‘On-the-fly’ optical encoding of combinatorial peptide libraries for profiling of protease specificity

Abstract

Large solid-phase combinatorial libraries currently play an important role in areas such as infectious disease biomarker discovery, profiling of protease specificity and anticancer drug discovery. Because compounds on solid support beads are not positionally-encoded as they are in microarrays, innovative methods of encoding are required. There are many advantages associated with optical encoding and several strategies have been described in the literature to combine fluorescence encoding methods with solid-phase library synthesis. We have previously introduced an alternative fluorescence-based encoding method (“colloidal barcoding”), which involves encoding 10–20 μm support beads during a split-and-mix synthesis with smaller 0.6–0.8 μm silicacolloids that contain specific and identifiable combinations of fluorescent dye. The power of this ‘on-the-fly’ encoding approach lies in the efficient use of a small number of fluorescent dyes to encode millions of compounds. Described herein, for the first time, is the use of a colloid -barcoded library in a biological assay (i.e., protease profiling) combined with the use of confocal microscopy to decode the colloidal barcode. In this proof-of-concept demonstration, a small focussed peptide library was optically-encoded during a combinatorial synthesis, incubated with a protease (trypsin), analysed by flow cytometry and decoded viaconfocal microscopy. During assay development, a range of parameters were investigated and optimised, including substrate (or probe) loading, barcode stability, characteristics of the peptide-tagging fluorophore, and spacer group configuration. Through successful decoding of the colloidal barcodes, it was confirmed that specific peptide sequences presenting one or two cleavage sites were recognised by trypsin while peptide sequences not cleavable by trypsin remained intact.

---

I. Introduction

Large combinatorial libraries currently play an important role in areas such as infectious disease biomarker discovery, profiling of protease specificity and anticancer drug discovery.^1–3 The most powerful combinatorial library production method is the iterative ‘split-and-mix’ synthesis on insoluble microscopic beads.⁴ In a split-and-mix process, libraries containing millions of different compounds can be synthesised onto solid support beads in a small number of reaction cycles, resulting in each bead containing a different compound.

Because compounds on solid support beads are not positionally-encoded as they are in microarrays, innovative methods of encoding are required. The conventional strategy for encoding split-and-mix libraries involves covalent binding of chemical ‘identifier’ tags (e.g.nucleic acids, secondary amines, fluorophenyl ethers)⁵ to the beads in parallel with the compound synthesis. However, there are many advantages associated with the use of optical encoding, rather than chemical encoding, and several attempts have been made to combine fluorescence encoding methods with solid-phase library synthesis.^6–10 These methods generally use “pre-encoded” support beads; that is, the beads are encoded before the library synthesis proceeds.

We have previously introduced an alternative fluorescence-based encoding method (“colloidal barcoding”), which involves encoding large support beads during split-and-mix synthesis with smaller silicacolloids that contain specific and identifiable combinations of fluorescent dye.^11–14 Demonstrated here, for the first time, is the use of a colloid -barcoded library in a high-throughput biological assay (i.e., protease profiling) as well as the use of confocal microscopy to decode the colloidal barcode. The unique concept of colloidal barcoding is briefly outlined below, with the complete assay process described in Fig. 1.

Fig. 1 Application of colloidal barcoding in split-and-mix synthesis and bead-based protease profiling. (1) Support beads (10 μm) are fractioned into several portions. Polyelectrolyte-coated fluorescent silica ‘reporter’ particles (0.65 μm) are attached to the beads while a specific amino acid is reacted. The beads are recombined to complete the cycle and fractioned again. A distinguishable and unique batch of reporters is used to encode each individual reaction in the split-and-mix synthesis. At the end of the synthesis, the N-terminus of each peptide is labelled with a fluorophore which is spectrally distinguishable from the reporters. (2) Beads are incubated with a protease, leading to the eventual proteolysis of the cleavable substrates and thus the release of the terminal fluorophore into solution. (3) Beads are run through a flow cytometer to identify and sort those that contain bioactive peptides according to the decrease in on-bead fluorescence. Fluorescence detection techniques enable decoding of the colloidal barcode, thereby identifying the complete sequence of the substrate synthesised on each bioactive bead.

The colloid -barcoded libraries differ radically from libraries which have been “pre-encoded” with fluorescent dyes (including those libraries used in our previous protease profiling work^7,10). The colloidal barcoding approach has several powerful advantages: (i) the reaction history is encoded onto the beads in parallel with a split-and-mix synthesis; (ii) large libraries can be produced in a low number of reaction cycles and with a low number of encoding dyes; (iii) the procedure produces large libraries in a format that is ready to read and decode. Indeed, the power of this method is such that with just 6 fluorescent dyes, a random octapeptide library containing every combination of 8 chosen amino acids (over 16 million peptides) could feasibly be completely and uniquely encoded.¹¹

