Investigating Fluorescence Dyes in Fluorescence-Assisted Screenings

Abstract

Screening of bead-based peptide libraries against fluorescence-labeled target proteins was found significantly influenced by the dye characteristics. Commercially available red fluorescence dyes with net negative charges adversely showed strong interactions with library beads. The introduction of zwitterionic dyes significantly reduced the unwanted interactions, which sheds light upon using the right fluorescence probe for acquisition of reliable results in various fluorescence-assisted applications.

Combinatorial one-bead-one-compound (OBOC) peptide libraries introduced by Lam et al.¹ allow for identification of binding peptides for targets of interest such as proteins. Several methods have been used to obtain positive beads from the screening, such as use of colorimetric detections by biotinylated proteins² or monoclonal antibodies,³ use of fluorescence-labeled proteins⁴ and magnetic isolation by magnetic nanoparticle-conjugated antibody.⁵ Amongst them, conjugation of a fluorescence dye to a target protein is simple and straightforward associated with its easy chemical and biochemical platform in applying to high-throughput screenings. We have previously demonstrated an ultrahigh-throughput screening campaign by employing an automatic bead sorter (COPAS PLUS) that allows for rapid collection of positive beads sorted by the fluorescence intensity.^4,6,7

However, screening of a bead-based peptide library against a dye-conjugated target protein has been found significantly influenced by the fluorescence dye associated with nonspecific interactions such as hydrophobic and charge-charge interactions. As addressed by Camperi et al.,⁴ the true positive beads are difficult to be distinguished from the false positive beads interfered by unwanted interactions between the peptides and the fluorescence dye in the use of the COPAS system because the automatic sorting is only triggered by a certain level of fluorescence intensity. Although Camperi et al., could distinguish between the true and the false positive beads by applying two different fluorescence modes (halo appearance vs. bright homogeneous fluorescence), their method depends upon the size of target protein and the nature of solid support.⁴ Herein, we investigate the adverse effects of the charged commercially available red fluorescence dyes in the screening of short peptide libraries and thereby introduce zwitterionic physicochemical properties to minimize the background interactions from the dye.

Initially, commercially available red fluorescence dyes such as AlexaFluor 647 (AL) and DyLight 650 (DY) were used for labeling the target proteins to avoid strong autofluorescence⁸ from TentaGel beads. It is noteworthy that both dyes append multiple sulfate groups for increased solubility and stability in aqueous media. In order to examine background interactions, a portion of 5-mer library beads, comprising of 18 D-amino acids excluding methionine and cysteine (ca. 100 mg) was prepared and screened against the AL-labeled lysine and AL-labeled human carbonic anhydrase II (hCAII), respectively (Fig. 1).⁹ The results from the AL-labeled lysine revealed that positively charged amino acids, i.e. arginine and lysine, were dominant in every position of the positive peptides (Fig. 1d). This trend might be caused mainly by strong binding of negatively charged fluorophores to positively charged residues.¹⁰ From screening against AL-labeled hCAII, the histogram appeared similar to that from screening against the AL-labeled lysine (Fig. 1d and 1e), implying that the AL dye itself is significantly involved in binding with library beads. This trend was also observed with screenings against DY-labeled lysine and DY-labeled hCAII (see SI, Fig. S3). Labeled with AL, several other proteins such as bovine carbonic anhydrase II (bCAII), C-reactive protein (CRP), α-fetoprotein (AFP) and prostate specific protein (PSA) were also examined and resulted in high frequency of positively charged entities at each position of the positive peptides (see SI, Fig. S4). It is apparent that the undesired strong background bindings of the charged dyes jeopardize the identification of true positive peptides regardless of the protein types.

Fig. 1.

COPAS sorting images (X-axis: time of flight (TOF); Y-axis: fluorescence intensity, a.u.) from a) a 5-mer comprehensive peptide library itself, and from its screenings against b) AL-labeled lysine, and c) AL-labeled hCAII ([dye-labeled target] = 100 nM; Detection conditions: Excitation at 640 nm solid laser, Gain = 1, PMT = 450); Font histograms from positive peptides for d) AL-labeled lysine, and e) AL-labeled hCAII.

With an aim to see whether other features also account for the high background bindings, Texas Red dye (TR) with net-zero charge was investigated. Although positive peptides from screening against TR-labeled lysine and TR-labeled hCAII showed reduced abundance of positively charged amino acids, those positive peptides contained considerably increased population of aromatic amino acids such as tryptophan, phenylalanine and tyrosine, presumably due to the high aromaticity of the TR dye (see SI, Fig. S5). Several positive peptides resulted from the screening against the TR-labeled hCAII still showed strong cross-bindings to the TR-labeled lysine (see SI, Fig. S6). These observations clearly indicate that the structure of the fluorescence dye has a great influence on fluorescence-assisted screenings.

As an effort to reduce the undesired background bindings from the conventional red dyes, a library without arginine, and another library without both arginine and lysine in diversity element were investigated in the first place. However, in both cases the dominance of positively charged entities still remained with lysine and histidine most abundant, respectively (see SI, Fig. S7). Towards the reliable screenings, a new type of dye with low background bindings is highly demanded. One possible clue is to modify the Cy5 dye with zwitterionic moieties, well-known for the beneficial role of reducing the nonspecific interactions.^11,12 Although the modified dye (ZWCy5) with a zwitterionic moiety introduced on Cy5 showed reduced background bindings compared with the DY dye, a high level of background bindings still resided (Fig. 2a and 2b). When a pair of zwitterionic moieties was symmetrically deployed on the pentamethine cyanine core, the novel dye (ZW)^11c resulted in only negligible level of background bindings under the identical sorting conditions (Fig. 2c). These results are well in concordance with those from an in vivo application.¹¹ The ZW dye exhibits excellent optical properties with a high extinction coefficient (≈ 200,000 M⁻¹cm⁻¹) and a high quantum yield (≈ 20%), comparable to AL and DY dyes. The geometrically balanced net zero charge contributes to lowering the adverse background bindings presumably due to the charge shield effect as shown in their previous applications.^11,13

Fig. 2.

