Evaluating the Effect of Dye-Dye Interactions of Xanthene-Based Fluorophores in the Fluorosequencing of Peptides

Abstract

A peptide sequencing scheme utilizing fluorescence microscopy and Edman degradation to determine the amino acid position in fluorophore-labeled peptides was recently reported, referred to as fluorosequencing. It was observed that multiple fluorophores covalently linked to a peptide scaffold resulted in a decrease in the anticipated fluorescence output and worsened the single-molecule fluorescence analysis. In this study, we report an improvement in the photophysical properties of fluorophore-labeled peptides by incorporating long and flexible (PEG)₁₀ linkers at the peptide attachment points. Long linkers to the fluorophores were installed using copper-catalyzed azide-alkyne cycloaddition conditions. The photophysical properties of these peptides were analyzed in solution and immobilized on a microscope slide at the single-molecule level under peptide fluorosequencing conditions. Solution-phase fluorescence analysis showed improvements in both quantum yield and fluorescence lifetime with the long linkers. While on the solid support, photometry measurements showed significant increases in fluorescence brightness and 20 to 60% improvements in the ability to determine the amino acid position with fluorosequencing. This spatial distancing strategy demonstrates improvements in the peptide sequencing platform and provides a general approach for improving the photophysical properties in fluorophore-labeled macromolecules.

Introduction

Fluorescence imaging has been integral to the development of several impactful technologies, especially in the chemical biology toolset, including activity-based probes^1-11 and methods for visualizing cellular structures.^12-21 Small-molecule fluorophores are uniquely enabling in analytical applications due to their size, ease of use, and modularity.^22,23 In combination with single-molecule fluorescence microscopy techniques, these probes have shed light on a number of challenging biological problems.^24-27 One of the most widely used families of fluorescent probes are based on the xanthene scaffold. Ever since their initial discovery in the late 19th century, the xanthene fluorophores, fluorescein (1) and rhodamine (2), have found abundant use (Figure 1A).²⁸ The parent compounds themselves are highly modifiable at the xanthene core and through substituents at the pendant aryl group or aniline. In the case of rhodamine, variants have also been prepared by substituting the central oxygen atom to give fluorophores such as the carbopyronine Atto 647N (3),^29-32 among others.^33-37 Xanthene-based dyes generally possess desirable spectral properties, including high extinction coefficients and quantum yields (QYs), fluorescence in the visible region, and overall chemical stability.³⁸ However, a key drawback to these fluorescent probes is their tendency for fluorescence self-quenching.

Figure 1.

Common xanthene fluorophores and overview of peptide fluorosequencing. (A) Structures of xanthene fluorophores and the two fluorophores used in the initial fluorosequencing report (2 and 3). (B) Bis-cysteine peptide with fluorophores at positions 3 and 4 being monitored with single-molecule fluorescence microscopy. A single fluorescent spot is representative of the di-labeled peptide, with corresponding loss of fluorescence for that spot occurring during the third and fourth Edman degradation cycles. (C) Initial report had short linkers between the fluorophores and the peptide backbone. The longer (PEG)₁₀ linker shown is discussed herein.

The self-quenching of 1 is a well-documented observation in the fluorescence microscopy analysis of fluorescein-labeled biomolecules^39-41 and has also been shown to occur with rhodamine fluorophores.^42-45 Increasing the number of fluorophores on a scaffold gives rise to a nonlinear increase in fluorescence intensity and a shortened fluorescent lifetime (FLT),^46,47 and in extreme cases, actually leads to inhibition of fluorescence.⁴⁸ The fluorescence self-quenching of fluorescein has been attributed to interactions through a fluorescence resonance-energy transfer (FRET) mechanism.^49,50 In contrast, rhodamines not only demonstrate dramatic fluorescence quenching primarily by dimerization through J- and H-aggregates depending upon the solvent but can also engage in FRET. Comparatively, carbopyronines such as 3 remain underexplored in this regard.^51-53 These two quenching mechanisms can operate in systems where multiple fluorescent probes of the same identity are covalently linked to scaffolds such as peptides⁵⁴ and thus can complicate analyses that rely upon reading the fluorescence output.