In our typical split-and-mix process, 10 μm TentaGel solid support beads in dimethylformamide (DMF) are fractioned into several portions and mixed with a unique reporter suspension containing 0.65 μm, fluorescent colloidal particles (‘reporters’) in DMF (see Fig. 1). The TentaGel beads in each portion become coated with multiple reporters (>50 reporters per bead) of the same type (i.e., the same dye combination). (Enhancement of the adhesion between the reporters and the solid support beads is achieved through layer-by-layer accumulation of polyelectrolyte multilayers on the reporters prior to their addition into the library synthesis.^11–14 Optimisation of electrostatic forces and polymeric flocculation under solvent conditions enables permanent adhesion of the reporters to the support beads, thereby preventing particle dislodgment or exchange which would interfere with barcode reading.^14,15) Each portion is then reacted with a different amino acid and the portions are washed and recombined to complete a cycle. The above process is repeated for a chosen number of cycles, using a unique reporter suspension for each reaction. This results in an optically-encoded peptide library.

In order to optically record the reaction history on the surface of each support bead (i.e., by forming the colloidal barcode), a distinguishable and unique batch of reporters is used to encode each individual portion throughout the entire library synthesis.

After the synthesis, the N-terminus of each peptide in the library is labelled with a fluorophore which is spectrally distinguishable from the reporters. Libraries are incubated with a protease (e.g. trypsin in this proof-of-concept demonstration), leading to the eventual proteolysis of the cleavable substrates and thus the release of the terminal-fluorophore into solution. This release can be measured through the increase of the amount of the fluorophore in the supernatant (e.g. viafluorimetry), as well as a decrease in the fluorescence intensity of any bead on which fluorophore-labelled, ‘bioactive’ peptide sequences had been cleaved through proteolysis (e.g. viaflow cytometry).

Deciphering the barcode on any individual support bead is achieved simply by using either fluorescence microscopy or any automated fluorescence detection instrument, which can detect both the location of each reporter on the bead and the colour combination within each reporter particle. This is typically accomplished by the application of multiple optical filters. By investigating the presence or absence of each fluorescent dye within each reporter on the bead, the identity of each reporter becomes known. The beads in a hexamer library, for example, will have 6 different reporter types (each containing a unique combination of fluorescent dyes) and there will be multiple copies (20–200) of each reporter type attached in order to provide sufficient redundancy. Through reference to the original reaction schema for the combinatorial synthesis, it is possible to determine the order in which each reporter type was added, the identity of the amino acid that was reacted when this reporter type was attached to the bead and therefore the exact peptide sequence which was synthesised on the bead.

In this proof-of-principle demonstration, a focussed combinatorial peptide library was synthesised and encoded via the colloidal barcoding method. After incubation with a model protease, trypsin, confocal microscopy imaging of the bioactive beads successfully enabled decoding of each colloidal barcode, thereby revealing the peptide sequence synthesised onto the support bead. A variety of parameters were also investigated, including optimisation of the loading of probe biomolecules on the support bead, barcode stability, characteristics of the peptide-tagging fluorophore, and spacer group configuration.

II. Results

II.1 Stability study of the reporters

Fluorescent silica reporters were synthesised through a modified Stöber procedure.¹⁵ A succinimidyl ester dye was first coupled to a silane coupling agent (3-aminopropyl)trimethoxysilane (APS) to form an APS–dye conjugate. The conjugate is then incorporated into a tetraethoxysilane (TEOS) silica matrix during its condensation. Resulting particles were monodisperse (650 ± 20 nm) after washing and filtration procedures. Reporters were synthesised using amine reactive Bodipy fluorescent dyes (Invitrogen, USA), including Bodipy® 493 (B493), Bodipy® 530 (B530), Bodipy® 580 (B580) and Bodipy® 650 (B650).

In order to assess the reporter stability towards Fmoc–tBupeptide synthesis reagents, a large excess of polyelectrolyte-coated B530 reporters was mixed with TentaGel (TG) beads in dimethylformamide (DMF) (approximately 10⁹ reporters for 10⁴ TG beads) in the first cycle only. One hexapeptide polylysine was then synthesised onto TG by solid-phase synthesis using Fmoc–tBu chemistry (without further addition of reporters). The average number of reporters per TG bead was counted after each step of the peptide synthesis by confocal microscopy (Table 1). Only a small number of reporter particles were desorbed after each cycle of peptide synthesis, i.e. coupling–washings–piperidine treatment (C₁ to C₆). In fact, the final TFA treatment did not result in the desorption of reporters, thus confirming the strong electrostatic association between the reporter particles and beads throughout the synthetic process. Fluorimetric measurements after each Fmoc-deprotection demonstrated that the fluorescent reporter particles were not affected by the basic conditions of the 20% piperidine in DMF. Extended experiments (data not shown) established that continuous immersion of the reporter-coated TG beads in the 94% TFA cleavage solution did not significantly affect the reporter fluorescence over a 24 hour period (corresponding to the attachment of 19 amino acids in our experimental conditions).

