Noninvasive optical detection of Granzyme B from natural killer cells using enzyme-activated fluorogenic probes

Despite many studies on the cytotoxic protease granzyme B, key aspects of its function remain unexplored due to the lack of selective probes for its activity. In this study, we fully mapped the substrate preferences of GrB using a set of unnatural amino acids, demonstrating previously unknown GrB substrate preferences that we then used to design novel substrate-based inhibitors and a GrB-activatable activity-based probe. We showed that our GrB probes react poorly with caspases, making them ideal for the in-depth analysis of GrB localization and function in cells. With our quenched fluorescence substrate, we determined GrB within the cytotoxic granules of human YT cells. When used as cytotoxic effectors, YT cells loaded with the GrB attack MDA-MB-231 target cells, and active GrB influences its target cell killing efficiency.


Introduction
In the last two decades, significant advances in the understanding of natural killer (NK) cells have been made 1 . These cytotoxic lymphocytes are key effectors of innate immunity and are involved in viral infection responses as well as controlling several types of tumors. The activation of these cells is initiated by major compatibility complex (MHC) class I protein loss in compromised cells 1 . NK cell activation changes the balance between the activating and inhibiting receptors on cell surfaces. Activated cells rapidly and quickly secrete cytokines: tumor necrosis factor α (TNFα) and interferon γ (IFNγ), leading to subsequent stimulation of the immune system. Reciprocal interactions with dendritic cells, macrophage T cells, and endothelial cells also enhance the immune system response. To prevent autoimmune damage, the previously reported that lengthening the peptide substrate improves the substrate hydrolysis rate for some proteases 14 . To define the optimal peptide length for which GrB is most active, six fluorescent substrates were designed based on literature data. The S1 binding pocket of GrB has an unusual preference for aspartic acid 10,15 ; therefore, we incorporated Asp at P1, and we elongated the peptide to up to six amino acids. The investigated peptides were as follows: Ac-AAIEPD-ACC, Ac-AIEPD-ACC, Ac-IEPD-ACC, Ac-EPD-ACC, Ac-PD-ACC and Ac-D-ACC. We measured the activity of GrB against each of the new substrates (at equal concentrations) and observed that neither the tripeptide, the dipeptide or the single amino acid-based substrates were hydrolyzed. The longer tetra-, penta-and hexapeptides were cleaved by GrB. The pentapeptide and hexapeptides were more efficiently cleaved by GrB than was the classic tetrapeptide substrate (Fig. 1A).
Because there was no significant difference in the efficiency of the cleavage of the penta-and hexapeptides, we selected the pentapeptides as the most appropriate chain length for activity-based probes and substrates.

Catalytic preferences in nonprime S1-S5 enzyme pockets of GrB
To further explore the substrate specificity of GrB and allow better optimization of GrB substrates, we analyzed the S1-S5 pocket preferences using both combinatorial chemistry methods and the screening of defined peptides possessing natural and diverse unnatural amino acids 16,17 . The GrB S1 pocket almost exclusively recognizes aspartic acid, and this feature is also seen in cysteine protease caspases 9,10,17 . To reduce substrate cross-reactivity, especially with caspases that exhibit the most similar substrate preferences to GrB, we tested whether any acidic and nonacidic modifications of natural amino acid residues are accommodated by the GrB S1 pocket. For this purpose, we designed and synthesized (using a solid-phase peptide synthesis method, see Supplementary  Data Fig. S1) a library of 95 defined peptides that share the same leading sequence and differ only by one amino acid residue at P1 (for the structures see Supplementary Data Fig. S2). ACC (7-aminocoumarin-4-acetic acid) was applied as a fluorescent leaving group, allowing the substrate hydrolysis rate to be determined based on the increase in fluorescence. The N-termini of the peptide substrates were acetylated to reduce the potential for hydrolysis by aminopeptidases commonly found in cells. In the remaining positions (P4-P2), defined natural amino acid residues were incorporated based on literature data (Ac-Ile-Ser-Pro-P1-ACC). We tested the activity of GrB against the new P1 library and determined that GrB almost exclusively requires aspartic acid in the S1 pocket ((L-Asp, 100%); it can accommodate the methylated derivative (L-Asp(O-Me), 30%), and less potently, tyrosine (L-Tyr) and its derivatives (<10%) can be tolerated (Fig. 1B and Fig. S3A). This confirms the literature data where, similar to caspases 10,17 , the GrB S1 pocket is essentially restricted to aspartic acid due to its interaction with the positively charged guanidine group of Arg 226. Additionally, the carboxyl group of Asp forms hydrogen bonds with three water molecules within the S1 subsite of GrB 18 . Not only the charge but also the shape of this pocket dictates amino acid binding since only Asp or uncharged Asp derivatives with minimal modifications (Asp(O-Me)) can occupy this pocket, despite the lack of a negative charge, while more sizable Asp derivatives (Asp(O-Chx) and Asp(O-Bzl)) are not recognized by GrB, revealing the limited capacity of the S1 subsite and the involvement of additional interactions in peptide binding. The shape of Asp allows it to fit perfectly in the S1 pocket, and this is supported by the interaction between the guanidine group and the COOH group (Fig.  S3).
Several approaches, including proteomics and PS-SCL (positional scanning substrate combinatorial libraries) 10,19 , have demonstrated that GrB hydrolyzes after P1 Asp. Although this feature distinguishes GrB from other granzymes, many GrB substrates are also recognized by caspases 10 , so conventional strategies for enzyme activity analysis cannot be used to selectively follow GrB activity in cells. We therefore sought a substrate sequence that distinguishes GrB activity from caspases. To this end, we next determined the extended catalytic preferences of GrB based on S4-S2 using a well-established HyCoSuL strategy (Hybrid Combinatorial Substrate Library) incorporating a wide range of different nonproteinogenic amino acids 16,17,20 (Fig. S3). We observed that bulky hydrophobic proline derivatives such as octahydroindolecarboxylic acid (L-Oic), 1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid (Tic) and O-benzyl-L-hydroxyproline (L-Hyp(Bzl)) are strongly preferred by GrB in P2, while substrates with hydroxyproline (L-Hyp) bearing unprotected hydroxyl groups are not hydrolyzed. Additionally, bulky hydrophobic 6-benzyloxy-L-norleucine (L-Nle(O-Bzl)), benzyl-L-histidine (L-His(Bzl)) and benzyl-L-serine (L-Ser(Bzl)) were tolerated in the S2 pocket, indicating that amino acids with benzyl groups can be accommodated within this pocket (Fig. 1B, Fig. S3) and revealing that the S2 pocket is very capacious. The crystal structure of GrB complexed with an inhibitor (Ac-IEPD-CO) with proline at P2 shows a cavity formed by the side chains of Phe 99, Tyr 94, Asp 102, His 57, and the main chain of Pro 96 is located beyond the pentameric pyrrolidine ring of the proline 18 . We speculate that this pocket is filled by the flexible cyclohexane group of proline derivative L-Oic or the benzyl groups of L-Nle(O-Bzl) or L-Ser(Bzl), enabling the amino acid residues to perfectly fill the S2 pocket. The HyCoSuL strategy also revealed that the S3 pocket of GrB has a broad substrate scope; however, it has strong preferences for glutamic acid (L-Glu, 100%), hydrophobic tyrosine bearing a benzyl group (L-Tyr(Bzl)), 72%) and mono-oxidized methionine L-Met(O) (Fig. 1B and Fig. S3). The strong preference for an acidic amino acid is due Asn 218 and Lys 192 in the S3 pocket of GrB, which interact with the side chain of L-Glu and stabilize the positively charged side chain of lysine 18 . Unlike the S2 pocket, the S3 pockets of all caspases 10 share GrB's preferences for glutamic acid, confirming the structural similarity of these enzymes 17 . Our results demonstrated that the S4 pocket also possesses a broad substrate scope. It can accommodate branched amino acids such as isoleucine (L-Ile, 72%) or valine (L-Val, 36%) or linear hydroxyl-L-norvaline (L-Hnv, 82%) but also a proline derivative (1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid, L-Tic, 100%), benzyloxymethyl-L-histidine (L-His(3-Bom), 82%) and some basic amino acids such as 2,4-diaminobutyrylic acid (L-Dab, 39%), citrulline (L-Cit, 10%) and 1,3diaminopropionic acid (L-Dap, 10%) ( Fig. 1B and Fig. S3). According to Rotonda et al., this GrB pocket is a "shallow hydrophobic depression formed by aromatic rings" (Tyr174 and Tyr215) and the side chain of Leu172 18 ; therefore, there is not enough space for phenylalanine within this pocket. Interestingly, our data from the HyCoSuL screening revealed that a bulkier amino acid (L-His(3-Bom)) was hydrolyzed by GrB (please see Supplementary Data Fig. S3).
To find the optimal amino acid for P5, we synthesized a combinatorial library of pentapeptides. For this purpose, based on the literature data and our results related to GrB specificity, we designed a library consisting of (1) defined amino acids at the P1 (Asp) and P3 (Glu), (2) equimolar mixtures (X) at P2 and P4 (a mixture of natural amino acids with L-Nle replacing L-Met and L-Cys), (3) and one of the 174 defined amino acids at P5 (Ac-P5-X-Glu-X-Asp-ACC) (for the structures, please see Supplementary Data Fig. S2). GrB was tested against the library, and we observed that it displays no S5 substrate specificity and is capable of hydrolyzing most substrates regardless of the residue at this position; however, the addition of an extra amino acid to the substrate (P5) clearly leads to a dramatic increase in substrate hydrolysis.

