bLf) and molecular weights. This well elucidated the contribution of hydrogen-bond interactions in the glycoprotein adsorption by MA-PI.

Generally, a salt in solution could remove water molecules around protein structures and facilitate the exposure of inner hydrophobic regions of proteins.^24,25 Thus, it is expected that the introduction of salt into the protein adsorption system would facilitate the contact between the glycan groups of the glycoproteins and MA-PI, giving rise to an improvement of the adsorption. ConA is selected as a model glycoprotein to investigate the effect of salts in glycoprotein adsorption. This has been clearly demonstrated in Fig. 5, which illustrates the variation of adsorption performance of ConA with ionic strength, i.e., various concentrations of NaCl. A sharp increase of adsorption efficiency, i.e., from 31.4% to 97.4%, is observed when varying the NaCl concentration from 0 to 0.4 mol L⁻¹.

Fig. 5 Dependence of ConA adsorption efficiency on the variation of ionic strength within a range of 0–0.8 mol L⁻¹ NaCl. Concentration/volume of protein solution: 70 mg L⁻¹/1.0 mL; pH: 7; amount of adsorbent: 1.0 mg; adsorption time: 30 min.

3.3 Selective removal of a highly abundant glycoprotein from egg white

The above observations demonstrate well that glycoproteins could be selectively bound onto the surface of the as-prepared MA-PI via hydrogen-bonding and hydrophobic interactions under controlled conditions. This indicates that MA-PI could be used as a promising adsorbent for the adsorption of glycoproteins by mimicking the lectin-affinity.

Taking into consideration that glycoproteins Ova and ConA coexist with non-glycoprotein Lys in egg white, their dynamic adsorption behaviors onto MA-PI are further investigated at room temperature within a concentration range of 10–500 μg mL⁻¹ in a neutral medium with an ionic strength of 0.4 mol L⁻¹ NaCl. The results are illustrated in Fig. 6.

Fig. 6 The adsorption isotherms of Ova, ConA and Lys on MA-PI. Protein solution: 0–500 μg mL⁻¹, 1.0 mL; amount of adsorbent: 1.0 mg; pH 7; adsorption time: 30 min.

The experimental data are fitted with a Langmuir adsorption model as expressed in the following equation, with C* and Q* as the protein concentration in the aqueous solution and the amount of protein retained by MA-PI. Q_m is the maximum adsorption capacity and K_d denotes the dissociation constant.

The maximum adsorption capacity for Ova and ConA are deduced to be 1666.7 mg g⁻¹ and 769 mg g⁻¹ respectively. The super high adsorption capacity facilitates their efficient removal from other protein species.

For the comparison of protein adsorption performance, Table 1 summarizes the adsorption equilibrium time and adsorption capacities for Ova by some other adsorbents reported very recently. It can be seen that MA-PI exhibits fast adsorption and a much improved adsorption capacity towards Ova with respect to other adsorbent materials attributed probably to its abundant binding sites and structural regularity.

Table 1 Comparisons of the adsorption capacity of Ova by various adsorbents, including triazinyl polyimide (MA-PI)

Adsorbents	Adsorption time	Adsorption capacity (mg g^−1)	Ref.
Cobalt mono-substituted silicotungstic acid doping aniline (SiW11Co/PANI)	10 min	200.0	27
Chitosan-based amphoteric membrane	12 h	172.0	29
Boronic acid functionalized Fe3O4 NPs	Overnight	777.9	30
Polymeric ionic liquid–graphene composite	30 min	917.4	31
Polymeric ionic liquid@SiO2 composite	30 min	333.3	6
Mesoporous silica nanoparticles	4 h	72.0	32
Triazinyl polyimide (MA-PI)	30 min	1666.7	This work

---

To demonstrate the practical usefulness of MA-PI for other glycoproteins in the presence of non-glycoproteins, a protein mixture with 200 μg mL⁻¹ Ova, 100 μg mL⁻¹ IgG, ConA, bLf and Lys before and after adsorption by MA-PI is adopted for the SDS-PAGE assay. After treatment with MA-PI, the bands for Ova and ConA disappeared, while those for IgG, bLf and Lys diminished slightly or remained virtually unchanged, demonstrating the selectivity of MA-PI for glycoprotein.