Table 1 Influence of the peptide synthesis cycles on reporter fluorescence intensity and coverage of TG beads

Step	No. of B530 reporter per bead^a	Normalised fluorescence intensity^b (%)
a B530 reporter-bound beads were imaged by confocal microscopy with 488 nm laser excitation. Only the reporters visible on the apparent area of the beads were counted. The values indicated are an average of 10 images of different beads. b Images obtained were quantified with Image J software.
t 0	195 ± 13	100 ± 4
C1	182 ± 9	99 ± 2
C2	159 ± 10	100 ± 2
C3	147 ± 12	100 ± 4
C4	115 ± 8	100 ± 1
C5	102 ± 7	98 ± 4
C6	78 ± 13	100 ± 3
TFA	81 ± 6	83 ± 8

---

To permit the accurate identification of barcodes during a split-and-mix synthesis, the cross-contamination between beads encoded by a different combination of reporters must be prevented. A high level of cross-contamination would indeed make the decoding impossible. The reporter adhesion to the TG beads was confirmed by mixing a population of support beads encoded with B493-reporters with a population of support beads encoded with B530-reporters in DMF for 7 days. No cross-contamination was observed, and the B493-labelled beads were easily discriminated from the B530-labelled beads (data not shown). The resilience of the reporter adhesion to the TG beads during the peptide synthesis process was investigated by preparing individual populations of B493-coated beads and B530-coated beads and then mixing the portions together. These beads were exposed to six peptide synthesis cycles, and the level of cross-contamination was evaluated in a fluorescence microscope by counting the number of B493-reporters on the B530-labelled beads and vice versa. The level of cross-contamination during the 6 cycles of amino acid coupling and Fmoc cleavage was extremely low with only 1 to 3 reporters desorbed after TFA treatment (Table 2).

Table 2 Cross-contamination study

Step	No. of B493 reporters on B530-labelled beads	No. of B530 reporters on B493-labelled beads
t 0	/	/
C1	0	0
C2	0	0
C3	0	0
C4	0	0
C5	1	0
C6	1	0
TFA	1	3

---

II.2 Attachment of Atto-488 and AZ164 onto different amine-loaded TG beads

To provide control over the optimal terminal-fluorophore loading for this proteaseassay , the peptide loading was controlled by selective activation of the TG beads. The number of available amine sites for peptide conjugation was reduced by reacting the TG beads with a pre-activated mixture of Boc–Ala–OH and Fmoc–Ala–OH. An increase of the Boc amino acid concentration led to a decrease of the amine loading. Amine loading after the removal of the Fmoc group was determined through the Kaiser test and correlated with theoretical calculations. Two carboxylic acid functionalised fluorophores were reacted in excess onto different amine-loaded set of TG: a green-emitting dye Atto-488 (λ_ex 501/λ_em 523 nm) and a blue-emitting dye AZ164 (λ_ex 350/λ_em 423 nm). The flow cytometric analysis of the on-bead fluorescence after reaction demonstrated that the fluorescence of Atto-488 was high and stable as a function of the loading whereas the fluorescence of AZ164 gradually increased. These results indicate that Atto-488 tended to predominantly physisorb onto the TG. Comparatively, AZ164 was covalently attached and was thus selected as the terminal-fluorophore used in the rest of the study.

II.3 Flow cytometric detection of two distinct environments for bead-bound fluorophore

A polylysine peptide was synthesised onto TG in order to assess the trypsin affinity for substrates borne onto solid beads. Trypsin (23.8 kDa) is a well-characterised protease that cleaves specifically the C-terminal side (or carboxyl side) of lysine (Lys) and arginine (Arg) residues. During the initial flow cytometric analysis of peptides (thus exposing several cleavage sites to trypsin), 2 peaks appeared in the UV channel before and after incubation of trypsin. It appeared that the fluorescence intensity gradually increased in the same proportions for both peaks and that, depending on the amine loading, one peak appeared first before reversing and almost disappearing for the 100% loading. In addition, discrimination of both populations via sorting with the flow cytometer was not possible, showing that both peaks were simultaneously resulting from the same bead.

It was inferred from this evidence that the peptide-AZ164 was coupled within two distinct environments in the structural matrix of each TG bead: one throughout the internal matrix and the other being the exterior surface of the bead. In order to limit the potential existence of several diffusion coefficients and consequently a complicated set of kinetic parameters during the trypsin assays , the beads were bifunctionalised using a method described by Lam and co-workers.¹⁶ This permitted the selective capping of the inner amine sites while leaving the outer amine sites free for subsequent peptide synthesis. The success of this approach was evidenced in the subsequent disappearance of one peak in the UV channel on application of this procedure.

II.4 Optimisation of the peptide sequence

The peptide sequence was optimised during trypsin assays to determine the influence of the peptide length as well as the number of amino acids surrounding the cleavage sites.