These challenges were highlighted in a recent report by our group that detailed a strategy for identifying the amino acid sequence of short peptides by analyzing the fluorescence of xanthene fluorophores covalently attached to amino acid side chains in a scheme known as “fluorosequencing” (Figure 1B).⁵⁵ Notably, fluorosequencing is just one of many emerging techniques in the field of single-molecule protein sequencing.⁵⁶ While the next-generation techniques in the area of genomic sequencing have revolutionized diagnostics, the analogous technique for proteins is still lacking. In the fluorosequencing approach, fluorophore-labeled peptides are covalently attached to a solid support via the C-terminus, where fluorescence is analyzed using total internal reflection fluorescence (TIRF) microscopy at single-molecule resolution.⁵⁷ During analysis, peptides are subjected to Edman degradation chemistry, which removes the N-terminal amino acid in each Edman cycle.⁵⁸ Thus, when a nonlabeled amino acid is cleaved, no change in the fluorescence output is observed, but when a fluorophore-labeled amino acid is cleaved, the fluorescence signal for that residue is lost, which indicates the identity of the amino acid residue.⁵⁹

While a broad range of fluorophores were evaluated for this sequencing strategy, given the harsh conditions of Edman degradation,⁶⁰ we found that only xanthene-based dyes tetramethyl rhodamine (2) and carbopyronine Atto 647N (3) possessed the desired survivability and photophysical properties required for fluorosequencing (Figure 1A).⁶¹ However, as expected, peptides labeled with multiple fluorophores of the same identity were more challenging to sequence. ⁶² Notably, the difficulty in sequencing was accentuated when the fluorophores were spatially close on the peptide backbone. As a result, we sought a strategy for minimizing the quenching interactions between appended fluorophores. The two avenues we considered to apply were the use of metals to enhance fluorescence, as demonstrated by a number of groups,^63-66 or to alter the length of the fluorophore linker, a surprisingly understudied area.⁶⁷ For the experimental setup, it was anticipated that metal additives could complicate analysis and potentially be degraded under Edman conditions. Thus, despite the considerable additional synthetic effort required to incorporate, we developed a methodology for attaching fluorophores 2 and 3 that possessed long, flexible linkers to cysteine residues using copper-catalyzed click chemistry. This report will detail our approach, synthetic efforts, and fluorescence analysis of peptides labeled with fluorophores using (PEG)₁₀ linkers. While unclear at the outset, it was hoped that this strategy would lead to a reduction in fluorescence quenching by minimizing the ability of fluorophores on a peptide scaffold to interact and thus improve the efficiency of peptide fluorosequencing. In our initial report, peptides were prepared by modifying the cysteine residues with iodoacetamide derivatives of TMR (4) and Atto 647N (5) (Figure 2A). This conjugation handle is selective for labeling cysteine residues,⁶⁸ and these derivatives are commercially available. The iodoacetamide handle is tethered to the fluorophores by linkers that are relatively short, as in 4 and 5, albeit the linker in 5 is longer than that in 4. Thus, when peptides, such as GACCAGRYD (7) are labeled with fluorophores at the third and fourth residues, the fluorophores are in quite close proximity and can engage in interactions such as aggregation that leads to fluorescence quenching.

Figure 2.

Design and synthesis of short- and long-linker peptides used to study photophysical and peptide fluorosequencing properties.

