Discovery of a Novel Class of Covalent Dual Inhibitors Targeting the Protein Kinases BMX and BTK

The nonreceptor tyrosine TEC kinases are key regulators of the immune system and play a crucial role in the pathogenesis of diverse hematological malignancies. In contrast to the substantial efforts in inhibitor development for Bruton’s tyrosine kinase (BTK), specific inhibitors of the other TEC kinases, including the bone marrow tyrosine kinase on chromosome X (BMX), remain sparse. Here we present a novel class of dual BMX/BTK inhibitors, which were designed from irreversible inhibitors of Janus kinase (JAK) 3 targeting a cysteine located within the solvent-exposed front region of the ATP binding pocket. Structure-guided design exploiting the differences in the gatekeeper residues enabled the achievement of high selectivity over JAK3 and certain other kinases harboring a sterically demanding residue at this position. The most active compounds inhibited BMX and BTK with apparent IC50 values in the single digit nanomolar range or below showing moderate selectivity within the TEC family and potent cellular target engagement. These compounds represent an important first step towards selective chemical probes for the protein kinase BMX.


Introduction
The TEC kinase family constitutes the second largest family of nonreceptor tyrosine kinases including the five members Bruton's tyrosine kinase (BTK), interleukin (IL)-2-inducible T-cell kinase (ITK or TSK/EMT), tyrosine kinase expressed in hepatocellular carcinoma (TEC), bone marrow tyrosine kinase on chromosome X (BMX or ETK) and tyrosine-protein kinase (TXK or RLK) [1]. All TEC family The presence of a nonconserved cysteine within the solvent-exposed front region at the Nterminus of the αD helix, often referred to as the αD-1 position [9], in all five TEC kinases (C496 in BMX; C481 in BTK) opens an opportunity for irreversible targeting. However, cysteines at the αD-1 As exemplified by BTK, irreversible inhibition represents an excellent strategy for targeting TEC kinases. Nevertheless, the development of inhibitors for other members of this family remains challenging due to the subtly different nature of their binding sites as well as the lack of understanding on these pockets. Design of specific ITK inhibitors, for example, is hindered by lower cysteine reactivity [25] and the larger gatekeeper residue. Only few reports describing the development of selective inhibitors of the less prominent TEC kinases, namely TXK, TEC and BMX, have been published so far. While no inhibitors addressing TXK or TEC as the primary target are known, two structural series of covalent BMX inhibitors targeting C496 have been disclosed. In 2013, Gray and colleagues identified the benzo[h] [1,6]naphthyridin-2(1H)-one derivative BMX-IN-1 (6, Figure 1), a dual covalent BMX/BTK inhibitor [26]. This compound has recently been complemented by a series of analogs from Bernardes and co-workers [27] (preprint on Chemrxiv), which feature increased potency but limited selectivity against JAK3, BLK and within the TEC family. JS25 (7a), the most potent BMX inhibitor from this set (IC50 = 3.5 nM, kinact/KI = 19 µM −1 s −1 ) showed increased cellular target engagement compared to 6 in a NanoBRET TM assay [28] while covalent binding has been demonstrated for the close analog JS24 (7b) by mass spectrometry and crystal structure. A structurally distinct series of covalent BMX inhibitors has been published in 2017 by Liu and colleagues [29]. These compounds exemplified by CHMFL-BMX-078 (8) showed high affinity for the inactive (nonphosphorylated) kinase while affinity for the active (phosphorylated) kinase was weak (apparent Kd > 10 µM). It is assumed that the latter compound binds BMX and related kinases in a type II binding mode addressing the hydrophobic back-pocket sometimes referred to as "deep pocket" [30], which is only accessible in the DFG-out conformation. However, there is still a need for expanding the scope of novel BMX/TEC family kinase inhibitors with high selectivity or complementary selectivity patterns to study biology. Here we present a novel and structurally unrelated series of covalent BMX/TEC-family kinase inhibitors featuring a distinct selectivity profile among the kinases with a αD-1 cysteine.