As a practical example, egg-white is taken for illustrating the adsorption of glycoproteins by MA-PI. In egg-white, the content of non-glycoprotein Lys is quite low with respect to that of glycoproteins Ova and ConA, and the efficient removal of Ova and ConA is quite challenging. The as-prepared MA-PI exhibits an excellent adsorption capacity and favorable selectivity towards glycoproteins, it is adopted for the removal of Ova and ConA from chicken egg white. Egg white is 10-fold diluted using 0.4 mol L⁻¹ NaCl solution, following centrifugation at 8000 rpm for 20 min. 67 μL of the supernatant, 5 mg of MA-PI and 933 μL of 0.4 mol L⁻¹ NaCl solution are sequentially added into a 1.5 mL centrifuge tube. The mixture is then shaken for 30 min, to allow the adsorption of the protein, followed by centrifugation at 8000 rpm for 8 min, and afterwards the supernatant is collected and employed for the SDS-PAGE assay. It should be noted that the low content of Lys in the egg white leads to difficulty for its identification in the electrophoretogram. Thus 100 μg mL⁻¹ of Lys is spiked into the diluted egg white prior to the adsorption process to further demonstrate the adsorption selectivity, as illustrated in Fig. 7. It is obvious that clear bands of Ova and ConA are observed in the diluted egg white in lane 2 and lane 4. After treatment with MA-PI, the bands of Ova and ConA disappear while that for Lys and its intensity remain virtually unchanged in lane 3 and lane 5. These results suggest that Ova and ConA of high abundance in egg white have been efficiently removed using the adsorption onto MA-PI, while Lys remains in the sample solution. This observation well demonstrated the selectivity and practical utility of MA-PI for the adsorption of glycoproteins.

Fig. 7 SDS-PAGE assay results. Lane 1: marker (kDa); lane 2: 150-fold diluted original egg white; lane 3: the supernatant after adsorption using the MA-PI; lane 4: 150-fold diluted original egg white spiked with 100 μg mL⁻¹ Lys; lane 5: the egg white sample spiked with 100 μg mL⁻¹ Lys and its supernatant undergoing adsorption using MA-PI; lane 6: Ova and ConA standard solution of 200 μg mL⁻¹.

4. Conclusions

A highly crystalline polyimide is prepared using a facile green one-step approach through an imidization reaction based on thermal condensation. Through hydrogen bonds between the O/N atoms of MA-PI and the glycan groups of the glycoprotein, along with the aid of hydrophobic interactions, MA-PI exhibits favorable adsorption performance towards glycoproteins by mimicking the lectin–carbohydrate behaviors. The successful removal of high abundant glycoproteins Ova and ConA from complex biological sample matrices has been demonstrated with MA-PI as the adsorbent. This result not only demonstrates the great potential of polyimide materials in the adsorption/extraction of glycoproteins, but also provides an attractive strategy for the elimination of high abundance glycoproteins when low abundant non-glycoprotein species are under investigation.

Acknowledgements

The authors appreciate the financial support of the Natural Science Foundation of China (21275027, 21235001 and 21475017), and the Fundamental Research Funds for the Central Universities (N140505003, N150502001).