For the current study, parent peptide 7 was selected to test our fluorophore spacing postulate. Based on our prior work,⁵⁵ we found peptides with fluorophores at adjacent positions particularly challenging to sequence. As such, we chose to place the cysteine residues next to one another in an effort to maximize the interaction, anticipating that the longer linker would decrease interactions and thus lead to a greater improvement in amino acid identification. Labeling at the third and fourth positions meant that the fluorophores would only need to undergo four rounds of Edman degradation to sequence the peptides, lowering overall experimental run time.⁶⁹ Additionally, the nine-residue oligopeptide was large enough to enable simple synthesis by solid-phase peptide synthesis (SPPS) and chemical manipulation. As our initial report was intended to demonstrate the sequencing of model peptides that could be derived from real human proteins, the sequence was intentionally made less than 20 amino acids and terminated in an aspartic acid residue to mimic protein digestion by Endoproteinase GluC. In particular, peptides were synthesized and covalently bound to fluorophores through both short-linker units (as shown in 4 and 5) and longer (PEG)₁₀ linkers (as shown in 8 and 9) (Figure 2). While the short-linker fluorophores could be incorporated through standard cysteine alkylation, the long-linker fluorophores were installed onto the alkyne-modified peptide 11 using copper-catalyzed azide-alkyne cycloaddition (CuAAC) conditions. Peptide 11 was accessed from Fmoc-protected variant of peptide 7 by alkylation with click handle 12.⁷⁰ The specific (PEG)₁₀ spacers were chosen due to commercial availability of the linker as a hetero-functionalized polyethylene glycol unit.⁷¹ The free amine of the linker allowed for simple incorporation into the fluorophores through amide coupling, while the azide functional group allowed for the click installation to the peptides.⁷² To evaluate the photophysical differences between the short-linker fluorophore-labeled peptides and the long linker peptides, we performed both solution-phase and single-molecule fluorescence analyses.

Peptides for bulk fluorescence analysis, which included FLT and QY measurements, were acetylated at the N-terminus.⁷³ Short-linker peptides for this purpose were 13-16, and long-linker peptides were 19-22. Within these sets of peptides, both single fluorophore-labeled and di-labeled peptides were prepared. The single fluorophore peptides represent the baseline for the photophysical properties that could be compared to the di-labeled peptides, where it was expected that the fluorescence output would be modified by the introduction of the second fluorophore onto the peptide scaffold. The differences in FLT and QY between these sets of peptides then allowed quantification of linker length effects on ensemble photophysics.

Separately, another group of peptides were synthesized that possessed removable N-terminal protecting groups (Boc or Fmoc) for use in the single-molecule fluorosequencing routine: 17, 18, 23, and 24. For these peptides, the removable protecting group was necessary to allow for solid-support attachment followed by Edman degradation to carry out fluorosequencing.

Peptide synthesis

As shown in Figure 2, a total of 12 peptides were prepared for this study, each of which was used in the manner presented above. The short-linker peptides used in ensemble analysis, 13, 14, 15, and 16, were prepared from acetylated peptide 6 using reaction conditions (a) (Figure 2A). Fluorophores were installed by alkylation under anhydrous conditions with the corresponding fluorophore-iodoacetamide derivative (4 or 5). Notably, single fluorophore peptides 14 and 16 were prepared in a two-step procedure by first alkylating with 1 equivalent of fluorophore followed by the corresponding 2-iodoacetamide (giving inseparable mixtures of position 3 and 4 labeling).⁷⁴ Short-linker peptides for fluorosequencing analysis were prepared with a Boc protecting group on the N-terminus. Using reaction conditions (b) (Figure 2A), alkylation of peptide 7 with fluorophore 4 or 5 in aqueous buffer followed by subjection to Boc anhydride under anhydrous conditions gave peptides 17 and 18, respectively.

The long-linker peptides for this study were prepared from alkyne-functionalized peptides 10 and 11 (Figure 2B), which were accessed by alkylation of 6 or a variant of 6 (R = Fmoc, 28, see SI) with iodoacetamide click handle 12. The fluorophore-labeled peptides for ensemble analysis, 19, 20, 21, and 22, were prepared by subjecting peptide 10 to copper-catalyzed conditions using TBTA (Tris[(1-benzyl-1H-1,2,3- triazol-4-yl)methyl]amine) as a ligand with azide-fluorophores 8 or 9. As with the short-linker peptides, two peptides labeled with a single fluorophore and one nonfluorescent handle on the second cysteine residue were prepared, 20 and 22. The peptides for fluorosequencing were prepared from Fmoc-protected peptide 11,⁷⁵ which was incorporated from the SPPS. Again, subjecting 11 to the same copper-catalyzed click conditions with 8 or 9 gave access to peptides 23 and 24, respectively. Overall, long-linker peptide preparation generally proceeded by way of lower yield than the corresponding short-linker peptides, likely due to the challenge of CuAAC occurring between such large subunits (1 KDa each). However, due to the very low concentrations required for both solution phase and single-molecule analyses, even low yielding reactions on sub milligram quantities enabled this study.