Discovery of a New Class of Covalent JAK3/TEC Family Kinase Inhibitors
Our previous study demonstrated the successful use of the tricyclic imidazo [5,4-d]pyrrolo [2,3b]pyridine scaffold for the design of selective covalent-reversible JAK3 probes FM-381(5a, see Figure  1) and FM-409 (5b) targeting JAK3 C909 via an α-cyano acrylamide warhead [21,22]. Based on this, we aimed to expand the inhibitor series and explore their potentials as irreversible inhibitors for JAK3 and other kinases that harbor a cysteine at the equivalent position. With no success in generating acrylamide-derived irreversible inhibitors based on the aforementioned tricyclic scaffold, we tested alternative hinge binding motifs and synthesized 1H-pyrrolo [2,3-b]pyridine-derived compound 9a (Figure 2, left). This derivative retained excellent JAK inhibitory potency (IC50 = 0.2 nM) and high selectivity against the other JAK family members and EGFR, demonstrating the hinge binder replacement as a potential strategy for optimizing the parent inhibitor towards other kinases (Table 1).    50 values were determined in a radiometric assay (Reaction Biology HotSpot™ [31]). Data were obtained as five-dose singlicate with 10-fold serial dilution starting at 10 µM.
We hypothesized that the 1H-pyrrolo[2,3-b]pyridine (7-azaindole) core in 9a might adopt an inverted binding orientation when compared to 5a,b or common 7H-pyrrolo [2,3-d]pyrimidine derived JAK inhibitors like 4 with two hydrogen bonds between the 7-azaindole and the backbone of L905 ( Figure 3). On this basis, we introduced a carboxamide group in the 5-position of the 7-azaindole scaffold to establish a third hydrogen bond towards the hinge region [32]. As expected, the resulting compound 10a (Figure 2, right) exhibited an increased potency for JAK3 (IC 50 < 0.1 nM) while maintaining a similar degree of selectivity against other JAKs. Interestingly, we observed that this compound demonstrated increased selectivity against EGFR compared to 9a. In line with covalent binding, nonreactive analogs 9b,c and 10b,c showed a strongly decreased potency with JAK3 IC 50 's in the micromolar range (not shown). The determined crystal structures of 9a and 10a ( Figure 4) in complex with JAK3 confirmed the predicted hinge interaction pattern and covalent modification of C909. As expected, the 7-azaindole core in both compounds was anchored to the hinge region via two hydrogen bonds with the L905 backbone NH and carbonyl oxygen atom, respectively. Compound 10a established an additional hydrogen bond between a carboxamide NH and the E903 backbone carbonyl oxygen atom. The second amide proton of the 5-carboxamide was involved in a presumably weak interaction with the methionine sulfur atom. The warhead amide was present in two flipped conformations (see Figure 4b) and involved in water-mediated hydrogen bonds, most notably with the backbone of C909. The latter interactions may assist in pre-orienting the Michael acceptor to facilitate cysteine addition. 1 IC50 values were determined in a radiometric assay (Reaction Biology HotSpot™ [31]). Data were obtained as five-dose singlicate with 10-fold serial dilution starting at 10 µM.
We hypothesized that the 1H-pyrrolo[2,3-b]pyridine (7-azaindole) core in 9a might adopt an inverted binding orientation when compared to 5a,b or common 7H-pyrrolo [2,3-d]pyrimidine derived JAK inhibitors like 4 with two hydrogen bonds between the 7-azaindole and the backbone of L905 ( Figure 3). On this basis, we introduced a carboxamide group in the 5-position of the 7-azaindole scaffold to establish a third hydrogen bond towards the hinge region [32]. As expected, the resulting compound 10a (Figure 2, right) exhibited an increased potency for JAK3 (IC50 < 0.1 nM) while maintaining a similar degree of selectivity against other JAKs. Interestingly, we observed that this compound demonstrated increased selectivity against EGFR compared to 9a. In line with covalent binding, nonreactive analogs 9b,c and 10b,c showed a strongly decreased potency with JAK3 IC50′s in the micromolar range (not shown). The determined crystal structures of 9a and 10a ( Figure 4) in complex with JAK3 confirmed the predicted hinge interaction pattern and covalent modification of C909. As expected, the 7-azaindole core in both compounds was anchored to the hinge region via two hydrogen bonds with the L905 backbone NH and carbonyl oxygen atom, respectively. Compound 10a established an additional hydrogen bond between a carboxamide NH and the E903 backbone carbonyl oxygen atom. The second amide proton of the 5-carboxamide was involved in a presumably weak interaction with the methionine sulfur atom. The warhead amide was present in two flipped conformations (see Figure 4b) and involved in water-mediated hydrogen bonds, most notably with the backbone of C909. The latter interactions may assist in pre-orienting the Michael acceptor to facilitate cysteine addition.   An ethylene glycol molecule (depicted in gray) fills the space between the ligand and the DFG motif forming a hydrogen bond to D967 but no direct hydrogen bonds with the ligand. Certain water-mediated hydrogen bonds as well as an alternative pose with a flipped amide conformation were omitted for clarity; (b) Compound 10a covalently bound to JAK3-C909. Two poses with a flipped warhead amide conformation were observed (depicted in salmon and blue). The carbonyl group of the 5-carboxamide moiety forms a hydrogen bond to an ethylene glycol molecule in the back pocket, which is further anchored by hydrogen bonding to E871 in the αC-helix and to the DFG motif. Certain water-mediated hydrogen bonds were omitted for clarity.
To examine whether our replacement strategy of the hinge binding motif for expanding the profile of the JAK3 covalent inhibitors to other kinases succeeded, we characterized the activity of inhibitors 9a and 10a against all kinases featuring the αD-1 cysteine. This screening revealed that the compounds strongly inhibited BMX and TXK with no or less significant binding to other kinases at 200 nM (see Table 2). Notably, the observed selectivity pattern did not seem to rely on differences in the gatekeeper moiety which is methionine in JAK3 and MKK7 and a threonine in the HER and TEC family kinases (except ITK).  An ethylene glycol molecule (depicted in gray) fills the space between the ligand and the DFG motif forming a hydrogen bond to D967 but no direct hydrogen bonds with the ligand. Certain water-mediated hydrogen bonds as well as an alternative pose with a flipped amide conformation were omitted for clarity; (b) Compound 10a covalently bound to JAK3-C909. Two poses with a flipped warhead amide conformation were observed (depicted in salmon and blue). The carbonyl group of the 5-carboxamide moiety forms a hydrogen bond to an ethylene glycol molecule in the back pocket, which is further anchored by hydrogen bonding to E871 in the αC-helix and to the DFG motif. Certain water-mediated hydrogen bonds were omitted for clarity.
To examine whether our replacement strategy of the hinge binding motif for expanding the profile of the JAK3 covalent inhibitors to other kinases succeeded, we characterized the activity of inhibitors 9a and 10a against all kinases featuring the αD-1 cysteine. This screening revealed that the compounds strongly inhibited BMX and TXK with no or less significant binding to other kinases at 200 nM (see Table 2). Notably, the observed selectivity pattern did not seem to rely on differences in the gatekeeper moiety which is methionine in JAK3 and MKK7 and a threonine in the HER and TEC family kinases (except ITK).

Evaluation of N-Substitution in the 5-Carboxamide Series
To exploit the steric nature of the gatekeeper moiety as an additional selectivity filter, we introduced N-substituents at the 5-carboxamide's nitrogen atom to preserve the triple hydrogen bonding interaction with the hinge region while forcing a steric clash with the bulky methionine gatekeeper in JAK3 or MKK7 and the phenylalanine in ITK. The first analog from this series with a relatively bulky N-cyclopentyl substituent (11a, Figure 5) did not show substantial activity at 200 nM on any of the kinases harboring an αD-1 cysteine. Nevertheless, we determined the IC 50 values for JAK3, BMX and TXK to compare acrylamide 11a with nonreactive propionamide 11b (Table 3). While 11a still showed residual activity on BMX, no JAK3 inhibitory activity was observed (IC 50 > 10 µM). In contrast, analog 11b did not inhibit any of the latter kinases up to 10 µM.  1 Residual activities were determined in a radiometric assay (Reaction Biology HotSpot™ [31]). Data were obtained from duplicate measurements and reported as the arithmetic mean of the individual values.