The influence of altering the poly(ethylene glycol) (PEG, 27 units each) spacer length between the beads and the first amino acid was investigated initially (Table 3). A typical assay with 2 mg of TG support per experiment was deemed sufficient for the remainder of this study. We synthesised the ●-(PEG)_x-KKKKKK-γaba-AZ164 sequence with x = 0, 1 or 3. Samples were typically incubated with trypsin for 24 h at 37 °C. After centrifugation, supernatants were analysed by fluorimetry at λ_ex = 350 nm. Results indicate that the PEG length has a detrimental effect on proteolysis, although the peptide exposed 6 cleavage sites. A 5-fold decrease of the fluorescence was observed between the sequence containing no PEG and 3 PEG units.

Table 3 Influence of the peptide sequence on the trypsin activity. Potential cleavage sites are indicated by a vertical arrow

Label	Sequence C-ter → N-ter	Fluo intensity/supernatant (a.u.)^a
a Measurements were repeated 5 times for each sample at 423 nm.
1	●-↓K↓K↓K↓K↓K↓K-γaba -AZ164	923 ± 5
2	●-(PEG)1-↓K↓K↓K↓K↓K↓K-γaba -AZ164	485 ± 6
3	●-(PEG)3-↓K↓K↓K↓K↓K↓K-γaba -AZ164	179 ± 5
4	●-↓K-GG-γaba-AZ164	32 ± 8
5	●-G-↓K-GG-γaba-AZ164	8 ± 10
6	●-GGG-↓K-GG-γaba-AZ164	14 ± 5
7	●-↓K-GGG-γaba-AZ164	197 ± 4
8	●-GGGG-↓K-GGG-γaba-AZ164	183 ± 6
9	●-GGGGGGGG-↓K-GGG-γaba-AZ164	76 ± 3

---

The next strategy was to sythesise sequences exhibiting different numbers of glycine residues around the C- or N-terminal of a single lysine. Comparison of the fluorescence intensities observed for the peptides 4 and 7 demonstrate that proteolysis is favoured by increasing the number of glycine residues toward the N-terminal side of the lysine. However, variation of the number of residues toward the C-terminal side (peptides 4 → 6 and 7 → 9) showed little or no influence on trypsin activity.

II.5 Influence of amine loading on trypsin assays

A peptide with the sequence GAKAGA-AZ164 was attached to TG beads in sets representing amine loadings ranging from 0 to 100%. Each set was then incubated with trypsin for 24 h. The outcome of these assays was analysed by flow cytometry with the negative control being a protease free assay . The corresponding supernatants from these assays were analysed in parallel by fluorimetry as an indirect route to validate the efficiency of the proteolysis.

Unmodified beads (100% peptide loading) resulted in a 10% increase, rather than an expected decrease, in bead fluorescence intensity after 24 h digest (Fig. 2). Meanwhile, the increase in the supernatant fluorescence showed indirectly that trypsin had efficiently cleaved the peptide. It was deemed likely that self-quenching was the likely cause of this discrepancy and is discussed further in the Discussion section below.

Fig. 2 Impact of the bead surface amine loading on the trypsin activity. (a) Fluorescence intensity of the UV-dye released into the solution after on-bead proteolysis. (b) Mean fluorescence intensity of the beads before and after digestion, as measured by flow cytometry. On average 100 000 beads per condition were analysed.

A relationship between the increase of the non-digested peptide fluorescence with the decrease of peptide loading was observed which reached a plateau at 40%. The largest intensity shift post-trypsin digestion was observed with a 40% peptide loading producing a 74% decrease in fluorescence on the supports; this was confirmed by the indirect measurement of fluorescence in the supernatant. Beads loaded with 20% peptide exhibited low fluorescence initially and a small absolute difference after digestion.

II.6 Tryptic split-and-mix library

In light of the outcomes in Section II-5, an optimised split-and-mix library directed against trypsin was generated using 40% amine-loaded TG beads. To illustrate the concept of barcoding, 7 different types of fluorescent silica reporter were used to encode the library, detailed in Table 4. Each portion of TG beads was thus encoded through each cycle of the peptide synthesis. Each type of reporter contained one or a combination of the following fluorescent dyes: Bodipy®493/503 (B493), 530/550 (B530), 581/591 (B580) and 650/665 (B650). The cleavage rate was determined after digestion through the flow cytometric analysis of the on-bead fluorescence (Fig. 3a) and the fluorescence spectrophotometric analysis of the UV-dye released in solution. Selected populations were then separated viaflow cytometry sorting to decode the corresponding barcode by confocal microscopy (Fig. 3b).

Fig. 3 (a) Fluorescence microscopy image of the peptide-bound bead library after trypsin incubation. The weak fluorescence of several TG beads clearly indicates that proteolysis occurred for specific peptide sequences on those beads. (b) Confocal microscopy image of the TG beads which presented a reduced UV fluorescence as measured and sorted by the flow cytometer. The reporters corresponding to the peptide sequences A and E are distinguishable. The composite image results from the use of the 3 lasers (488/543/633 nm) and filters that permits the discrimination of the fluorescence emissions. Each channel is assigned a false colour for the clarity of pictures. Green circles = B493, yellow circles = B530, pink circles = B580, red circles = B650, blue circles = B493/650. (c) Analysis of the trypsin digestion of the split-and-mix peptide library. The loss of fluorescence (black columns) was quantified by flow cytometry for the whole library. Beads were then sorted and decoded in order to attribute the fluorescence changes. In parallel, each peptide was digested individually in order to analyse the release of UV dye in solution by fluorescence spectrophotometry (grey columns).