COPAS sorting images from bead-based screenings of a 5-mer comprehensive peptide library against a) DY-labeled lysine, b) ZWCy5-labeled lysine, and c) ZW-labeled lysine (Detection conditions: Excitation at 640 nm solid laser, Gain = 3, PMT = 700). The concentration of each dye-labeled target for screening was adjusted to exhibit the identical emission maximum.

When a comprehensive 5-mer peptide library was screened against ZW-labeled hCAII, the overall fluorescence level in the sorting by COPAS was significantly decreased compared with the screening against AL-labeled hCAII under the identical conditions (see SI, Fig. S8). Analysis of 120 positive beads sorted by the fluorescence intensity produced completely different results from the screenings against the ZW-labeled lysine as well as the AL-labeled hCAII (Fig. 3d, 3e and 1e). To obtain the peptides of reasonable binding affinity, a focused library was generated by recruiting only dominant amino acids at each position (see SI, Fig. S9). Through another round of screening, 11 peptides were selected, synthesized and evaluated for validations. Surface plasmon resonance (SPR) experiments revealed that the dissociation constants (K_D) of fplsk, lpypd and hrtsa are 5.8, 5.3 and 4.3 μM, respectively, which were determined by two-state model fitting found in Biacore T100 evaluation software (see SI, Fig. S10). The micromolar binding affinity is within the acceptable values for short peptides.¹⁴ In particular, fplsk showed excellent specificity for hCAII in the presence of PSA, AFP and CRP (see SI, Fig. S11). These results imply that the ZW dye can provide a great tool for reliable high-throughput screening of OBOC peptide libraries.

Fig. 3.

COPAS sorting images from a) a 5-mer comprehensive peptide library itself, and its screenings against b) ZW-labeled lysine, and c) ZW-labeled hCAII ([dye-labeled target] = 100 nM; Detection conditions: Excitation at 640 nm solid laser, Gain = 3, PMT = 750); Font histograms from positive peptides for d) ZW-labeled lysine, and e) ZW-labeled hCAII.

To expand the applicability of the ZW dye to other targets, we selected Ki-67 as a new target protein, a proliferation marker in early-stage breast cancer. The screening against AL-labeled Ki-67 generated a histogram with a great dominance of arginine and lysine at all positions, as similarly as the other targets (see SI, Fig. S12). Binding peptides with reasonable affinity could not be identified by validation of positive peptides. With the ZW-labeled Ki-67 used in the meantime, several binding peptides could be easily identified in conjunction with the negligible influence from the ZW dye. The positive 5-mer peptides showed several μM binding affinity by SPR. (see SI, Fig. S13).

To further elucidate the dye effects, a binding peptide (Ac-fplsk) for hCAII was elaborated on TentaGel beads. Each portion of the beads was incubated with hCAII or lysine, respectively labeled with 3 different dyes as similarly as in the screening experiments. After thorough washings, the ZW-labeled lysine was not detected on the beads under the given image acquisition conditions, confirming that the ZW dye involves only negligible level of background bindings (Fig. 4b). However, the beads incubated with the AL-labeled lysine displayed high fluorescence under the identical image acquisition conditions stemming from its strong bindings with the peptides (Fig. 4c). Whilst the AL- and the DY-labeled hCAII resulted in extremely high fluorescent beads, the ZW-labeled hCAII brightened only the outer sphere of the beads upon incubation under the identical conditions (Fig. 4d–4e).⁶ In fact, the similar halo fluorescence patterns were also observed with the former two derivatives as the image acquisition parameters were adjusted to reduce the fluorescence level. Therefore, it turned out extremely difficult to distinguish the true positive beads from the false when the conventional red fluorescence dyes are employed. We envision that it is not in general plausible to develop the binding peptides for the target proteins with AL used for labeling, although some previous approaches managed to identify the binding peptides for bCAII using AL for labeling.^9,15 However, those peptides were also found to interact strongly with the AL dye (see SI, Fig. S14). In fact, for screenings against the other AL-labeled proteins such as AFP, Ki67 and PSA, we could not identify any binding peptides with reasonable affinity. These results clearly illustrate that the ZW dye provides an excellent tool for reliable high-throughput screenings assisted by fluorescence.

Fig. 4.

a) Microscopic images of TentaGel beads appending Ac-fplsk; Confocal images of TentaGel beads modified with Ac-fplsk after incubation with b) ZW-labeled lysine, c) AL-labeled lysine, d) ZW-labeled hCAII, e) AL-labeled hCAII, and f) DY-labeled hCAII. [dye-labeled target] = 100 nM.

In summary, commonly used red fluorescence dyes negatively charged associated with the appending multiple sulfate groups adversely affect the screening of OBOC peptide libraries through strong background interactions with peptides on beads. A symmetric placement of zwitterionic moieties on both ends of the Cy5 dye molecule significantly reduced the background interactions, which could lead to successful identification of short peptides of reasonable binding affinity. Our results suggest that using the appropriate fluorescence dye should be of utmost importance to acquire reliable results in various fluorescence-assisted applications such as screening, assay and imaging.