Notes and references

1. G. Such-Sanmartin, E. Ventura-Espejo and O. N. Jensen, Anal. Chem., 2014, 86, 1543–1550.
2. J. W. Liu, T. Yang, S. Chen, X. W. Chen and J. H. Wang, J. Mater. Chem. B, 2013, 1, 810–818.
3. Y. Zhang, J. W. Liu, X. W. Chen and J. H. Wang, J. Mater. Chem. B, 2015, 3, 983–989.
4. Y. Zhang, L. G. Xing, X. W. Chen and J. H. Wang, ACS Appl. Mater. Interfaces, 2015, 7, 5116–5123.
5. Q. X. Mao, H. Wang, Y. Shu, X. W. Chen and J. H. Wang, RSC Adv., 2014, 4, 8177–8182.
6. L. Han, Y. Shu, X. F. Wang, X. W. Chen and J. H. Wang, Anal. Bioanal. Chem., 2013, 405, 8799–8806.
7. S. S. Tang, W. J. Wang, X. W. Chen and J. H. Wang, Anal. Bioanal. Chem., 2011, 401, 3283–3292.
8. S. Kandambeth, V. Venkatesh, D. B. Shinde, S. Kumari, A. Halder, S. Verma and R. Banerjee, Nat. Commun., 2015, 6, 6786.
9. Y. P. Zhu, Y. L. Liu, T. Z. Ren and Z. Y. Yuan, Nanoscale, 2014, 6, 6627–6636.
10. Z. Y. Gu, C. X. Yang, N. Chang and X. P. Yan, Acc. Chem. Res., 2012, 45, 734–745.
11. K. Ohtsubo and J. D. Marth, Cell, 2006, 126, 855–867.
12. R. S. Haltiwanger and J. B. Lowe, Annu. Rev. Biochem., 2004, 73, 491–537.
13. Y. Zhang, M. Yu, C. Zhang, Y. L. Wang, Y. Di, C. C. Wang and H. J. Lu, Chem. Commun., 2015, 51, 5982–5985.
14. J. Tang, Y. C. Liu, D. W. Qi, G. P. Yao, C. H. Deng and X. M. Zhang, Proteomics, 2009, 9, 5046–5055.
15. S. Elgavish and B. Shaanan, Trends Biochem. Sci., 1997, 22, 462–467.
16. S. Nathan, Trends Biochem. Sci., 1993, 18, 221–226.
17. H. Ohya, V. V. Kudryavtsev and S. I. Semenova, Polyimide membranes: applications, fabrications, and properties, CRC Press, 1996, p. 1.
18. D. J. Liaw, K. L. Wang, Y. C. Huang, K. R. Lee, J. Y. Lai and C. S. Ha, Prog. Polym. Sci., 2012, 37, 907–974.
19. K. Yano, A. Usuki and A. Okada, J. Polym. Sci., Part A: Polym. Chem., 1997, 35, 2289–2294.
20. N. Ritter, M. Antonietti, A. Thomas, I. Senkovska, S. Kaskel and J. Weber, Macromolecules, 2009, 42, 8017–8020.
21. R. R. Gu, J. Yao, X. Fu, W. Zhou and D. H. Qu, Chem. Commun., 2015, 51, 5429–5431.
22. S. Chu, Y. Wang, Y. Guo, P. Zhou, H. Yu, L. Luo, F. Kong and Z. Zou, J. Mater. Chem., 2012, 22, 15519–15521.
23. M. L. Chabinyc, M. F. Toney, R. J. Kline, I. McCulloch and M. Heeney, J. Am. Chem. Soc., 2007, 129, 3226–3237.
24. D. Turkmen and A. Denizli, Colloids Surf., B, 2014, 123, 859–865.
25. J. A. Queiroz, F. A. P. Garcia and J. M. S. Cabral, J. Chromatogr. A, 1996, 734, 213–219.
26. J. De Groot, H. A. Kosters and H. H. J. de Jongh, Biotechnol. Bioeng., 2007, 97, 735–741.
27. D. D. Zhang, Q. Chen, L. L. Hu, X. W. Chen and J. H. Wang, J. Mater. Chem. B, 2015, 3, 4363–4369.
28. C. Q. Wang, M. Eufemi, C. Turano and A. Giartosio, Biochemistry, 1996, 35, 7299–7307.
29. Z. C. Feng, Z. Z. Shao, J. R. Yao, Y. F. Huang and X. Chen, Polymer, 2009, 50, 1257–1263.
30. S. T. Zhang, X. W. He, L. X. Chen and Y. K. Zhang, New J. Chem., 2014, 38, 4212–4218.
31. J. W. Liu, M. M. Wang, Y. Zhang, L. Han, X. W. Chen and J. H. Wang, RSC Adv., 2014, 4, 61936–61943.
32. D. Mahony, A. S. Cavallaro, F. Stahr, T. J. Mahony, S. Z. Qiao and N. Mitter, Small, 2013, 9, 3138–3146.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra05411k

---