Design, synthesis and kinetic analysis of pentapeptide substrates for GrB
To validate the results of the HyCoSuL screening, we selected the most promising amino acid residues for the S4-S2 positions (P4: L-Tic, L-His(3-Bom), and L-Ile; P3: L-Glu, and L-Tyr(Bzl); and P2: L-His(Bzl), L-Oic, L-Tic, and L-Hyp(Bzl)) and synthesized eighteen different fluorogenic tetrapeptides using a previously described method 21 . Afterwards, we tested the activity of GrB on the new substrates and observed that peptides with Ile at P4 were exclusively hydrolyzed by GrB (TJ40-44),  while substrates with other selected amino acids, such as Tic (TJ2, TJ4, TJ6, and  TJ35-39) and L-His(3-Bom) (TJ3 and TJ30-34), were not recognized by GrB under these assay conditions. Additionally, we confirmed that sequences containing L-Nle(O-Bzl) or L-Oic at S2 were preferred by GrB, confirming the large size of S2 (Fig.  S4A).
With this in mind, we performed a detailed kinetic analysis of the most hydrolyzed and promising substrates, namely, TJ40, TJ41, TJ42, TJ43, and TJ44 (Ac-Ile-Glu-Oic-Asp-ACC, Ac-Ile-Glu-Hyp(Bzl)-Asp-ACC, Ac-Ile-Glu-Tic-Asp-ACC, Ac-Ile-Glu-His(Bzl)-Asp-ACC, and Ac-Ile-Glu-Nle(O-Bzl)-Asp-ACC, respectively) (Fig. S4), and we compared those results with those of reference substrate TJ7 (Ac-Ile-Glu-Pro-Asp-ACC). Substrates TJ40, TJ43 and TJ44 were cleaved more rapidly by GrB than was TJ7, and their kinetic constants were k cat /K M = 379.09 ± 25.15 M -1 s -1 , 267.89 ± 20.75 M -1 s -1 , and 685.51 ± 26.67 M -1 s -1 , respectively, and that of the reference is k cat /K M = 76.95 ± 7.72 M -1 s -1 . This was in agreement with our initial substrate screening (Fig. S4A). Since L-Oic and other proline derivatives are poorly or not recognized by caspases 10,17,22 , we decided to use the sequence of substrate TJ40 as the core for our future substrates and probes.   The optimization of the P4-P1 GrB peptide sequence resulted in a champion substrate that is well recognized by GrB and is less likely to show cross-reactivity with other granzymes and caspases. From the broad range of investigated amino acids, we selected the most promising structures (L-Lys(TFA), L-Nva, L-Ile, L-Phe(2-Cl) and L-hPhe), and we synthesized ten defined pentapeptide substrates, incorporating in the sequences previously selected for P4-P1 (L-Ile-L-Glu-L-Oic/L-Nle(O-Bzl)-L-Asp). After an initial screening of GrB activities against the new substrates (TJ46-TJ56) (Fig. 1C), we selected substrates TJ47, TJ49, TJ52, and TJ55 as being hydrolyzed with the highest efficiency, and we studied their kinetics in detail (Fig. 1D). Substrate TJ49 possesses the lowest K M value (32.33 ± 10.81 µM) and, at the same time, the highest k cat parameter (0.152 ± 0.017 s -1 ), with 40% higher activity (k cat /K M of 4960 ± 565 M -1 s -1 ) than that of the optimal tetrapeptide. Exchanging L-Nle(O-Bzl) at P2 (in TJ49) for L-Oic (affording TJ55) caused the enzyme activity to decrease by approximately a factor of four; however, we used this sequence in further analyses to minimize the cross-reactivity with majority of caspases 17 .