Evaluation of N-Substitution in the 5-Carboxamide Series
To exploit the steric nature of the gatekeeper moiety as an additional selectivity filter, we introduced N-substituents at the 5-carboxamide's nitrogen atom to preserve the triple hydrogen bonding interaction with the hinge region while forcing a steric clash with the bulky methionine gatekeeper in JAK3 or MKK7 and the phenylalanine in ITK. The first analog from this series with a relatively bulky N-cyclopentyl substituent (11a, Figure 5) did not show substantial activity at 200 nM on any of the kinases harboring an αD-1 cysteine. Nevertheless, we determined the IC50 values for JAK3, BMX and TXK to compare acrylamide 11a with nonreactive propionamide 11b (Table 3). While 11a still showed residual activity on BMX, no JAK3 inhibitory activity was observed (IC50 > 10 µM). In contrast, analog 11b did not inhibit any of the latter kinases up to 10 µM.  1 IC50 values were determined in a radiometric assay (Reaction Biology HotSpot™ [31]). Data were obtained as five-dose singlicate with 10-fold serial dilution starting at 1 µM (for 10a/11a) or 10 µM (11b).
To test whether smaller N-substituents could prevent the loss in BMX potency, we prepared the corresponding N-methyl and N-ethyl analogs 11c and 11d (Table 4). In addition, we synthesized compounds 11e-g to evaluate whether a methylene linker would be suitable to direct apolar cyclic moieties of variable size to the hydrophobic pocket behind the threonine gatekeeper often termed "hydrophobic region I" [34]. The binding of these derivatives was assessed using a thermal shift (∆Tm) assay, which unfortunately revealed a dramatic decrease of ∆Tm for all compounds in relation to the starting point 10a (Table 4). These results suggested that the applied modification strategy was not suitable to generate potent BMX inhibitors with high selectivity against JAK3.   [31]). Data were obtained as five-dose singlicate with 10-fold serial dilution starting at 1 µM (for 10a/11a) or 10 µM (11b).
To test whether smaller N-substituents could prevent the loss in BMX potency, we prepared the corresponding N-methyl and N-ethyl analogs 11c and 11d (Table 4). In addition, we synthesized compounds 11e-g to evaluate whether a methylene linker would be suitable to direct apolar cyclic moieties of variable size to the hydrophobic pocket behind the threonine gatekeeper often termed "hydrophobic region I" [34]. The binding of these derivatives was assessed using a thermal shift (∆T m ) assay, which unfortunately revealed a dramatic decrease of ∆T m for all compounds in relation to the starting point 10a (Table 4). These results suggested that the applied modification strategy was not suitable to generate potent BMX inhibitors with high selectivity against JAK3.

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6). 21

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6). 10

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6). 10

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5-carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6).
We thus prepared a set of analogs with an N-linked amide moiety (compounds 12a-i, see Figure 6 and Table 5) to validate this hypothesis. Satisfactorily, the compounds from this series showed increased BMX thermal shifts while JAK3 thermal shifts remained in a moderate range suggesting low inhibitory potency on the latter. To confirm these results, the IC 50 values of a subset of compounds were determined for BMX and JAK3. In agreement, all tested inhibitors exhibited high inhibitory activities for BMX with IC 50 values in a low nanomolar range. Moreover, all compounds displayed excellent selectivity against JAK3, which was most pronounced for the most active analog 12c (>9000-fold selectivity). We thus prepared a set of analogs with an N-linked amide moiety (compounds 12a-i, see Figure  6 and Table 5) to validate this hypothesis. Satisfactorily, the compounds from this series showed increased BMX thermal shifts while JAK3 thermal shifts remained in a moderate range suggesting low inhibitory potency on the latter. To confirm these results, the IC50 values of a subset of compounds were determined for BMX and JAK3. In agreement, all tested inhibitors exhibited high inhibitory activities for BMX with IC50 values in a low nanomolar range. Moreover, all compounds displayed excellent selectivity against JAK3, which was most pronounced for the most active analog 12c (>9000fold selectivity).
We selected potent cyclopentanoic amide 12a and compound 12c with a thiophene 2-carboxylic acid amide moiety for further profiling against the kinases that have an equivalent cysteine at the αD-1 position. At a concentration of 200 nM, 12a strongly inhibited BMX, BTK and TXK while TEC, BLK, HER4 and EGFR were less affected ( Table 6). As expected, no inhibition of JAK3, MKK7 and ITK was observed, which may be attributed to their bulkier gatekeeper incapable of forming the predicted hydrogen bond to the inverted amide moiety. In comparison, we observed that compound 12c had a slightly better potency than 12a, yet both shared a similar selectivity profile. Moreover, saturated analog 12b expectedly showed a much weaker BMX inhibition with an IC50 value of 582 nM while being devoid any significant activity on other kinases in this set except BTK and TXK (70% and 89% residual activity at 200 nM, respectively). For a better selectivity ranking, we determined IC50 values of 12a and 12c against all kinases that were significantly inhibited at 200 nM ( Table 6). The results confirmed that both compounds were highly potent inhibitors of BMX demonstrated by IC50 values of 2 nM and 1 nM, respectively. However, these inhibitors also inhibited BTK with similar We selected potent cyclopentanoic amide 12a and compound 12c with a thiophene 2-carboxylic acid amide moiety for further profiling against the kinases that have an equivalent cysteine at the αD-1 position. At a concentration of 200 nM, 12a strongly inhibited BMX, BTK and TXK while TEC, BLK, HER4 and EGFR were less affected ( Table 6). As expected, no inhibition of JAK3, MKK7 and ITK was observed, which may be attributed to their bulkier gatekeeper incapable of forming the predicted hydrogen bond to the inverted amide moiety. In comparison, we observed that compound 12c had a slightly better potency than 12a, yet both shared a similar selectivity profile. Moreover, saturated analog 12b expectedly showed a much weaker BMX inhibition with an IC 50 value of 582 nM while being devoid any significant activity on other kinases in this set except BTK and TXK (70% and 89% residual activity at 200 nM, respectively). For a better selectivity ranking, we determined IC 50 values of 12a and 12c against all kinases that were significantly inhibited at 200 nM ( Table 6). The results confirmed that both compounds were highly potent inhibitors of BMX demonstrated by IC 50 values of 2 nM and 1 nM, respectively. However, these inhibitors also inhibited BTK with similar potencies, suggesting that the moderate selectivity against BTK observed earlier for the parent compounds 9a and 10a was unfortunately compromised. Nevertheless, moderate selectivity of 12a and 12c over the other kinases in this test set, which included TXK, TEC, EGFR, HER2, HER4 and BLK, was evident and both had a slightly different selectivity profile. In comparison to 12c, which exhibited 8−30-fold lower potencies for the tested kinases (>100-fold only for HER2), compound 12a showed a better selectivity against HER family kinases (72-, 280-and 48-fold against EGFR, HER2 and HER4, respectively), TEC (14-fold) and BLK (24-fold), while selectivity against TXK (6-fold) remained moderate.

Design and Structure-Activity Exploration of the 5-Acylamino Series
In search for alternative approaches exploiting the gatekeeper residue as a selectivity filter, molecular modeling studies were performed. Docking simulations suggested that inverting the 5carboxamide substituent would enable a favorable NH···O hydrogen bond to the hydroxy group of the threonine gatekeeper in BMX. While the third hydrogen bond to the hinge region would not be compatible with this arrangement, the interaction with the gatekeeper was predicted to direct moieties attached at the carbonyl group towards the hydrophobic region I behind the gatekeeper moiety (see Figure 6). BLK, was evident and both had a slightly different selectivity profile. In comparison to 12c, which exhibited 8−30-fold lower potencies for the tested kinases (>100-fold only for HER2), compound 12a showed a better selectivity against HER family kinases (72-, 280-and 48-fold against EGFR, HER2 and HER4, respectively), TEC (14-fold) and BLK (24-fold), while selectivity against TXK (6-fold) remained moderate. 1 Mean of at least three independent measurements ± standard deviation. 2 IC50 values were determined in a radiometric assay (Reaction Biology HotSpot™ [31]). Data were obtained as five-dose singlicate with 10-fold serial dilution starting at 10 µM (JAK3) or 1 µM (12d,g,i vs. BMX). For 12a,c vs. BMX, 5-fold serial dilution starting from 0.5 µM was used.
[ATP] to 10 µM. analog 10a showed decreased potency, which might be a result of lower cell penetration due to the primary amide functionality. As expected, and in line with the data from enzymatic assays, N-alkylated analog 11a and non-reactive control compounds 9b, 10b, 11b and 12b showed weak or no affinity for both kinases in this cellular model.