Table 4 Scheme for colloidal barcoding of a 5 member peptide library

Label	Sequence C-ter → N-ter								Note
A	●	Ala	Gly	↓Lys	Ala	Gly	γaba	AZ164	1 cleavage site
		B493	B530	B580	B650
B	●	Ala	Pro	↓Lys	Ala	Gly	γaba	AZ164	Blocked site
		B493	B493/530	B580	B650
C	●	Ala	Gly	Ala	Ala	Gly	γaba	AZ164	No cleavage site
		B493	B530	B493/580	B650
D	●	Ala	Pro	Ala	Ala	Gly	γaba	AZ164	No cleavage site
		B493	B493/530	B493/580	B650
E	●	Ala	Gly	↓Lys	↓Lys	Gly	γaba	AZ164	2 cleavage sites
		B493	B530	B580	B493/650

---

The rate of hydrolysis by trypsin is dependent on the sequence of amino acids adjacent to Lys. For example, the pyrrolidine ring in Pro is known to cause structural deformations in the peptide chain, causing the peptide to kink.¹⁷ The presence of Pro adjacent to Lys should consequently restrict the ability of trypsin to cleave the C-terminal side of Lys. Indeed, the presence of Pro adjacent to Lys on the C-terminal side of the peptide B in this study restricted the activity of trypsin as expected, yielding a decrease in fluorescence intensity of less than 7% (Fig. 3c). In the absence of Lys (peptide C and D), the fluorescence decrease was also low with values of 5% and 4%, respectively, indicating that cleavage was minimal. In contrast, peptides A and E exposing one or two cleavage sites exhibited strong fluorescence decreases of 69% and 74%, respectively.

III. Discussion

TentaGel polymer beads comprise a polystyrene backbone, 1–2% cross linked with divinyl benzene and grafted with polyethylene glycol. Generally speaking, the rate of any reaction onto such a support will depend on the diffusion rate of reagents or biomolecules that enter and leave this polymer matrix and their hydrophobicity or hydrophilicity. One study reported that trypsin (23.8 kDa) could penetrate the matrix of 90 μm TG beads¹⁸ while another study showed that chymotrypsin (22 kDa) could not reach and hydrolyse peptide substrates located within the interior of TG beads.¹⁹ In addition, Lowe and co-workers reported a low level of substrate hydrolysis by papain suggesting that only peptides present at the external region of TG beads were accessible for cleavage by proteases.²⁰

The outcome of the investigation of controlled amine loading on TG beads established that, after attachment of a fluorescent peptide, there were two separate populations of fluorescence peptides which exist in two distinct environments on each bead. The selective blocking of the amine functions located within the interior matrix of the beads prior to peptide synthesis removed one population. This strategy reduced the potential impact of multiple diffusion rates as reagents or biomolecules access the interior of the TG bead during either the synthesis of peptides or during proteolytic digestion.

The amino acid sequence optimisation experiments were directed at determination of the specificity of the peptide sequence toward proteolysis. It has been demonstrated that a library of penta- to octapeptides is a suitable length in order to map protease specificity on beads^21,22 while studies of peptides in solution can be effective with only tri-peptides (e.g. in the case of viral proteases²³). However, the protease affinity is generally sensitive toward substitutions in P′ positions, highlighting the importance of residues surrounding the cleavage site P1–P1′. In this study, bead-based assays were used to investigate the role of adjacent amino acids around the C- or N-terminal of the bead-bound main motif P1–P1′ in trypsin recognition.

Originally it was thought that synthesis of a PEG spacer onto the TG beads would enable better access by the protease to the substrate because the substrate would be further from the bead surface. However, the synthesis of an additional PEG spacer onto the TG beads was found to be detrimental toward trypsin activity. This is likely to be due to the characteristic loose coil structure of PEG units in water²⁴ increasing the susceptibility to hindrance of the trypsin access to the substrate sequence.

Further peptide constructs demonstrated that the number of amino acid residues between the beads and the cleavage site were of little importance compared to the impact of the length of the domain between the cleavage site and the terminal fluorophore. It was found that the longer the peptide chain between the cleavage site and the fluorophore, the greater the efficiency of the proteolysis. Such impact is likely due to the proximity of the terminal fluorophore. This fluorophore has a molecular weight of 326 g mol⁻¹ and is composed of three conjugated rings of which two are unsaturated. This bulky structure is sufficient to limit proteolysis through steric hindrance.