Granzyme B activity-based probe design, synthesis and evaluation
The most valuable tools for GrB investigations are activity-based probes, both inhibitor-and substrate-based. However, because GrB and caspases share similar substrate specificity 10 , the development of potent and selective GrB probes is challenging. To address this limitation, we first designed an inhibitor-like molecule (activity-based probes) that is selective for GrB with low cross-reactivity with caspases. We designed our probe to be built from three main functional segments. First, a reactive functional group that covalently binds to the active site of the enzyme. We selected diphenyl phosphonate, which mimics natural amino acids and potently reacts with serine proteases 23,24 , thereby reducing the possibility of the probe binding with caspases. Second, we attached a peptide sequence to the warhead based on the optimal GrB catalytic preferences to ensure the best fitting of the probe within the active site of GrB. For this purpose, we selected our champion substrate (TJ55) (Fig.  1D) since GrB showed its highest activity with this substrate, and Oic at P2 disfavors binding with caspases. Third, we attached an affinity (biotin) (TJ55.Bt) or a fluorescent tag (Cy5) (TJ55.5) at the N-terminus for a pull-down assay or ingel/membrane detection, respectively ( Fig. 2A). The probes were synthesized using a mix of classic solid-phase and solution-phase peptide synthesis methods. Briefly, the peptide sequence with Cy5 or biotin was attached to the warhead in solution following a procedure described elsewhere (see Supplementary Method Fig. S5).
Using the same method, we obtained a covalent inhibitor of GrB with the general structure of Ac-Nva-Ile-Glu-Oic-Asp P (OPh) 2 . As we noticed from the calculation of the inhibition kinetic constants, GrB strongly prefers biotin as a tag relative to the Cy5 derivative, and it binds with TJ55.Bt almost 70 times more rapidly than it does to TJ55.5 ( Fig. 2A). We speculate that this substantial difference in enzyme binding to biotinylated versus fluorescent probes is due to the presence of a biotin binding exosite in the GrB structure, and biotin binding is preferred due to the threedimensional enzyme structure and probe interactions. We tested the binding of our new probes to purified GrB and showed that both probes were capable of GrB labeling (TJ55.Bt Fig. 2B and TJ55.5 Fig. 2C). First, we optimized the enzyme-probe incubation time for detecting GrB. After 5 minutes of incubation, we noticed a strongly labeled species between 35 and 40 kDa, corresponding to the size of purified GrB (Fig. 2B, lane 2 and Fig. 2C, lane 2). As a control, to test whether the probe covalently binds in the GrB active site, we preinhibited GrB with a covalent inhibitor (Ac-Nva-Ile-Glu-Oic-Asp P (OPh) 2 ) prior to probe addition (Fig. 2B, lane 8 and Fig. 2C. lane 8). The inhibitor prevented probe binding; therefore, we concluded that TJ55.Bt and TJ55.5 bind to the active site of GrB. Additionally, TJ55.5 binds less potently than TJ55.Bt, which is consistent with their inhibition kinetic constants (Fig. 1A). In addition, to confirm that the 35 and 40 kDa species represent GrB, we applied an anti-GrB monoclonal antibody, and the signals from the probes and antibodies overlapped exactly ( Fig. 2B and 2C, yellow band), confirming that our probe labels GrB. Furthermore, to test whether the sample autofluoresces in our assay conditions, we ran one sample without an activity-based probe and only detected a signal from the antibody

Active GrB detection in cell lysates
Since our activity-based probes allow efficient GrB labeling, in the next step, we verified their specificity in a complex system of cell lysates. We selected the human NK cell line YT since it constitutively expresses and releases GrB 25 to observe if the probe binds with enzymes or cellular components other than the target enzyme. First, we optimized the concentration of the probe for efficient GrB labeling (Fig. 2C) and found that 5 nM is sufficient for GrB detection in complex systems. As indicated, even at a very high probe concentration (500 nM) and a long incubation time (60 minutes, Fig. 2C, lane 7), the probe binds almost exclusively with GrB; a signal from an additional band between 70-100 kDa was observed, but it does not have proteolytic activity 26 . Additionally, we observed that TJ55.Bt probe binds with GrB from YT cell lysates rapidly, and the complex was detectable after one minute of incubation (Fig.  2D, lane 2). Pretreatment of the cell lysates with a competitive inhibitor of GrB activity prevented probe from binding (Fig. 2D, lane 9). (The ~100 kDa species seen in Fig. 2D and E is a nonspecific product of the streptavidin reagent, as it appears in samples not containing TJ55.Bt (lane 7).) To verify whether the observed species corresponds to the target enzyme, we performed additional staining using an anti-GrB monoclonal antibody, and we noted that the signals from the probe and antibody overlapped, confirming probe binding with GrB. We also observed an additional species at 70 kDa when using the antibody. We speculate that this represents a complex between GrB and the cytosolic serpin PI-9 (Serpinb9), which forms postlysis and is detected by the monoclonal antibody of GrB.

standard deviation and represent at least 2 independent experiments. B and C) Optimization of the recombinant enzyme incubation time with TJ55.Bt (A) or TJ55.5 (B). GrB was incubated with TJ55.Bt or TJ55.5 for the indicated times (only the probe or only the enzyme was used in the controls). Afterwards, samples were analyzed by SDS-PAGE, followed by transfer to the membrane and streptavidin conjugate and antibody labeling. D and E)
YT cell lysates corresponding to 1 × 10 7 cells/mL were treated with TJ55.Bt for the indicated time (C) or samples were incubated with varying concentrations of GrB probe (from 5-500 nM) for 60 minutes (D). Afterwards, samples were analyzed by SDS-PAGE followed by transfer to the nitrocellulose membrane and immunoblotting using anti-GrB. As a control, lysates were pretreated with a competitive inhibitor prior to probe addition, or the probe and the lysates were run separately. The data reflect at least three separate biological replicates.
Since our probe is specific to GrB even in a complex system of YT cell lysates, we speculated that it can be applied for GrB detection in other cells, especially cancer cell lines. To test that hypothesis, we screened GrB activity in different cell lines, such as YT (as a control cell line), MDA-MB-231, Su-DHL-4, Jurkat-T, NK92, MG63, SEMK2, REM, and NALM-6, using classic SDS-PAGE, and we observed that GrB is present in its active form mainly in the NK-like cells YT and NK92. Interestingly, there was far less GrB detected in NK92 cells (below the detection level of the antibody), emphasizing the sensitivity of this active site probe. GrB was also detected in leukemia cell lines SEMK2, REM, and NALM but only in an inactive form, as it was strongly labeled with the antibody and not labeled with the activity-based probe (Fig. 3). We think that this may be due to the presence of inactive GrB within these cells or nonspecific antibody binding.