Validation of Covalent Modification and Binding Mode Prediction
Covalent adduction between BMX and the inhibitors was investigated by mass spectrometry (see Supplementary Table S1). When incubating highly potent acrylamide-derived inhibitors 9a, 10a and 12a,c,g,i at 1.5-fold molar excess with BMX at 4 °C, complete labeling was observed readily after 30 min indicating a high efficiency of covalent bond formation. In contrast, analog 11a, which was less potent in the activity assay required prolonged exposure (240 min) to approach full target modification. As expected, no modification was detectable for propionamides 9b, 10b, 11b and 12b, which were employed as negative controls. To examine general reactivity towards thiols, compound 12c was tested for its stability against glutathione (GSH). In the presence of 5 mM GSH at pH 7.4., 12c showed a half-life of approx. 10 h which compares favorably to Afatinib (<1 h) tested under the same conditions (see Supplementary Figure S2). It can therefore be concluded that covalent target modification is facilitated by the preceding reversible binding event and not a result of nonspecific thiol addition. To gain a detailed insight in binding interactions, we aimed to crystallize the complex of BMX and the inhibitors, yet unsuccessfully. Thus, covalent docking analyses were instead performed. Suggested covalent binding modes of the two representative compounds 12a and 12c are

Validation of Covalent Modification and Binding Mode Prediction
Covalent adduction between BMX and the inhibitors was investigated by mass spectrometry (see Supplementary Table S1). When incubating highly potent acrylamide-derived inhibitors 9a, 10a and 12a,c,g,i at 1.5-fold molar excess with BMX at 4 • C, complete labeling was observed readily after 30 min indicating a high efficiency of covalent bond formation. In contrast, analog 11a, which was less potent in the activity assay required prolonged exposure (240 min) to approach full target modification. As expected, no modification was detectable for propionamides 9b, 10b, 11b and 12b, which were employed as negative controls. To examine general reactivity towards thiols, compound 12c was tested for its stability against glutathione (GSH). In the presence of 5 mM GSH at pH 7.4, 12c showed a half-life of approx. 10 h which compares favorably to Afatinib (<1 h) tested under the same conditions (see Supplementary Figure S2). It can therefore be concluded that covalent target modification is facilitated by the preceding reversible binding event and not a result of nonspecific thiol addition. To gain a detailed insight in binding interactions, we aimed to crystallize the complex of BMX and the inhibitors, yet unsuccessfully. Thus, covalent docking analyses were instead performed. Suggested covalent binding modes of the two representative compounds 12a and 12c are depicted in Figure 8. As expected, the docking poses show the dual hinge interaction pattern along with an additional hydrogen bond between the amide NH and the hydroxy group of T489 orienting the nonpolar substituent at the 5-acylamino group towards the hydrophobic region I. Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 13 of 38 depicted in Figure 8. As expected, the docking poses show the dual hinge interaction pattern along with an additional hydrogen bond between the amide NH and the hydroxy group of T489 orienting the nonpolar substituent at the 5-acylamino group towards the hydrophobic region I.
(a) (b) Figure 8. Covalent docking models of compounds 12a (panel (a)) and 12c (panel (b)). Poses were generated using the BMX crystal structure with the PDB-code 3SXR as the template.
Although we did not experimentally prove selective modification of C496 by tryptic digestion/MS experiments, crystal structure or mutation of the target cysteine, the efficiency and stoichiometry of covalent modification in conjunction with modeling data and the binding modes of analogs 9a and 10a in JAK3 strongly support the specific labeling of this residue, which is the only accessible cysteine in or proximal to the BMX ATP binding site. In addition, much lower activity of non-reactive analog 12b compared to acrylamide 12a on all enzymes tested complies with a covalent mode of action on other kinases with an equivalent cysteine placement.

Compound Synthesis
The key transformation in the synthetic route to the inhibitors disclosed herein was the late stage introduction of the entire N-aryl acrylamide warhead/linker fragment via a Suzuki coupling. This reaction could be performed under very mild conditions preventing possible side reactions involving the electrophilic Michael acceptor system. The synthesis of the key boronic acid ester 14 was realized as a high yielding two step sequence starting from 3-bromoaniline. The latter was smoothly converted to the pinacol boronate 13 under Miyaura conditions [35] followed by the acylation with acryloyl chloride to deliver the bench stable building block 14 in a good overall yield of 71% (Scheme 1, top).
For the facile derivatization in position 5 of the 7-azaindole scaffold, the corresponding 5-amino and 5-carboxy functionalized key intermediates 18 and 22 were prepared. The synthesis of 18 started with the SEM-protection of commercially available 5-bromo-7-azaindole (15) to deliver compound 16. The carboxy group was then introduced via lithium-halogen-exchange followed by quenching the intermediate aryllithium species with gaseous carbon dioxide at low temperatures. The treatment with elemental bromine in methylene chloride converted acid 17 cleanly to the desired intermediate 18 in excellent yields (Scheme 1, middle).
The corresponding 5-amino derivative 22 was prepared from 5-nitro-7-azaindole 19, which is accessible via a previously reported scalable three-step process [36]. Starting from intermediate 19, the 3-bromo substituent was introduced via electrophilic bromination with NBS (20) followed by the  )). Poses were generated using the BMX crystal structure with the PDB-code 3SXR as the template.
Although we did not experimentally prove selective modification of C496 by tryptic digestion/MS experiments, crystal structure or mutation of the target cysteine, the efficiency and stoichiometry of covalent modification in conjunction with modeling data and the binding modes of analogs 9a and 10a in JAK3 strongly support the specific labeling of this residue, which is the only accessible cysteine in or proximal to the BMX ATP binding site. In addition, much lower activity of non-reactive analog 12b compared to acrylamide 12a on all enzymes tested complies with a covalent mode of action on other kinases with an equivalent cysteine placement.