Ensemble fluorescence analysis

With the peptides synthesized, we first characterized the ensemble fluorescent properties. Using short-linker peptides 13-16 and long-linker peptides 19-22, we sought to evaluate trends in the FLT (τ) and QY (Φ) between the two sets of peptides. This analysis would allow us to glean insights into the possible quenching pathways occurring in these peptides. Aggregation-induced fluorescence quenching occurs via a static mechanism, where the two or more fluorophores form a nonfluorescent species, decreasing fluorescence intensity and QY but minorly altering FLT.^46,47 In contrast, FRET-based quenching is a dynamic, through-space mechanism, where a decrease in FLT is generally observed.^50,51 The single fluorophore-labeled peptides represent the baseline of photophysical properties, with the di-labeled peptides representing an increase in the relative fluorophore concentration. The observed changes in QY and FLT upon an increase in fluorophore concentration can then give insight into the dye- dye interactions that are occurring.

The experimentally determined QY and FLTs⁷⁶ of peptides 13-16 and 19-22 are reported in Figure 3. To begin the analysis, starting from the mono-short-TMR peptide 14, the introduction of a second fluorophore, as in di-short-TMR peptide 13, resulted in a decrease in both the FLT and QY. The differences between the mono-fluorophore values and the di-fluorophore values for the short-linker TMR peptides are ΔFLT_ST and ΔQY_ST (ST = short-linker, TMR). To evaluate the improvement in photophysical properties seen by using the (PEG)₁₀ linkers, the values of ΔFLT_ST and ΔQY_ST are compared to the same values determined for the long-linker peptides of the same fluorophore (e.g., ΔFLT_LT and ΔQY_LT). In this way, the effect of long-linker tethers on the QY and FLT could be quantified.

Figure 3.

Solution-phase analysis of peptide QY (Φ) and FLT (τ, ns). Nomenclature for fluorophore-labeled peptides: ST = short-linker, TMR; LT = long-linker, TMR; SA = short-linker, Atto 647N; and LA = long-linker, Atto 647N.

The mono-short-TMR peptide 14 was found to have a FLT of 2.53 ns and a QY of 0.55. Upon the installation of a second TMR fluorophore as in di-short-TMR (13), a FLT of 2.19 ns and a QY 0.33 were observed. The introduction of the second fluorophore resulted in FLT decrease of 13% (ΔFLT_ST) and a QY decrease of 40% (ΔQY_ST). The same calculations were performed for the long-linker TMR peptides 19 and 20, to determine the effect of the linker change. For the mono-long-TMR peptide 20, a FLT of 2.56 ns and a QY of 0.51 were found. Upon installation of a second TMR fluorophore, as in di-long-TMR (19), a FLT of 2.31 ns and a QY of 0.39 were observed. In this case, introducing the second fluorophore led to a FLT decrease of 9.8% (ΔFLT_LT) and a QY decrease of 24% (ΔQY_LT).

For TMR-labeled peptides, a 13% decrease in FLT with the short-linker peptides (ΔFLT_ST) was found, compared to a 9.8% decrease in FLT with the long-linker peptides (ΔFLT_LT). The longer linkers resulted in an improvement in fluorescence lifetime by a factor of 1.3. More pronounced, however, is the improvement in the QY. The QY decrease for the short-linker TMR peptides of 40% (ΔQY_ST) compared to the 24% QY decrease for the long-linker peptides (ΔQY_LT) represents an improvement in the QY of the peptide-fluorophore conjugate by a factor of 1.7. Taken together, for TMR peptides, the improvements in both FLT and QY using the long-linker strategy validate our hypothesis that increasing the spacing and flexibility of peptide attachment points would improve the photophysical properties of the appended fluorophores.