The design of a quenched fluorescence substrate probe for GrB
Biotinylated and fluorescent probes are valuable tools, but this type of inhibitor-like active site probe presents problems for enzyme monitoring in real time due to inhibition of the enzyme activity potentially leading to changes in its function. In addition, classic inhibitor-like probes are frequently equipped with unquenched fluorescence tags, which are always "on" despite the enzyme being active, and they emit a signal regardless of binding with the enzyme. This may cause false positive results since the cells, by pinocytosis or another mechanism, may take up the probe, and the signal will be detected regardless of binding with the target enzyme. Therefore, to avoid these limitations, we designed another type of molecule, a quenched fluorescent substrate probe, that emits fluorescence only after hydrolysis by the targeted enzyme. In addition, the probe will be substrate-based and therefore will not modify the biological function of the enzyme upon covalent binding.   To generate this probe, we first optimized its leading sequence. For this purpose, we used two previously optimized amino acid sequences for the P5-P1 positions based on TJ49 (Nva-Ile-Glu-Nle(O-Bzl)-Asp) and TJ55 (Nva-Ile-Glu-Oic-Asp), and we utilized variations of the P1'-P3' sequence (Phe-Gly-Arg or Gly-Gly-Gly). Additionally, at the P6 position, we attached a PEG(4) linker (or not) to increase the distance between a recognition sequence and the fluorophore. To identify the most active sequence for GrB, we utilized ACC and Lys(Dnp) as the fluorescence donor-acceptor pair. After synthesizing the twelve substrates (Table 1), we tested their hydrolysis rates by GrB, and we observed that in this type of GrB substrate, the predicted Oic (instead of the Nle(O-Bzl) group that was indicated in the HyCoSuL screening), is crucial at P2 for GrB substrate detection, and moreover, the S1'-S3' pockets are significant for GrB activity. We selected two of the most promising sequences, TJ65: H 2 N-ACC-Peg(4)-Nva-Ile-Glu-Oic-Asp-Phe-Gly-Arg-Lys(Dnp)-CO(NH 2 ) and TJ71: H 2 N-ACC-Nva-Ile-Glu-Oic-Asp-Phe-Gly-Arg-Lys(Dnp)-CO(NH 2 ), and we calculated their exact kinetic parameters. Table 1 shows the structures and kinetic constants for all substrates. Regardless of the presence of a linker at P6, the substrates possess high k cat /K M kinetic rates (TJ65 and TJ71). We also noticed a 33% reduction in K M if PEG(4) was incorporated at P6. Surprisingly, the k cat /K M for GrB on IQF substrates was dramatically higher than what was observed with shorter, classic substrates (TJ49 and TJ55). To test the selectivity of the most promising sequence (TJ71), we tested its activity with caspases since these enzymes share similar substrate specificity 10 , and we noticed robust activity with GrB but only minimal hydrolysis of TJ71 by caspase-6 and caspase-8 ( Fig. 4B and Table 2); other caspases did not lead to fluorescence increases. Thus, the peptide sequence TJ71 was selected as a scaffold for future synthesis.

The GrB quenched fluorescent substrate is activated by recombinant human GrB
The ACC moiety is not a suitable fluorophore for the live imaging of enzymes within cells since its fluorescence emission is close to the natural autofluorescence of cells, which may cause a false positive result. Additionally, ACC possesses a poor quantum yield compared to other fluorophores; therefore, to use our quenched fluorescent substrate for in-cell GrB investigation, we exchanged ACC with cyanine derivative Cy3, which is more stable and convenient for cell-based analysis. Classic, "always-on" probes contain a fluorophore that exhibits a signal regardless of binding with the targeted enzyme. To prevent false positive signals and minimize the background fluorescence from the fluorophore of classic "always-on" probes, we applied a nonfluorescent quencher in the fluorescent substrate sequence that efficiently silences the fluorescence signal of the unhydrolyzed substrate. After hydrolysis, due to the separation of the donor-acceptor pair, Cy3 fluorescence is activated by the enzyme, and the substrate is released. We applied Black Hole Quencher ® 2 (BHQ2) as the quencher since it is characterized by a high quenching yield of Cy3, and we speculated that it can be utilized in cell-based analyses. The lack of measurable activity toward caspases made this sequence (TJ71) an ideal candidate for the recognition sequence of the quenched fluorescent substrate. Importantly, the fluorophore was attached to the C-terminus of the substrate and the quencher was attached to the N-terminus since upon substrate hydrolysis by an enzyme, the product containing Cy3 will be amplified in place of hydrolysis (Fig. 4E).
The qTJ71 quenched fluorescent substrate was synthesized using a mixture of solidphase and solution-phase synthesis techniques (Supplementary data Fig. S7).
First, the peptide sequence was synthesized using classic SPPS on Rink amide followed by BHQ2 coupling at the N-terminus. Afterwards, the peptidic sequence with a quencher was cleaved from the resin, and Cy3-NHS ester was attached to the amine group of lysine with DIPEA/DMSO. We tested the activation of qTJ71 by GrB utilizing kinetic analysis and as indicated in Fig. 4C. The total RFU from the entire peptide hydrolysis increased proportionally to the substrate concentration, confirming our hypothesis that qTJ71 is activated by GrB and that the fluorescence quenched by BHQ2 is released upon hydrolysis. Since caspases have also been reported to recognize Asp at the P1 position, we evaluated the hydrolysis of our substrate with these enzymes. For this purpose, we monitored the fluorescence increases as a function of time upon qTJ71 hydrolysis with GrB and caspases (at equal concentrations). As indicated in Fig. 4D, qTJ71 is hydrolyzed exclusively by GrB. Therefore, in the next step, we calculated k cat /K M parameters for the hydrolysis of qTJ71 by GrB to be as high as 66500 M -1 s -1 . Next, we tested the cleavage site of qTJ71 by GrB, and observed that, as expected, the cleavage site was after the Asp residue ( Fig. 4E and Supplementary Data Fig.  S8).
qTJ71 was designed to be utilized in cell culture-based assays; however, first, we tested the utility of both quenched fluorescent substrate qTJ71 and unquenched TJ55.5 probe in a simple dot-blot assay with purified enzymes, and we confirmed if a signal specific to GrB could be detected and if any cross-reactivity of qTJ71 with caspases would be detected. Therefore, we added GrB or caspases-3, -4, -5, -6, -7, -8, -9, and -10 (separately) to the qTJ71 or TJ55.5, and afterwards, we performed dotblot analysis. We observed increases in the signal only in the samples containing GrB and qTJ71, and as a control, we utilized an untreated quenched fluorescent substrate in an assay buffer and did not observe fluorescence. In the same test using unquenched TJ55.5, we observed a strong false positive signal in every sample from the so-called "always-on" activity-based probe (TJ55.5), in which fluorescence is emitted regardless of binding with the enzyme. Therefore, we concluded that our qTJ71 is selective for GrB and can be utilized in cell culture assays for real-time imaging of GrB within cells (Fig. 4F). qTJ71 quenched fluorescent substrate cellular uptake qTJ71 possesses selectivity toward GrB and is not recognized by active caspases; therefore, in the next step, we used it to follow GrB activity in YT cells. First, as a pilot experiment, we tested qTJ71 cellular uptake over time (Supplementary data Fig.  S9). We performed live cell imaging of GrB in YT cells (in growth media) treated with 500 nM of qTJ71 and noticed that the quenched fluorescent substrate allows detection of GrB immediately after addition; however, the optimal labeling time for living cells was 15 minutes. Next, we addressed whether the fluorescence from qTJ71 cleavage is GrB-specific based on if it can be prevented by covalent inhibition of GrB. YT cells were pretreated with TJ55i and subsequently labeled with qTJ71. We observed a strong fluorescence increase in YT cells labeled with qTJ71 but no fluorescence in labeled cells pretreated with TJ55i, demonstrating that qTJ71 hydrolysis is prevented by the GrB covalent inhibitor (Fig. 5A). We concluded that this compound is selective to GrB and can be utilized for future GrB monitoring in living cells.