Compound Synthesis
The key transformation in the synthetic route to the inhibitors disclosed herein was the late stage introduction of the entire N-aryl acrylamide warhead/linker fragment via a Suzuki coupling. This reaction could be performed under very mild conditions preventing possible side reactions involving the electrophilic Michael acceptor system. The synthesis of the key boronic acid ester 14 was realized as a high yielding two step sequence starting from 3-bromoaniline. The latter was smoothly converted to the pinacol boronate 13 under Miyaura conditions [35] followed by the acylation with acryloyl chloride to deliver the bench stable building block 14 in a good overall yield of 71% (Scheme 1, top). SEM-protection of the indole nitrogen atom (21). Finally, the nitro group was reduced under mild conditions with zinc and ammonium formate to obtain the key building block 22 (Scheme 1, bottom). The early hit compounds 9a and 10a and their corresponding unsubstituted phenyl analogs 9c and 10c were synthesized following slightly modified routes (Schemes 2 and 3). Starting from plain 7-azaindole, the bromination of position 3 was performed with NBS to give compound 23 in almost quantitative yield. In the following step, the indole nitrogen atom was protected with a tosyl group via deprotonation with sodium hydride and reaction with tosyl chloride to obtain intermediate 24.
The latter was coupled with phenylboronic acid or building block 14 under Suzuki conditions to yield the compounds 25 and 26, respectively. Tosyl deprotection under basic conditions in methanol or tert-butanol, respectively, delivered the final compounds 9c and 9a. The corresponding 5-amino derivative 22 was prepared from 5-nitro-7-azaindole 19, which is accessible via a previously reported scalable three-step process [36]. Starting from intermediate 19, the 3-bromo substituent was introduced via electrophilic bromination with NBS (20) followed by the SEM-protection of the indole nitrogen atom (21). Finally, the nitro group was reduced under mild conditions with zinc and ammonium formate to obtain the key building block 22 (Scheme 1, bottom).
The early hit compounds 9a and 10a and their corresponding unsubstituted phenyl analogs 9c and 10c were synthesized following slightly modified routes (Schemes 2 and 3). Starting from plain 7-azaindole, the bromination of position 3 was performed with NBS to give compound 23 in almost quantitative yield. In the following step, the indole nitrogen atom was protected with a tosyl group via deprotonation with sodium hydride and reaction with tosyl chloride to obtain intermediate 24.
The latter was coupled with phenylboronic acid or building block 14 under Suzuki conditions to yield the compounds 25 and 26, respectively. Tosyl deprotection under basic conditions in methanol or tert-butanol, respectively, delivered the final compounds 9c and 9a. The early hit compounds 9a and 10a and their corresponding unsubstituted phenyl analogs 9c and 10c were synthesized following slightly modified routes (Schemes 2 and 3). Starting from plain 7-azaindole, the bromination of position 3 was performed with NBS to give compound 23 in almost quantitative yield. In the following step, the indole nitrogen atom was protected with a tosyl group via deprotonation with sodium hydride and reaction with tosyl chloride to obtain intermediate 24.  The series of N-substituted azaindole 5-carboxamides (compounds 11a and 11c-g) was accessible via a three-step procedure starting from key intermediate 18 (Scheme 4). First, amides 31ag were prepared via CDI-mediated acid activation and reaction with the corresponding amines. The latter intermediates were then coupled with 14 under mild Suzuki conditions at ambient temperature to give 32a-g in favorable yields. The final compounds 11a and 11c-g were obtained after acidpromoted SEM-cleavage using TFA in DCM at ambient temperature (Scheme 4). The final inhibitor series 12a and 12c-i featuring an inverted amide functionality was prepared via a similar route as described for 11a and 11c-g (Scheme 5). Building block 22 was reacted with the corresponding acid chlorides to deliver the amide derivatives 33a-i in moderate to excellent yields. The Suzuki coupling with boronate ester 14 was performed under the mild conditions established before to afford precursors 34a-i. The latter were converted to the final compounds 12a and 12c-i under the same acidic conditions as described above (Scheme 5). The series of N-substituted azaindole 5-carboxamides (compounds 11a and 11c-g) was accessible via a three-step procedure starting from key intermediate 18 (Scheme 4). First, amides 31a-g were prepared via CDI-mediated acid activation and reaction with the corresponding amines. The latter intermediates were then coupled with 14 under mild Suzuki conditions at ambient temperature to give 32a-g in favorable yields. The final compounds 11a and 11c-g were obtained after acid-promoted SEM-cleavage using TFA in DCM at ambient temperature (Scheme 4).  The series of N-substituted azaindole 5-carboxamides (compounds 11a and 11c-g) was accessible via a three-step procedure starting from key intermediate 18 (Scheme 4). First, amides 31ag were prepared via CDI-mediated acid activation and reaction with the corresponding amines. The latter intermediates were then coupled with 14 under mild Suzuki conditions at ambient temperature to give 32a-g in favorable yields. The final compounds 11a and 11c-g were obtained after acidpromoted SEM-cleavage using TFA in DCM at ambient temperature (Scheme 4). The final inhibitor series 12a and 12c-i featuring an inverted amide functionality was prepared via a similar route as described for 11a and 11c-g (Scheme 5). Building block 22 was reacted with the corresponding acid chlorides to deliver the amide derivatives 33a-i in moderate to excellent yields. The Suzuki coupling with boronate ester 14 was performed under the mild conditions established before to afford precursors 34a-i. The latter were converted to the final compounds 12a and 12c-i under the same acidic conditions as described above (Scheme 5). The final inhibitor series 12a and 12c-i featuring an inverted amide functionality was prepared via a similar route as described for 11a and 11c-g (Scheme 5). Building block 22 was reacted with the corresponding acid chlorides to deliver the amide derivatives 33a-i in moderate to excellent yields. The Suzuki coupling with boronate ester 14 was performed under the mild conditions established before to afford precursors 34a-i. The latter were converted to the final compounds 12a and 12c-i under the same acidic conditions as described above (Scheme 5). To access the propionamide analogs of selected inhibitors as negative control compounds, the corresponding acrylamides were hydrogenated in the presence of a Pd/C catalyst. Due to the absence of sensitive functional groups, these conversions usually proceeded smoothly and gave the desired propionamides 9b, 10b, 11b and 12b in good to excellent yields (Scheme 6).