Interestingly, we noted that enzymatic cleavage of the fluorescently labelled peptides onto unmodified TG beads (i.e. 100% loading) led to an increase of fluorescence intensity. This increase is believed to be a result of the high density of peptide loading giving rise to self-quenching of the fluorophores. Self-quenching occurs as a result of the transfer of energy between neighbouring fluorophore molecules (one in the excited state and one in the ground state) without emission of radiation when the distance separating them is less than 10 nm. The Förster theory shows that the transfer efficiency varies as the sixth power of the distance between the two molecules.²⁵

A simple geometric model was used to estimate the surface-bound fluorophore distances from one another to examine the probability of self-quenching. The 10 μm beads used in this study have an average loading of 0.24 mmol g⁻¹. The length of the peptide chains is regarded to be negligible compared to the radius of the bead. This gives a volume of 524 μm³. Considering the capacity per bead of 1 pmol (as given by the supplier), there are approximately 7.8 × 10¹⁰ reactive sites per bead. This gives a volume of 6.7 nm³ per site and thus a distance of 2.3 nm between each site. These data are in agreement with the Förster radius provided by the supplier for which the efficiency of the donor–acceptor transfer is 50%. With a decrease in peptide loading from 100, 80, 60, 40 to 20%, this distance increases from 2.3, 2.5, 2.8, 3.2 to 4 nm. Our data show that a 40% loading allows for the largest decrease in fluorescence, both for the flow cytometry or fluorescence spectrophotometry analysis, after digestion of the beads. This indicates that a distance of approximately 3.2 nm between fluorophores is required to reduce self-quenching significantly.

Finally, the outcomes from each of the optimisation studies were applied in a proof-of-concept demonstration of a small encoded split-and-mix library directed against the trypsin protease. The flow cytometric analysis combined with the decoding of reporters by confocal microscopy demonstrated that the presence of the reporter particles did not alter the trypsin activity. The barcodes on the TG beads were found to be very stable throughout all the processes that were encountered. Specific peptide sequences presenting one or two cleavage sites were recognised while non-tryptic peptide sequences remained uncleaved.

The application of confocal microscopy to decoding offers major advantages over fluorescence microscopy by permitting the simultaneous acquisition of several (up to 8) reporters per group of beads The separate laser lines of the Zeiss LSM 510 Meta confocal microscope are activated simultaneously and up to 8 channels can be defined, each grouping any of the 32 detectors that resolve wavelengths in 20 nm intervals. The wavelength detection is completely under user control and can be set for any fluorophore, so that in only four passes (1 pass per encoded residue) a 32-channel spectrum can be collected. A frame size of 1024 pixel per line combined with a scan speed of 2.6 μs per pixel allowed to collect the composite image of 5 to 10 beads with a resolution sufficient to distinguish the reporters. The typical image acquisition time was about 5–10 seconds and is considerably shortened compared to the laborious manual change of emission filters of a fluorescence microscope. In addition, the confocal microscopy technique has become widespread amongst many research groups and is a powerful and available tool for decoding.

The maximum number of library members is dependent on the availability of a sufficient number of different reporter types. This is because a different reporter type is used to encode each individual amino acidaddition reaction in the combinatorial peptide synthesis. Using the encoding method presented here, 126 reporters are sufficient to encode an 18-mer peptide library containing every combination of 7 different amino acids (i.e., 10¹⁵peptides). It is possible to synthesise these 126 different reporter types from just seven fluorescent dyes (2⁷ − 1 = 127) (unpublished data).

It is anticipated that redundancy of 20–50 reporters per 10 μm bead is sufficient to enable accurate determination of peptide sequence by this method. A maximum of just over 900 reporters of 650 μm diameter could potentially be adsorbed as a monolayer on each 10 μm bead (with a surface area of approximately 300 μm²). Therefore, if 50 copies of each reporter type remain attached to each bead, theoretically there is sufficient surface area on the bead to enable a peptide sequence containing 18 amino acids to be encoded. This is an absolute extrema, however, with a more realistic maximum peptide sequence being 12–15 amino acids in length.

IV. Conclusions

In summary, trypsin protease profiling was demonstrated on a solid-phase split-and-mix peptide library which was optically-encoded via the colloidal barcoding approach. This unique optical encoding process enabled a library to be encoded ‘on-the-fly’, during a combinatorial peptide synthesis. This study demonstrated that TentaGel beads were a good support for the flow cytometric screening of encoded peptide libraries, provided several key variables were controlled. A high concentration of fluorophore on-bead caused self-quenching, however, careful control of the amine loading prevented this effect. A range of other parameters were investigated and optimised, including barcode stability, characteristics of the peptide-tagging fluorophore, as well as spacer group configuration. Through successful decoding of the colloidal barcodes, it was confirmed that specific peptide sequences presenting one or two cleavage sites were recognised by trypsin while peptide sequences not cleavable by trypsin remained intact. This encoding method is extremely powerful for applications where the consensus sequence of a protease is unknown, for example, for new viral proteases. For this application, random libraries containing all possible combinations of amino acids are required and can be encoded via the method presented herein. Indeed, from seven fluorescent dyes, it is possible to encode a hexamer peptide library containing essentially every combination of the 20 naturally-occurring amino acids, that is, a library containing 64 million peptides encoded with 120 different reporters (unpublished data). Moreover, this platform could potentially be extended to other biomolecular families such as oligonucleotides and sugars.