GrB activity monitoring in direct NK cell recognition of tumor targeted cells
GrB is considered to be involved in target cell killing, acting in concert with perforin, and by entering the target cell and, among other things, hydrolyzing caspase substrates, it can cause target cell death. Therefore, our next goal was to observe the accumulation of GrB activity in the breast cancer cell line MDA-MB-231 (target cells, T) after the addition of YT cells (effector cells, E). As indicated previously using the unquenched TJ55.5 probe (Fig. 3), active GrB is not usually present in MDA-MB-231 cells, while YT cells produce a substantial amount of this enzyme.
However, to test the utility of qTJ71 in GrB imaging in living cells, we performed two controls. First, GrB was inactivated using a covalent GrB inhibitor prior to the addition of the qTJ71 quenched fluorescent substrate. We observed no fluorescence, but when we used the qTJ71 alone, fluorescence was detected. This demonstrates that other components in the cell do not hydrolyze our quenched fluorescent substrate and it therefore can be used in cell-based assays. Second, to test the selectivity of the qTJ71, we treated both MDA-MB-231 and YT cell lines separately with the quenched fluorescent substrate and analyzed the samples by flow cytometry. We also analyzed naive cells and cells stained with anti-GrB followed by a fluorescently labeled secondary antibody. We observed robust fluorescence in YT cells treated with qTJ71, and in contrast, MDA-MB-231 cells showed no fluorescence. This was confirmed with antibody staining, which indicated the absence of GrB in MDA-MB-231 and its presence in YT (Fig. 5B). These experiments demonstrated that our quenched fluorescent substrate can detect active GrB within cells using simple and straightforward assays and classic methods such as FACS and microscopy ( Fig. 5A and B).
GrB is considered one of the most important enzymes involved in target cell killing. Therefore, with a specific GrB inhibitor, we tested whether the inhibition of GrB in YT cell lysates influences its target cell killing efficiency. To do so, we measured MDA-MB-231 cell viability in the presence of YT cell lysates or YT cell lysates pretreated with GrB inhibitor (TJ55i). As indicated in Fig. 5C, the inhibition of GrB led to a 10% increase in cell viability but did not prevent target cell killing. With this in mind, we speculate that GrB is important but is not the only YT cell component that can cause target cell death.

Chemicals
All chemicals were purchased from commercial suppliers and were used without further purification.

Kinetic assays
All kinetic experiments were performed using a spectrofluorometer (SpectraMax Gemini XPS, plate reader, Molecular Devices) and analyzed using SoftMax ® software (Molecular Devices), Microsoft Excel ® and GraphPad Prism ® . The granzyme B assay buffer contained 50 mM Tris-base, 100 mM NaCl, 25 mM CaCl 2 , 0.1% Tween, pH 7.4. The buffer was prepared at room temperature, and the enzyme kinetic studies were performed at 37°C. a) A HyCoSuL screening was utilized to define the substrate specificity of granzyme B. P4, P3, and P2 fluorogenic substrate sublibraries (33 µM) with P1-Asp, with the general structure Ac-P4-X-X-Asp-ACC, Ac-X-P3-X-Asp-ACC, and Ac-X-X-P2-Asp-ACC (where P4-2 is a defined amino acid, "X" is an equimolar mixture of natural amino acids with Nle replacing Met and Cys) were scanned with GrB (48 nM). The final volume of the reaction mixture was 30 µL. The substrate hydrolysis (fluorescence release in the presence of enzyme) was monitored in real time at λ ex = 355 nm and λ em = 460 nm for a minimum of 30 minutes, but for each substrate, only the linear part of the curve was used to calculate the reaction rate (RFU/s, relative fluorescent units per second). Each sublibrary was screened twice, and the average value was calculated. For each sublibrary, the amino acid with the highest RFU/s was set to 100%, and the other amino acids were adjusted accordingly. The results are presented as a bar diagram (P4-P2) (Supplementary Data Fig. S3). Additionally, the adjusted values for amino acids sharing similar chemical properties were added and then divided by the number of events, showing the GrB preferences for a particular molecular feature. Data are presented as a heatmap (Fig. 1A). In the same manner, the P5 library (Ac-P5-X-E-X-D-ACC, 166 µM final concentration) was screened using 12 nM of GrB and analyzed as above. Data are presented as a violin diagram of GrB amino acid preferences.

Synthesis of P1 fluorogenic substrate library
To synthesize peptides with the sequence Ac-amino acids-ACC, we used the same method described previously 16,21 with small modifications. A total of 100 mg of Rink AM resin (0.74 mmol/g) was added to a glass reaction vessel containing DCM, and the mixture was stirred gently once every 10 minutes for 30 minutes. Then the mixture was filtered and washed 3 times with DMF. The Fmoc protecting group was then removed by treatment with 20% piperidine (v/v) in DMF for 5 minutes, 5 minutes, and then 25 minutes. After each cycle, the resin was washed with DMF and then filtered, and after the final deprotection, it was washed carefully six times with DMF. Afterwards, 2.5 eq of Fmoc-ACC-OH was preactivated with 2.5 eq of HOBt and 2.5 eq of DICI in DMF for 5 minutes, and that mixture was poured into the resin. The reaction was gently stirred for 24 hours at room temperature. The resin was then washed three times with DMF, and the Fmoc-ACC-OH coupling reaction was repeated using 1.5 eq of the above reagents to improve the yield of the coupling reaction. After this, the reaction mixture was removed, and the resin was washed five times with DMF, affording ACC-resin (1). As previously described, to remove the Fmoc protecting group, 20% piperidine in DMF (v/v) was added to the reaction vessel, and three deprotection cycles were conducted (5 minutes, 5 minutes, and 25 minutes). The mixture was filtered and washed with DMF after each cycle to afford H 2 N-ACC-resin (2). Next, 2.5 eq of Fmoc-P1-OH was preactivated with 2.5 eq of HATU and 2.5 eq of 2,4,6-trimethylpiridine in DMF, and this mixture was added to the reaction vessel with 2. The reaction was carried out overnight at room temperature with gentle stirring. Then, the resin was washed with DMF three times, and the Fmoc-P1-OH coupling reaction was repeated using 1.5 eq of the reagents to increase the reaction yield and to obtain Fmoc-P1-ACC-resin (3). The Fmoc protecting group was removed as described above, and the peptide chain elongation was continued using 2.5 eq of Fmoc-aa-OH (aa-amino acid), 2.5 eq of HOBt, and 2.5 eq of DICI until the desired peptide length was obtained. Each time, the coupling reaction or deprotection efficiency was tested using the ninhydrin test. At the end of this process, the N-terminus was protected with an acetyl group using 5 eq of AcOH, 5 eq of HBTU, and 5 eq of DIPEA in DMF with gentle stirring for 60 minutes at room temperature to afford Ac-peptide-ACC-resin (4). Afterwards, the resin was washed five times with DMF, three times with DCM, and twice with MeOH and dried over P 2 O 5 overnight. Next, peptide cleavage from the dry resin was performed with a mixture of cold TFA/TIPS/H 2 O (v/v/v; 95:2.5:2.5) for an hour at room temperature with gentle stirring once every 10 minutes. Then, the peptide was precipitated from cold Et 2 O for an hour and centrifuged. The supernatant was removed while the pellet was recrystallized from an additional portion of cold Et 2 O and centrifuged. The product, as a light-yellow pellet, was dried overnight at room temperature, purified by HPLC, and then lyophilized. Product purity was confirmed by analytical HPLC and HRMS analysis. All substrates were stored as 20 mM solutions at -80°C until use.