Discussion
Here we describe the discovery of a novel class of covalent inhibitors of the TEC family kinases, most notably BMX and BTK. Development started from compounds 9a and 10a belonging to an unprecedented class of covalent JAK3 inhibitors, which possessed potent off-target activity on the TEC kinases BMX and TXK while showing moderate to high selectivity against the other eight protein kinases with an equivalent cysteine placement. Guided by the crystal structures of 9a and 10a in complex with JAK3, modification of the hinge binding motif as well as back pocket binding moieties was successfully exploited as a strategy to fine tune the binding profiles of these irreversible inhibitors that target a cysteine located at the αD-1 position. This approach enabled the development of irreversible inhibitors for BMX and BTK, which showed a general preference for kinases with a less bulky threonine gatekeeper residue, enabling high selectivity over JAK3 and MKK7 that harbor a methionine, and ITK that possess a phenylalanine at this position.
The key step in the synthetic access to these inhibitors was the late stage introduction of the Nphenyl acrylamide warhead via Suzuki coupling under exceptionally mild conditions. This facilitated broad SAR evaluation and fast library synthesis. Starting from unsubstituted carboxamide 10a, we introduced N-alkyl substituents to induce a clash with bulkier gatekeeper moieties, but this modification led to a drop in potency on both JAK3 and BMX. In contrast, inversion of the carboxamide in the 5-position of the azaindole scaffold retained BMX inhibitory potency, while activity on JAK3 was strongly decreased. Besides the expected steric clash with the methionine gatekeeper moiety in JAK3, molecular modeling suggested inverted amides to form an additional To access the propionamide analogs of selected inhibitors as negative control compounds, the corresponding acrylamides were hydrogenated in the presence of a Pd/C catalyst. Due to the absence of sensitive functional groups, these conversions usually proceeded smoothly and gave the desired propionamides 9b, 10b, 11b and 12b in good to excellent yields (Scheme 6). To access the propionamide analogs of selected inhibitors as negative control compounds, the corresponding acrylamides were hydrogenated in the presence of a Pd/C catalyst. Due to the absence of sensitive functional groups, these conversions usually proceeded smoothly and gave the desired propionamides 9b, 10b, 11b and 12b in good to excellent yields (Scheme 6).

Discussion
Here we describe the discovery of a novel class of covalent inhibitors of the TEC family kinases, most notably BMX and BTK. Development started from compounds 9a and 10a belonging to an unprecedented class of covalent JAK3 inhibitors, which possessed potent off-target activity on the TEC kinases BMX and TXK while showing moderate to high selectivity against the other eight protein kinases with an equivalent cysteine placement. Guided by the crystal structures of 9a and 10a in complex with JAK3, modification of the hinge binding motif as well as back pocket binding moieties was successfully exploited as a strategy to fine tune the binding profiles of these irreversible inhibitors that target a cysteine located at the αD-1 position. This approach enabled the development of irreversible inhibitors for BMX and BTK, which showed a general preference for kinases with a less bulky threonine gatekeeper residue, enabling high selectivity over JAK3 and MKK7 that harbor a methionine, and ITK that possess a phenylalanine at this position.
The key step in the synthetic access to these inhibitors was the late stage introduction of the Nphenyl acrylamide warhead via Suzuki coupling under exceptionally mild conditions. This facilitated broad SAR evaluation and fast library synthesis. Starting from unsubstituted carboxamide 10a, we introduced N-alkyl substituents to induce a clash with bulkier gatekeeper moieties, but this modification led to a drop in potency on both JAK3 and BMX. In contrast, inversion of the carboxamide in the 5-position of the azaindole scaffold retained BMX inhibitory potency, while activity on JAK3 was strongly decreased. Besides the expected steric clash with the methionine gatekeeper moiety in JAK3, molecular modeling suggested inverted amides to form an additional

Discussion
Here we describe the discovery of a novel class of covalent inhibitors of the TEC family kinases, most notably BMX and BTK. Development started from compounds 9a and 10a belonging to an unprecedented class of covalent JAK3 inhibitors, which possessed potent off-target activity on the TEC kinases BMX and TXK while showing moderate to high selectivity against the other eight protein kinases with an equivalent cysteine placement. Guided by the crystal structures of 9a and 10a in complex with JAK3, modification of the hinge binding motif as well as back pocket binding moieties was successfully exploited as a strategy to fine tune the binding profiles of these irreversible inhibitors that target a cysteine located at the αD-1 position. This approach enabled the development of irreversible inhibitors for BMX and BTK, which showed a general preference for kinases with a less bulky threonine gatekeeper residue, enabling high selectivity over JAK3 and MKK7 that harbor a methionine, and ITK that possess a phenylalanine at this position.
The key step in the synthetic access to these inhibitors was the late stage introduction of the N-phenyl acrylamide warhead via Suzuki coupling under exceptionally mild conditions. This facilitated broad SAR evaluation and fast library synthesis. Starting from unsubstituted carboxamide 10a, we introduced N-alkyl substituents to induce a clash with bulkier gatekeeper moieties, but this modification led to a drop in potency on both JAK3 and BMX. In contrast, inversion of the carboxamide in the 5-position of the azaindole scaffold retained BMX inhibitory potency, while activity on JAK3 was strongly decreased. Besides the expected steric clash with the methionine gatekeeper moiety in JAK3, molecular modeling suggested inverted amides to form an additional hydrogen bond towards the hydroxy group of threonine gatekeeper moieties, which drives the carbonyl's substituent towards the hydrophobic region I.
The most promising compounds 12a and 12c were further profiled against the TEC kinase family and other kinases with an αD-1 cysteine. As expected, low activity was detected on kinases with a non-threonine gatekeeper. However, the promising selectivity pattern observed for starting compounds 9a and 10a was partially lost. Most notably, optimization was accompanied by a strong increase in BTK potency with compound 12a being equipotent on BMX and BTK, while 12c slightly favored BTK with subnanomolar inhibitory activity in the enzymatic assay. Moderate selectivity was observed against the other TEC kinases and BLK. However, compound 12c also showed significant off-target activity on HER family kinases while 12a maintained good selectivity against the latter. Comparison with unreactive analog 12b, which showed weak activity on all tested kinases underlined the importance of the covalent mode of action. Highly efficient covalent modification of BMX was experimentally shown for several analogs via mass spectrometry while potent cellular BMX and BTK engagement was demonstrated by means of a NanoBRET TM assay.
Targeting BMX has been proposed as a strategy for treatment of various diseases including cardiovascular disorders [37] and certain cancers [26,38,39]. Nevertheless, only three inhibitors have been developed to date: BMX-IN-1 (6, see Figure 1), the analog JS25 (7a) and the type II inhibitor CHMFL-BMX-078 (8). However, these inhibitors still exhibit pronounced affinities to several other kinases with different selectivity profiles. Off-target activities could thus limit their use as a sole tool for biological study. Moreover, their chemical structures suggest unfavorable physicochemical properties due to size and the high portion of sp 2 atoms. The dual covalent BMX/BTK inhibitors presented here thus complement the repository of available BMX inhibitors. It should be mentioned, however, that caution must be taken when comparing covalent inhibitors based on (apparent) IC 50 data since the latter depend on incubation time. For example, it has recently been demonstrated in the case of MKK7 and ibrutinib that a longer incubation could result in a misled nanomolar inhibitory activity observed for an inhibitor that has only weak reversible binding affinity to the kinase [16]. Nonetheless, determination of meaningful kinetic data, i.e., k inact and K I , to disentangle contributions of reversible binding and the subsequent bond-forming step remains highly elaborate, thus for BMX such values have only been reported for JS25 and a few related compounds from the same study [27]. Moreover, and in contrast to the aforementioned studies, the compounds presented herein have not been characterized in terms of wider kinome selectivity. Improvements also need to be made concerning the intra-TEC family selectivity. Future efforts will primarily focus on increasing the selectivity against BTK, for example by modifying the residues targeting the hydrophobic region I or by extending the latter towards the DFG-out pocket to generate type II inhibitors [16].
Overall, the presented optimization approach offers a strategy that can be exploited to fine tune existing covalent inhibitors towards unrelated kinases. Compared to previous compounds, the BMX inhibitors discovered here are based on an alternative and readily optimizable chemotype, which may be beneficial for further development of probes that exclusively target BMX. Currently, there is no selective inhibitor or chemical probe available for BMX. However, combinatorial use of our dual irreversible BMX/BTK inhibitors along with other inhibitors with different off-target profiles as a set of chemogenomic compounds would nevertheless allow dissecting biological functions of BMX as well as its role in disease development, which will enable further validation of this kinase as a potential therapeutic target.