The same analysis was performed for the Atto 647N-labeled peptides. The mono-short-Atto peptide 16 was found to have a FLT of 4.24 ns and a QY of 0.74. Comparatively, the di-short-Atto (15) had a FLT of 3.76 ns and a QY of 0.63. Taken together, for short-linker Atto 647N peptides, a decrease in FLT of 11% (ΔFLT_SA) and a QY decrease of 15% (ΔQY_SA) were observed when a second Atto 647N fluorophore was introduced to the peptide. The same calculations were then performed for the long-linker Atto 647N peptides. Mono-long-Atto peptide 22 had a FLT of 4.21 ns and a QY of 0.79 compared to di-long-Atto (21), which had a FLT of 4.01 ns and a QY of 0.75. Thus, for long-linker Atto 647N peptides, a FLT decrease of 4.8% (ΔFLT_LA) and a QY decrease of 5.1% (ΔQY_LA) were observed upon the introduction of the second fluorophore onto the peptide.

For short-linker Atto 647N-labeled peptides, the FLT decrease of 11% (ΔFLT_SA) compared to the 4.8% decrease for the long-linker peptides (ΔFLT_LA) shows an improvement in fluorescence lifetime by a factor of 2.3 using long (PEG)₁₀ linkers. Similarly, the decrease in QY for the short-linker peptides of 15% (ΔQY_SA) as compared to 5.1% in the long-linker peptides (ΔQY_LA) shows an improvement in the QY by a factor of 2.9. With these results, we reach a similar conclusion to the TMR-labeled peptides: the smaller decreases in fluorescence lifetime and QY reflect a decrease in fluorescence quenching and thus an improvement in photophysical properties.

In the TMR-labeled peptides, the relatively small improvements in FLT (1.3-fold) and QY (1.7-fold) found when using long-linkers can likely be rationalized by considering a fluorescence quenching mechanism that proceeds primarily by dimerization. It has been widely reported that rhodamine fluorophores engage in aggregation-induced quenching,^51,53 and this is the most likely interaction occurring in peptides 13 and 19. With the (PEG)₁₀ linker strategy, the long distance installed between the fluorophores can perturb the dimerization pathways, perhaps minimizing it or slowing the rate of dimerization due to the added degrees of freedom but not inhibiting it altogether. As such, improvements are observed in the photophysical properties, albeit modest ones. For the Atto 647N-labeled peptides, significantly larger improvements in both FLT (2.3-fold) and QY (2.9-fold) were observed when using the long-linker strategy. These larger changes observed in the (PEG)₁₀ linker molecules could be due to either disruption of a homo-FRET-type quenching mechanism or the same dimerization inhibition as with TMR. To the best of our knowledge, this is an understudied area of research with respect to carbopyronine fluorophores.

Single-molecule fluorosequencing analysis

As a companion to the ensemble fluorescence analysis, we also employed single-molecule fluorescence microscopy to evaluate the effect of linker length on peptide fluorosequencing efficiencies. The purpose of fluorosequencing is to identify the position of the fluorophore-labeled amino acid residues within a peptide sequence, giving a partial sequence. In modeling, our group has previously demonstrated that proteins in complex mixtures can be identified from the information contained in partial peptide sequences.⁷⁷ Fluorosequencing, like other single-molecule-based methods, enables decoupling of dynamic signal variance caused within a molecule from environmental variances observed between molecules.⁷⁸ Additionally, the fluorosequencing method enables the analysis of the same peptide before and after the removal of a fluorophore-labeled amino acid.

To this end, peptides 17, 18, 23, and 24 were subjected to single-molecule fluorosequencing, as described by us in detail previously.⁵⁵ This particular analysis provides a fluorophore count assignment for each peptide after each round of Edman degradation, which is used to determine the position of the labeled amino acids. For peptides of parent sequence GACCAGRYD, the expected fluorophore count assignments after five rounds of Edman degradation would be 2, 2, 1, 0, 0 with a drop from 2 to 1 after the third Edman cycle and 1 to 0 at the fourth Edman cycle, representing the loss of the fluorophore-labeled cysteines (see Figure 1B). Four unique fluorosequencing analyses were performed for each peptide 17, 18, 23, and 24.