Synthesis of internally quenched substrates
A total of 100 mg of Rink AM resin was added to a glass reaction vessel or 48-well reaction vessel containing DCM and stirred gently once every 10 minutes for 30 minutes, and then the mixture was filtered and washed 3 times with DMF. The Fmoc protecting group was then removed by treatment with 20% piperidine (v/v) in DMF for 5 minutes, 5 minutes, and 25 minutes. After each cycle, the piperidine was removed, and the resin was washed with DMF three times. Prior to transfer to the glass reaction vessel, 2.5 eq of Fmoc-Lys(Dnp)-OH was preactivated for 5 minutes by mixing with 2.5 eq of HOBt and 2.5 eq of DICI in DMF. Afterwards, the reaction was stirred for 12 hours at room temperature. The reaction mixture was filtered, and the resin was washed three times with DMF, and the Fmoc protecting group was removed from Fmoc-Lys(Dnp)-resin (5) as above using 20% piperidine in DMF (v/v) to afford H 2 N-Lys(Dnp)-resin (6). Next, 2.5 eq of Fmoc-AA-OH (P3' position) was preactivated with 2.5 eq HOBt and 2.5 eq DICI in DMF for 5 minutes and added to a reaction vessel with H 2 N-Lys(Dnp)-resin. The reaction was carried out for three hours with gentle stirring. Then, Fmoc-AA-Lys(Dnp)-resin (7) was filtered and washed with DMF three times. Afterwards, the Fmoc protecting group was removed as above, and the peptide chain elongation was continued until the peptide with the desired length was obtained (using 2.5 eq of HOBt and 2.5 eq of DICI as the coupling reagents). Next, 2.5 eq of Fmoc-ACC-OH was attached to H 2 N-peptide-Lys(Dnp)-resin (8) using 2.5 eq of HOBt and 2.5 eq of DICI, and the mixture was stirred for 24 hours at room temperature. Resin was then washed three times with DMF, and the reaction was repeated using 1.5 eq of the above reagents to improve the yield of the coupling of Fmoc-ACC-OH to (8). The final product, H 2 N-ACC-peptide-Lys(Dnp)-resin (9), was cleaved from Fmoc-ACC-peptide-Lys(Dnp)-resin protecting group and precipitated from Et 2 O as above. The product, a bright-yellow pellet, was dried overnight at room temperature, purified by HPLC, and lyophilized. Product purity was confirmed by analytical HPLC and HRMS analysis. All substrates were stored as 20 mM solutions at -80°C until use.

Fluorogenic and biotinylated activity-based probe and inhibitor synthesis (TJ55.5: Cy5-Gly-Nva-Ile-Glu-Oic-Asp P (OPh) 2 , TJ55.Bt: Biot-Nva-Ile-Glu-Oic-Asp P (OPh) 2 and TJ55i: Ac-Nva-Ile-Glu-Oic-Asp P (OPh) 2 )
In the first step, H 2 N-Nva-Ile-Glu-Oic-resin was synthesized using solid-phase synthesis on 2-chlorotityl chloride (CTC) resin. CTC resin (100 mg) was activated with 5 mL of anhydrous DCM and gently stirred once every 5 minutes for 30 minutes. Then, the mixture was filtered, and the filtrate was washed three times with DCM. Afterwards, 3 eq of Fmoc-Oic-OH (P2 position) was preactivated with 5 eq of DIPEA in anhydrous DCM and added to the glass reaction vessel with the CTC resin. The reaction was stirred gently in an argon atmosphere for 12 hours at room temperature. After that, the reaction mixture was filtered, and the resin was washed three times with DCM and two times with DMF to afford Fmoc-Oic-CTC (10). The Fmoc protecting group was then removed from 10 using 20% piperidine (v/v) in DMF for 5, 5 and 25 minutes, as described above, to afford H 2 N-Oic-CTC (11).
Peptide elongation with other amino acids (2.5 eq of P3: Fmoc-Glu(O-tBu)-OH, P4: Fmoc-Ile-OH, and P5: Fmoc-Nva-OH) was achieved with a series of coupling (with 2.5 eq of HOBt and 2.5 eq of DICI) and deprotection (20% PIP/DMF) reactions to obtain H 2 N-Nva-Ile-Glu(O-tBu)-Oic-CTC (11). Next, Boc-Gly-OH was introduced using 2.5 eq of HOBt and 2.5 eq of DICI (for TJ55.5) (12) or D-Biotin was introduced using 3 eq of DIPEA (for TJ55.Bt) (13). After the coupling of the last amino acid, the resin was washed five times with 2 mL portions of DMF, five times with 2 mL portions of DCM and three times with 2 mL portions of MeOH and dried over P  13 . In a round-bottom flask, 8.64 g of Meldrum's acid was mixed with 28 mL of triethyl orthoformate and stirred under reflux at 80°C. After three hours, the mixture was concentrated under reduced pressure to obtain a crude compound as a brown oil, which was stored at +4°C for 24 hours (it solidified at the temperature). The solid material (2.37 g) was vigorously stirred for one hour in 2 M HCl (34 mL). The suspension was then partitioned between diethyl ether and brine (3 × 100 mL). The organic phase was collected and dried over MgSO 4 , and the volatile components were evaporated under reduced pressure to give the product as a yellow solid. Next, to a 50 mL round-bottom flask were added formyl Meldrum's acid (1.3 g, 7.5 mmol), p-nitrobenzyl alcohol (1.2 g, 7.5 mmol), and toluene (15 mL). The reaction mixture was stirred under reflux in an oil bath for 30 minutes, and afterwards, the solvent was removed under reduced pressure. The obtained crude oil (1.5 g) was used in the next step without purification. In the next step, in a 50 mL round-bottom flask, 1.5 g of p-nitrobenzyl formylacetate, 0.9 g of benzyl carbamate, 2.1 mL of triphenyl phosphite, and 2.1 mL of glacial acetic acid were stirred for 1 hour under reflux at 80°C, and then the solvent was removed under reduced pressure. The obtained product (15) was purified over SiO 2 (2:3 ethyl acetate/hexane), and 1.2 g of pure compound was obtained. Afterwards, 15 was treated with 30% HBr in AcOH (10 mL) with stirring for 1 hour at room temperature. The solvent was then evaporated under reduced pressure, and the material was purified by reversedphase HPLC. Compound 16 was analyzed by HRMS and stored at -80°C until use.
In the next step, 2.5 eq (25 mg) of the aspartic acid phosphonate warhead (H 2 N-Asp P (OPh) 2, 16) was coupled with 12 or 13 using 2.5 eq of HATU and 2.5 eq of collidine in DMF. Each reaction was carried out for 2 hours at room temperature. Afterwards, each mixture was diluted in ethyl acetate and washed 2 times with 5% NaHCO 3 solution, 2 times with 5% citric acid solution, and 2 times with brine. The organic phase was collected and dried over MgSO 4 , and the volatile components were evaporated under reduced pressure. Next, deprotection of the aspartic acid carboxyl group from the warhead was performed by hydrogenolysis (H 2 , Pd/C, MeOH), and global deprotection was achieved in TFA/DCM (1:1, v/v; with 3% TIPS for 30 minutes) to afford H 2 N-Gly-Nva-Ile-Glu-Oic-Asp-Asp P (OPh) 2 (17) and Biot-Nva-Ile-Glu-Oic-Asp-Asp P (OPh) 2 (18). Crude 17 was purified by HPLC, analyzed using HRMS and lyophilized. In the next step, 1 eq of Cy5-NHS was dissolved in DMF, 5 eq of DIPEA was added, and the mixture was stirred occasionally. This mixture was then added to pure, dry 17. The reaction was carried out at room temperature for four hours with gentle stirring. The reaction progress was monitored by analytical HPLC. The final products, Cy5-Gly-Nva-Ile-Glu-Oic-Asp P (OPh) 2 (18) and Biot-Nva-Ile-Glu-Oic-Asp P (OPh) 2 (19), were purified on HPLC, analyzed using HRMS, lyophilized, dissolved to a concentration of 20 mM in DMSO and stored at -80°C until use.
To obtain the inhibitor, the amine group of 11 was acetylated with 5 eq of AcOH, 5 eq of DIPEA, and 5 eq of HBTU in DMF. The reaction was carried out at room temperature for one hour. The remaining steps were analogous to the synthesis of the activity-based probe.