General Information
Chemical synthesis was carried out using commonly applied techniques and general procedures. All starting materials and reagents were of commercial quality and were used without further purification. Thin layer chromatography (TLC) was carried out on Merck 60 F254 silica plates

General Synthetic Procedures
General Procedure A (Suzuki coupling): In a screw-top reaction vial were combined the corresponding boronic acid or ester and the aryl bromide under argon atmosphere. Dioxane and the aqueous base were added subsequently, and the mixture was degassed carefully via several vacuum/argon cycles. Finally, the catalyst system was added to the reaction and the degassing procedure was repeated. The vial was sealed and heated to the indicated temperature until reaction control (TLC or HPLC) showed complete consumption of starting materials. The reaction was cooled to ambient temperature, diluted with EtOAc and washed with brine. The organic phase was dried over Na 2 SO 4 and evaporated to dryness. The residue was usually purified via flash chromatography using an appropriate solvent system.
General Procedure B (SEM cleavage): The SEM-protected substrate was dissolved in dry DCM and TFA was added subsequently. The reaction was stirred until TLC indicated complete consumption of the starting material. The majority of TFA was removed under reduced pressure and the residue was taken up in EtOH and was basified with aqueous NH 3 solution. The mixture was stirred at ambient temperature until complete consumption of the hydroxymethyl intermediate (usually overnight) was observed. The volatiles were stripped of under reduced pressure and the residue purified via flash chromatography as indicated.
General Procedure C (amide coupling with CDI): Carboxylic acid 18 was dissolved in dry DMF (ca. 0.1 M) and CDI was added at ambient temperature. The mixture was stirred until gas evolution ceased (usually about 1 h) before the corresponding amine was added to the reaction. Stirring was continued until reaction control indicated complete conversion and the reaction was quenched by addition of water followed by dilution with EtOAc. The organic phase was washed with sat. NaHCO 3 (two times) and brine, prior to solvent evaporation and purification by flash chromatography.
General Procedure D (amide coupling with acid chlorides): To a solution of 22 and Et 3 N in dry DCM (ca. 0.1 M) was added the corresponding acid chloride under ice-cooling. After complete addition, the cooling bath was removed and stirring was continued at ambient temperature until reaction control (HPLC or TLC) indicated complete conversion. The reaction was quenched by addition of water followed by dilution with EtOAc. The organic phase was washed with sat. NaHCO 3 , and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was purified via flash chromatography with an appropriate eluent system.

Synthesis and Characterization of Intermediates and Final Compounds
3-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)aniline (13): In a flame-dried Schlenk flask were suspended 1.47 g dry KOAc (15.0 mmol), 860 mg 3-bromoaniline (5.00 mmol) and 1.27 g bis(pinacolato)diboron (5.00 mmol) in 20 mL dry dioxane. The mixture was carefully degassed via three vacuum/argon cycles and 9 mg XPhos Pd G4 (10 µmol) was added subsequently. The degassing procedure was repeated and then the reaction was heated to 90 • C oil-bath temperature overnight. After cooling to ambient temperature, the mixture was diluted with EtOAc and filtered over a bed of celite. The filtrate was washed with water (three times) and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was triturated with heptane and the precipitate isolated by filtration to yield 914 mg (84%) of the product as a slightly brownish solid. 1 (16): To an ice-cooled solution of 1.96 g 5-bromo-1H-pyrrolo[2,3-b]pyridine (10 mmol) in 10 mL dry DMF were added 518 mg sodium hydride (60%wt dispersion in min. oil, 12.9 mmol) and stirring was continued for about 30 min. Subsequently were added 1.95 mL SEM-Cl (10.9 mmol) and stirring was continued until TLC indicated complete conversion. The reaction was diluted with 120 mL EtOAc and transferred to a separatory funnel. The organic phase was successively washed with water (four times) and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was purified via flash chromatography (petrol ether/EtOAc (0-15%)) to obtain 2.55 g (78%) as a colorless oil. 1  After another 30 min, the reaction was quenched with water and slowly warmed to ambient temperature. After dilution with EtOAc, the organic phase was extracted with 0.5 M aqueous NaOH (three times). The combined extracts were acidified with 2M HCl and back-extracted with EtOAc (three times). The combined organic phases were washed with brine, dried over Na 2 SO 4 and evaporated to dryness. The crude product was purified via flash chromatography (petrol ether/EtOAc + 5% MeOH + 2% AcOH (25-75%)) to yield 1.14 g (51%) of pure 17 as an off-white semisolid. 1 (20): To an ice-cooled suspension of 1.8 g 5-nitro-1Hpyrrolo[2,3-b]pyridine (11.0 mmol) in 40 mL dry DMF were added 2.36 g N-bromosuccinimide (13.2 mmol) as solid in small portions. After complete addition, the cooling bath was removed and the reaction was stirred for 3 h at ambient temperature. The mixture was diluted with water and aqueous Na 2 S 2 O 3 and the precipitate was isolate by filtration. The filter cake was washed with water and dried at 60 • C in a convection oven to yield 1.86 g (70%) of the title compound as a yellow solid. 1 (21): To an ice-cooled solution of 2.14 g 20 (8.9 mmol) in 20 mL dry DMF were added 462 mg sodium hydride (60%wt dispersion in min. oil, 11.5 mmol) and stirring was continued for about one h. Subsequently were added 1.7 mL SEM-Cl (9.7 mmol) and stirring was continued until TLC indicated complete conversion. The reaction was diluted with 120 mL EtOAc and transferred to a separatory funnel. The organic phase was successively washed with water (four times) and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was purified via flash chromatography (petrol ether/EtOAc (0-20%)) to obtain 1.85 g (56%) 21 as a yellow oil. 1 (22): To a solution of 1.85 g 21 (5.0 mmol) in 40 mL abs. EtOH were added 1.63 g finely powdered zinc (24.9 mmol) and 1.57 g ammonium formate (24.9 mmol). The mixture was stirred for 22 h at 50 • C and was then filtered over a celite pad to remove unreacted zinc. The filtrate was concentrated under reduced pressure and taken up in EtOAc. The organic phase was washed with sat NH 4 Cl and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was purified via flash chromatography (petrol ether/EtOAc 20-60%) to obtain 0.86 g (51%) of the title compound as a brown oil. 1 (23): To an ice-cooled solution of 473 mg 1H-pyrrolo[2,3-b]pyridine (4.0 mmol) in 8 mL DMF were added 726 mg N-bromosuccinimide (4.1 mmol) in several portions. After complete addition, the cooling bath was removed and stirring was continued for 4 h. The reaction mixture was poured on sat. NaHCO 3/ ice and the resulting suspension was stirred for ca. 10 min until a homogenous precipitate was formed. The solids were collected by filtration, washed with water and dried in vacuo to yield 753 mg (96%) of the title compound as a white solid. 1 1 mmol) and stirring was continued until TLC indicated complete conversion. The reaction was diluted with 25 mL EtOAc and transferred to a separatory funnel. The organic phase was successively washed with water and brine, prior to drying over Na 2 SO 4 and evaporation. The residue was triturated with chilled MeOH and filtered to obtain 298 mg (85%) 24 as a white solid. 1 163.3, 149.1, 143.0, 139.5, 135.5, 132.0, 129.3, 127.4, 126.9, 123.7, 121.4, 117.2, 117.1, 116.6, 116.1 13 13