V. Experimental

V.1 Chemicals

All solvents used were of analytical grade. TentaGel N NH₂ (TG) beads used in the synthesis of peptide libraries were purchased from Rapp Polymere GmbH (Germany). 3-Aminopropyl-trimethoxysilane (APS) and tetraethoxysilane (TEOS) were purchased from Aldrich. The carboxylic acid functionalised dyes, AZ164 (λ_ex 350/λ_em 423 nm) and Atto-488 (λ_ex 501/λ_em 523 nm) which were used as terminal fluorophores, were purchased from Atto-tec GmbH (Germany). Succinimidyl dyes used for the reporter synthesis, namely Bodipy®493/503 (B493), Bodipy®530/550 (B530), Bodipy®581/591 (B580), Bodipy®650/665 (B650), were purchased from Molecular Probes (Invitrogen, USA). The polyelectrolytespoly(acrylic acid) (PAA, MW ≈ 250 K) and polydiallyldimethylammonium chloride (PDADMAC, MW ≈ 100–200 K) were purchased from Scientific Polymers and Aldrich, respectively. The coupling reagents HoBt (1-hydroxybenzotriazole), diisopropylethylamine (DIEA) and triisopropylsilane (TIPS) were purchased from Sigma. Fmoc-protected amino acids and spacers, γ-aminobutyric acid (γaba) and O-(benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HBTU) were purchased from Novabiochem. Piperidine and trifluoroacetic acid (TFA) were purchased from Auspep and SDS, respectively. Sequencing grade modified porcine trypsin was purchased from Promega Corporation.

V.2 Synthesis of fluorescent silica reporters

Reporter particles were synthesised through a modified Stöber procedure whereby a fluorescent dye–silane conjugate is incorporated into the matrix of a silica particle during its condensation, as described elsewhere.^15,26 Briefly, the isothiocyanate-functionalised dye is conjugated to APS and then mixed with TEOS during the process of hydrolysis and condensation. The resulting reporters are finally resuspended in water. This procedure was adopted to prepare reporters containing single or combinations of the following dyes: B493, B530, B580 and B650.

V.3 Polyelectrolyte coating

Polyelectrolyte coating of reporters was performed by suspending the reporters in a 1% aqueous solution of PDADMAC for 24 h and washing with water, and then resuspending the PDADMAC-coated reporters in a 1% aqueous solution of PAA followed by thorough washing. This procedure has been described previously.¹¹

IV.4 Blocking of the inner amine functions

Bifunctionalisation of the TG beads was performed following the method described by Lam et al.¹⁶ Briefly, TG beads were swollen in water for 24 h and then reacted with 9-fluorenylmethyloxycarbonyl-N-hydroxysuccinimide (FmocOsu) in DCM–diethyl ether and DIEA for 30 min. The free amines located in the interior of the beads were then capped with DMF–DIEA–acetic anhydride during 15 min before freeing the amines of the exterior with a 20% piperidine in DMF treatment.

V.5 Peptide synthesis

The peptide libraries were synthesised on TentaGel N NH₂ beads (10 μm, 0.24 mmol g⁻¹ substitution; Rapp Polymere GmbH, Germany) by using standard Fmoc chemistry and applying the split-and-mix method. Fmoc amino acids (10 mol equivalents) were activated by HoBt–HBTU–DIEA (10 mol equivalents) in DMF. The coupling reactions were carried out for 30 min with double coupling and monitored with ninhydrin. Specific reporters were introduced during the first coupling of each amino acid. After the Fmoc deprotection with 20% piperidine in DMF for 10 min, beads were combined and mixed in DMF before splitting into equal parts for the next cycle. After the last amino acid position, γ-aminobutyric acid was conjugated as a spacer group followed by labelling with a carboxy-functionalised terminal-fluorophore AZ164 which was spectrally distinct from the encoding dyes incorporated into the reporters. The side-chains of the peptides were cleaved with a cocktail of TFA–TIPS–DCM–H₂O 94 : 2 : 2 : 2% for 30 min followed by 4 washings with DMF.

V.6 Trypsin incubation

The 20 μg enzyme pellet was suspended in a resuspension buffer (50 mM acetic acid) and then in the incubation buffer (50 mM NH₄CO₃, pH 7.9). The peptide-coated beads were resuspended in the incubation buffer and then incubated with the trypsin solution at 37 °C (protease : peptide ratio 1 : 100 w/w) for 4 h. After digestion, the cleavage sites were determined using mass spectrometry analysis of the digest mixture with a Voyager DE-STR MALDI-TOF instrument for quality control purposes.