qTJ71 synthesis (BHQ2-Nva-Ile-Glu-Oic-Asp-Phe-Gly-Arg-Lys-Cy3)
In the first step, a peptide derivative was obtained using solid-phase synthesis with Rink amide ® (AMR) resin in the same manner as above; however, a quencher was coupled to the N-terminus. For that coupling, 12.

Inhibition kinetic assay
The inhibitory constants of TJ55i, TJ55. 5 [8][9][10] for 20 minutes at 37°C. Afterwards, 10 µL of each solution was dotted onto a dry nitrocellulose membrane (Bio-Rad, 0.2 µm) and allowed to dry for 5 minutes. Then, the membrane was scanned with a Sapphire TM Biomolecular Imager with two lasers dedicated to Cy5 (658 nm) and Cy3 (520 nm), and the data were analyzed using Azure Biosystems software. The experiment was repeated 3 times.

Cell culturing
SU-DHL1, YT, NALM-6 and Jurkat-T cells were cultured in 75 cm 3 flasks in RPMI1640 medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin; SEMK2 and MG63 were cultured on culture plates in Eagle's minimum essential medium (EMEM) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin; and MDA-MB-231 cells were cultured on culture plates in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin under incubating conditions at 37°C, 90% relative humidity and 5% CO 2 . The primary concentration for optimal growth was 1 × 10 5 cells/mL, and the media were changed every other day. For all further experiments, cells with a low passage number (up to 30) were taken approximately 24 hours after the last seeding.

Western blot analysis of the cell lysates
To prepare the cell lysates, 1 × 10 7 cells/mL were lysed with 1 mL of cold lysis buffer containing 15 mM KCl, 5 mM MgCl 2 , 10 mM Tris-HCl, 0.5% Triton-X 100 (v/v). Afterwards, the cells were sonicated (2.0 kJ for 10 seconds) and immediately treated with an activity-based probe (TJ55.5 or TJ55.Bt) for the indicated time (0 minutes to 1 hour) at 37°C. The reaction was stopped by the addition of 30 µL of 3 × SDS/DTT to Afterwards, 30 µL of sample was loaded on 4-12% Bis-Tris Plus gel (Life Technologies); electrophoresis was performed at 200 V for 30 minutes, followed by transfer to a nitrocellulose membrane (0.2 µm, Bio-Rad, 1620112) for 60 minutes at 10 V. Then, the membrane was blocked with 2% BSA in TBS-T (Tris-buffered saline with 0.1% (v/v) Tween-20) for 60 minutes at room temperature, and when TJ55.Bt was used, the membranes were treated with fluorescent streptavidin conjugate (Steptavidin, Alexa Fluor TM 647 conjugate, cat. no S21374, Invitrogen) (1:10 000) for 1 hour at room temperature, followed by rabbit recombinant monoclonal granzyme B antibody (Abcam, ab208586) and incubated overnight at 4°C. Then, the membrane was incubated with the secondary antibody (Alexa Fluor ® 532 goat anti-rabbit IgG [H+L], Invitrogen A11009) for 30 minutes at room temperature. The fluorescence was scanned at wavelengths of 649 nM for AF647 or Cy5 and 554 nM for AF532 using a Sapphire TM Biomolecular Imager and Azure Biosystems software. The blots were then analyzed using Image Studio software.

Flow cytometry
All experiments were performed using CyFlow Cube6 (Sysmex) and analyzed in Sysmex software. MDA-MB-231 cells (1 × 10 5 cells/mL) in culturing media were incubated with (or without, as a control) 250 nM of qTJ71 for 1 hour. YT cells (1 × 10 5 cells/mL) were incubated in cell culture media with (or without, as a control) 250 nM of qTJ71 for 1 hour. Afterwards, all samples were spun down, fixed with 4% PFA for 20 minutes, and washed twice with DPBS. Then, the cells were treated with 10% BSA in DPBS for 30 minutes, and the samples were spun down and treated with anti-GrB primary antibody (rabbit recombinant monoclonal granzyme B antibody, Abcam, ab208586) and incubated overnight at 4°C. After washing, DPBS secondary antibody (Alexa Fluor ® 532 goat anti-rabbit IgG [H+L], Invitrogen A11009) was added, and the mixture was incubated for 1 hour at 37°C. Then, samples were washed twice with DPBS, resuspended in 200 µL of DPBS, and analyzed with CyView ® software (Sysmex) using blue and red lasers and FL2 (580 nm) and FL4 (675 nm) filters. The experiment was repeated three times.