JAK3 Crystal Structure Determination
Recombinant JAK3 kinase domain was expressed and purified as described previously [21,22]. The protein (10 mg/mL) was incubated with the inhibitors at 1:1.2 molar ratio, and the complexes were crystallized using sitting drop vapor diffusion method at 4 • C and the conditions containing 25% PEG 3350, 0.1-0.2 M MgCl 2 and 0.1 M MES, pH 5.5. The crystals were cryo-protected using mother liquor supplemented with 20% ethylene glycol, and diffraction data were collected at SLS X06SA. The data were processed and scaled with XDS [40] and Aimless [41], respectively. Molecular replacement using Phaser [42] and the coordinate of JAK3 (pdb id: 5lwm) was performed. Model rebuilding alternated with refinement was performed in COOT [43] and REFMAC [44], respectively. Data collection and refinement statistics are summarized in Table 7.

Thermal Shift Assay
The kinase domains of JAK3 and BMX at 2 µM were mixed with the inhibitors at 10 µM, and subsequently SyPRO orange dye (Invitrogen Carlsbad, CA, USA) was added. The thermal shift assay and data evaluation were performed as described previously using a Real-Time PCR Mx3005p machine (Stratagene, La Jolla, CA, USA) [45,46].

Enzymatic Activity Assay
If not mentioned differently, in vitro profiling of compounds was performed at Reaction Biology Corporation using the HotSpot™ assay platform. IC 50 values were determined as singlicates using five doses with 5-or 10-fold serial dilution starting at 0.5 µM, 1 µM or 10 µM. Further details on the assay can be found on the supplier homepage (http://www.reactionbiology.com; see also Anastassiadis et al. 2011 [31]).

Cellular NanoBRET Assay
Full length BMX and BTK was cloned into pFC32K (Promega, Madison, WI, USA) for expression of a C-terminal NanoLuc fusion. The plasmids were transfected into HEK293T cells cultured in DMEM (Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Gibco, Waltham, MA, USA) and Penicillin/Streptamycin (Gibco, Waltham, MA, USA). NanoBRET assays were performed using the protocol published previously [47]. Briefly, after transfection and 20 h incubation cells were harvested and subsequently resuspended in OptiMEM (Gibco, Waltham, MA, USA). Cells were aliquoted onto 1534-well plates (Greiner, Kremsmünster, Austria), and inhibitors as well as 0.5 µM Tracer K4 (Promega) for BMX or 0.5 µM Tracer K5 (Promega, Madison, WI, USA) for BTK were added using an ECHO acoustic dispenser. The plates were incubated at 37 • C with 5% CO 2 for 2 h prior to the addition of both NanoBRET NanoGlo substrate (Promega, Madison, WI, USA) and extracellular NanoLuc inhibitor. BRET luminescence (450 nm for donor emission and 610 nm for BRET signal) was measured using PHERAstar FSX plate reader (BMG Labtech, Ortenberg, Germany). Milli-BRET units (mBU) were calculated as a ratio between BRET signal and the overall measured luminescence. A dose-response fitting was applied and IC 50 values were calculated using the Prism software. The affinity of both tracer molecules towards BMX and BTK was tested and showed similar values (EC 50 (Tracer 4, BMX) = 0.240 µM, EC50 (Tracer 5, BTK) = 0.231 µM) indicating no bias of the assay due to the higher affinity of one kinase. Experiments were performed as tripicates and repeated at least three times.

Mass Spectrometric Investigation of Covalent Binding to BMX
The kinase domain of BMX was diluted to 0.05 mM with buffer containing 20 mM Tris pH 8.0, 200 mM NaCl, 0.5 mM TCEP (pH 7.0) and was subsequently mixed with 0.075 mM inhibitors (ratio of protein to inhibitor of 1:1.5). The mixture was incubated at 4 • C. The samples were taken at different time points, the reaction stopped by diluting the sample with a 200-fold excess of 1% formic acid. The denatured protein and the adducts were assessed using electrospray time-of-flight (ESI-TOF) mass spectrometry.

Determination of Glutathione Reactivity
A total of 5 µL of a 10 mM compound stock solution in DMSO were added to 5 µL of a 4 mM Indoprofen solution in PBS as internal standard. The mixture was diluted with PBS buffer to a total volume of 1 mL of component A. Additionally, a freshly prepared solution of 10 mM GSH in PBS buffer was used as component B. A total of 250 µL of each component was mixed and immediately subjected to HPLC analysis for t 0 measurement. The probes were stored at 40 • C in between the respective injections. t 1/2 was determined by plotting natural logarithm of AUC/AUC 0 against time.

Molecular Modeling
All the modelling was performed using the Schrödinger Small-Molecule Drug Discovery Suite 2019-1 (Schrödinger, LLC, New York, NY, USA). Noncovalent docking was performed with Glide in the XP mode after protein preparation using standard settings. Covalent docking was performed with the CovDock module in the pose prediction mode using standard settings. The figures were prepared with PyMOL 1.8.