V.7 Fluorescence analysis

On-bead fluorescence was analysed using a BD LSRII flow cytometer (Becton Dickinson, USA). A UV laser emission line (355 nm, 20 mW) was used. The terminal-fluorophore AZ164 cleavage was detected in the 450 ± 25 channel. On average 1 000 000 beads per sample was analysed using BD FACSDiva software. Beads presenting a decrease of the UV fluorescence were separated through gating and sorting and then decoded by a Zeiss LSM 510 Meta confocal microscope. Imaging of the reporters and beads was done using a Zeiss Plan 100 × 1.4. Three different laser lines were used simultaneously during the acquisitions, namely the argon (488/514 nm), the green HeNe (543 nm) and the red HeNe (633 nm) lasers. Colloidal barcodes were identified using the standard set of filters supplied with the instrument. After trypsin digestion, fluorescence spectra of the supernatants were recorded in a Cary Eclipse fluorescence spectrophotometer (Varian). Fluorescence intensity was determined at the wavelength of 423 nm with an excitation of λ_ex = 350 nm.

Acknowledgements

This work was supported by the ARC (FF0455861; DP0555914); and the NHMRC (Industry Fellowship for BJB: 301267). The authors acknowledge Daniel Sangermani from the Centre of Advanced Light Microscopy (CALM) at UQ for his assistance with confocal microscopy.

Notes and references

1. O. H. Aina, R. Liu, J. L. Sutcliffe, J. Marik, C. X. Pan and K. S. Lam, Mol. Pharm., 2007, 4, 631–651.
2. H. M. Geysen, F. Schoenen, D. Wagner and R. Wagner, Nat. Rev. Drug Discovery, 2003, 2, 222–230.
3. O. Schilling and C. M. Overall, Nat. Biotechnol., 2008, 26, 685–694.
4. K. S. Lam, V. Krchnak and M. Lebl, Chem. Rev., 1997, 97, 411–448.
5. A. W. Czarnik, Curr. Opin. Chem. Biol., 1997, 1, 60–66.
6. B. J. Egner, S. Rana, H. Smith, N. Bouloc, J. G. Frey, W. S. Brocklesby and M. Bradley, Chem. Commun., 1997, 735–736.
7. L. Marcon, D. Kozak, B. J. Battersby, K. J. Chappell, D. P. Fairlie, P. Young and M. Trau, Anal. Biochem., 2008, 376, 151–153.
8. A. Nanthakumar, R. T. Pon, A. Mazumder, S. Yu and A. Watson, Bioconjugate Chem., 2000, 11, 282–288.
9. R. H. Scott and S. Balasubramanian, Bioorg. Med. Chem. Lett., 1997, 7, 1567–1572.
10. P. P. Surawski, B. J. Battersby, G. A. Lawrie, K. Ford, A. Rühmann, L. Marcon, D. Kozak and M. Trau, Mol. BioSyst., 2008, 4, 774–778.
11. B. J. Battersby, D. Bryant, W. Meutermans, D. Matthews, M. L. Smythe and M. Trau, J. Am. Chem. Soc., 2000, 122, 2138–2139.
12. B. J. Battersby, G. A. Lawrie and M. Trau, Drug Discovery Today, 2001, 6, 19–26.
13. B. J. Battersby and M. Trau, Trends Biotechnol., 2002, 20, 167–173.
14. L. Grøndahl, B. J. Battersby, D. Bryant and M. Trau, Langmuir, 2000, 16, 9709–9715.
15. G. Lawrie, L. Grondahl, B. Battersby, I. Keen, M. Lorentzen, P. Surawski and M. Trau, Langmuir, 2006, 22, 497–505.
16. R. Liu, J. Marik and K. S. Lam, J. Am. Chem. Soc., 2002, 124, 7678–7680.
17. B. Yan, P. C. Martin and J. Lee, J. Comb. Chem., 1999, 1, 78–81.
18. R. Quarrell, T. D. Claridge, G. W. Weaver and G. Lowe, Mol. Diversity, 1996, 1, 223–232.
19. J. Vagner, G. Barany, K. S. Lam, V. Krchnak, N. F. Sepetov, J. A. Ostrem, P. Strop and M. Lebl, Proc. Natl. Acad. Sci. U. S. A., 1996, 93, 8194–8199.
20. S. Leon, R. Quarrell and G. Lowe, Bioorg. Med. Chem. Lett., 1998, 8, 2997–3002.
21. S. L. Diamond, Curr. Opin. Chem. Biol., 2007, 11, 46–51.
22. M. Meldal, QSAR Comb. Sci., 2005, 24, 1141–1148.
23. N. H. Mueller, C. Yon, V. K. Ganesh and R. Padmanabhan, Int. J. Biochem. Cell Biol., 2007, 39, 606–614.
24. N. Vennemann, M. D. Lechner and R. C. Oberthür, Polymer, 1987, 28, 1738–1748.
25. T. Förster, Ann. Phys., 1948, 437, 55–75.
26. A. Van Blaaderen and A. Vrij, Langmuir, 1992, 8, 2921–2931.

---