Discussion
Granzyme B, a serine protease involved in programmed cell death through the cleavage of caspase substrates (BID and ICAM), is one of the key factors within NK cells and CTLs leading to cell death 10, 12 15 . GrB enters the target cell via perforin, and by hydrolyzing cellular proteins, it activates caspases, which are pivotal enzymes in cell death. The mechanism of GrB entrance into target cells using perforins remains unclear; however, a recent discovery led to the hypothesis that perforin may form a pore within the target cell that allows GrB to enter 27 . Within the cell, GrB is capable of cleaving different substrates, and since this enzyme can hydrolyze the same substrates as caspases, it can induce cell death through a caspase-independent cascade or the activation of caspases and is therefore involved in caspase-dependent cell death induction.
The classic artificial substrates of GrB described previously are tetrapeptide derivatives 13 . In our work, we analyzed the influence of peptide chain length with substrates containing from one to six amino acids. The incorporation of one or two additional amino acid residues increased GrB activity against the substrates, while shorter substrates (tri-or dipeptides) are not hydrolyzed by GrB. This analysis also indicated the importance of P4-Ile and P5-P6 amino acids in the interactions with the enzyme active site cavity and demonstrated that an additional amino acid reside at P5 of the substrate improves the interactions with the enzyme. We demonstrated that the P5 preference is broad but leans toward aliphatic residues such as Ile or Nva. Therefore, to construct our activity-based probe, we incorporated a pentapeptide as a leading sequence.
To precisely determine the GrB substrate specificity in the S1 pocket, we synthesized a set of substrates sharing the same P4-P2 sequence and differing at P1. We aimed to test whether Asp is the only amino acid that can be accommodated by the S1 pocket of GrB, and we observed that this enzyme almost exclusively hydrolyzes peptides after the Asp residue or its methylated derivative Asp(O-Me). This is due to the shape and chemical properties of the S1 pocket, which are controlled by the presence of a positively charged arginine 226 moiety oriented into the S1 cavity by the cis-proline conformation in the AA221-Pro224-Pro225-Arg226 motif, which is conserved between human and mouse GrB 18 . Therefore, GrB interacts with negatively charged Asp and its methylated derivative (Asp(O-Me)), but surprisingly, GrB does not cleave glutamic acid, which has one additional methylene group, and this is probably due to the shape and depth of the cavity, allowing for additional interactions with the Asp backbone.
On the other hand, the substrate specificity of the S2 pocket of GrB is similar to that of other serine proteases, and it recognizes proline and its derivatives. This specificity is characteristic of a majority of serine proteases, including the granzyme family 10 , neutrophil elastase 28 , cathepsin G 28 , proteinase 3 28 , chymotrypsin, and blood coagulation factor II. It is caused by Phe99 forming a wall in the active site of GrB, and as demonstrated previously, proline forms no specific interactions with the S2 pocket, but above the proline ring, an additional cavity was detected, and we speculate that this cavity can accommodate the proline extension because we observed that installing Oic at P2 dramatically improves substrate binding. Oic is a proline derivative modified with a cyclohexane ring that can adopt a threedimensional structure known as chair conformation, giving it flexibility. Interestingly, bulky hydrophobic Nle(O-Bzl) at P2 was also recognized, confirming the presence of an additional cavity around the S2 subsite. In our opinion, the S2 pocket is crucial for distinguishing between granzyme B and caspases since it is the only position in which these enzymes possess distinct specificity. With our peptide library screening of GrB, we confirmed that it recognizes glutamic acid at P3 and shares this feature with caspases 10,17,22,29 . The carboxyl group of glutamic acid interacts with Lys192 and Asn218 in the GrB S3 pocket, consistent with GrB specificity. Rotonda et al. circumvent this, we designed a quenched fluorescent substrate that fluoresces only after hydrolysis by GrB. To do that, we started be designing design an optimal and selective peptide sequence by a design, synthesis and kinetic evaluation of internally quenched substrates. Afterwards, we selected the most promising structure, which was characterized by a dramatic increase in the k cat /K M value, up to 100 000 M -1 s -1 relative to classic substrates (approximately 5000 M -1 s -1 ), and high specificity for GrB. We concluded that Oic at the P2 position is crucial for this specificity, while elongation with P1'-P3' improves the enzyme activity. Consequently, we exchanged a fluorophore/quencher pair (Lys(DNP)-ACC) for Cy3/BHQ2 since it is more suitable for cell-based experiments. This new sequence was specific to GrB and was barely recognized by caspase-6 and caspase-8.
The activity of GrB depends on the pH, and GrB stored in low pH granules is not enzymatically active. However, after treatment of YT cells with our qTJ71, we detected a strong fluorescence, which was confirmed by FACS analysis (Fig. 5A and  B). Signal was dismissed by the presence of GrB inhibitor, confirming that hydrolysis occurs indeed by GrB. These data suggest that YT cells constitutively release active GrB via granule exocytosis, consistent with a previous study 25 . We speculate that the treatment of YT with granule destabilizer LLOMe should markedly increase GrB binding 30 .
GrB delivery to target cells was previously reported to follow a mechanism involving perforin, also known as a cytoplasmic granule toxin, which allows appropriate GrB entrance to the cells, and once inside the cell, GrB triggers apoptosis. It was demonstrated that both perforin (PFN-SG) and GrB (GrB-SG) form complexes with anionic serglycine (Ser-Gly), and although PFN-SG is less membranolytic than free PFN, the complex displays a similar or even greater ability to deliver GrB-SG and free GrB to the target cell. Additionally, membrane pore formation is not mandatory for GrB delivery and subsequent apoptosis 31 . The other possible mechanism involves GrB arginine and lysine residues on the surface structure that can bind to heparan sulfate 32 or the negatively charged target cell membrane since the structure of GrB is highly basic (with a calculated pI of 10.4) 33 . We believe that our quenched fluorescent substrate will allow the monitoring of GrB delivery from NK cells to target cancer cell lines, which, as observed in MDA-MB-231 cells, do not have active GrB (Fig. 5B), and the observed fluorescence increase is associated with the substrate interaction with GrB. In summary, in our work, we designed and obtained a set of specific tools for GrB investigation: (1) substrates, (2) an inhibitor, (3) inhibitor-based activity-based probes, and (4) a quenched fluorescent substrate. The designed compounds are characterized by high selectivity for GrB and potent activity. With this, we were able to detect active GrB in lysates from different cell lines, and we noted the presence of this enzyme in NK-like cells: YT and NK92.