To evaluate quenching in these peptides, the count assignments from each sequencing experiment were collected for every Edman degradation cycle and then sorted by their fluorophore count based on photometry analysis. The number of one- and two-count assignments (i.e., loss of only one fluorophore and expected loss of two fluorophores, respectively) were plotted as separate histograms and fit with a Gaussian distribution to determine the mean spot intensity for short- (17, 18) and long-linker peptides (23, 24).⁷⁹ Representative examples of these histograms and fits are shown in Figure 4A for the di-labeled short-linker Atto 647N peptide 18 and in Figure 4B for the di-labeled long-linker Atto 647N peptide 24, where the fits for one-count assignments are shown as a black solid line and the fits for two-count assignments are shown as a dashed teal line. The effects of self-quenching can be visualized by first plotting the expected distribution of the calculated unquenched two-count signal as a doubling of the mean of the single-count distribution, represented on Figure 4A,B as a dashed purple line. Second, we compare the two-count signal (teal line) to this doubling of the mean for a one-count peptide. We find that the dotted lines better overlay with the longer linkers; in other words, the experimental distribution for the peptides with two counts (Figure 4B) fits the calculated distribution better than that with the shorter-linker tethers (Figure 4A). Hence, the (PEG)₁₀-linked attachment indeed improves the fluorosequencing results.

Figure 4.

Intensity distribution and Gaussian fits for (A) short-linker Atto (18) and (B) long-linker Atto (24). (C) Percent quenching of peptides 17, 18, 23, and 24, showing a clear decrease in quenching with long linkers. (D) Percent change in correct positional assignments in fluorosequencing from short to long linkers for TMR (17 to 23) and Atto 647N (18 to 24).

To quantify this improvement, the fluorescence quenching was calculated as the percent difference between the calculated, unquenched, two-fluorophore signal mean and the mean of the experimentally collected two-fluorophore signal distribution. The resultant values are plotted in Figure 4C and show a significant decrease in quenching from the short-linker peptides to long-linker peptides with TMR [from 22 ± 2% (17) to 16 ± 3% (23)]. A similar effect, albeit smaller, occurs with the Atto 647N-labeled peptides, where the fluorescence quenching drops from 17 ± 1% (18) to 15 ± 3% (24).

Lastly, the sequencing assignments were used to evaluate the degree to which the length of the linkers affects the correct positional assignments in the fluorosequencing analysis. Figure 4D shows the percent increase in correct assignments for long-linker peptides versus short-linker peptides. Here, the TMR case showed a dramatic effect, with a 60% improvement in sequencing long-linker peptide 23 compared to short-linker peptide 17. Again, for peptides labeled with Atto 647N, the change was less pronounced but still quite significant; a 23% improvement when sequencing long-linker peptide 24 compared to short-linker peptide 18. These results demonstrate that the strategy of spatially distancing fluorophores on a peptide backbone led to improvements in the photophysical properties of the peptide-fluorophore conjugates and ease of application.

Conclusions

Single-molecule techniques for the sequencing of biopolymers offer significant advantages for genomic and proteomic studies. For peptides, our method, referred to as fluorosequencing, uses iterative Edman degradation steps to remove fluorophore-labeled amino acid side chains, as monitored by TIRF microscopy. To improve the positional assignments of amino acids when proximal to one another by minimizing the static and dynamic quenching, we explored a strategy of increasing the spacing of the fluorophores from the peptide backbone by using longer linkers than those used in our first report. Using two different fluorophores, TMR and Atto-647N, the increased linker length (embodied by a PEG₁₀) led to significant improvements in both QYs and FLTs (factors of 1.3 to 2.9). More important, however, the improvements found in the correct positional assignments for fluorosequencing were significant. When adjacent, the correct assignments for two cysteine residues using TMR and Atto fluorophores improved by 60% and 23%, respectively. Thus, increasing the length and flexibility of the linkers between amino acid side chains and their corresponding fluorophores decreases quenching, thereby increasing QYs and FLTs, and importantly the ability to correctly sequence labeled amino acids when positioned proximally in a peptide. This advance will become a general strategy to incorporate into our fluorosequencing routines and may inform more general fluorophore attachment strategies in various scaffolds. Similarly, the use of long and flexible linkers could benefit other quantitative single-molecule applications that involve counting applications.^80,81 Although this study focuses on adjacent labels, other quenching processes could similarly benefit from increased freedom of motion and the resulting physical separation.