Insertion and interaction of transmembrane helices, from basic principles to rational design
The main objective of this thesis was to study α-helical membrane protein segments, from basic principles to rational design. Our intention was to investigate a wide spectrum of functions developed by the TMDs far beyond from its structural role anchoring proteins to membranes. The final goal of this thesis is to delve deep in our understanding of membrane protein biogenesis from insertion and topology determination to better characterize these ‘greasy’ proteins. In particular, we wanted to make progress on depicting how α-helical TMD-TMD interactions work and their importance in the regulation of different cellular processes like apoptosis and the control of cellular death. A better understanding of these regulation processes can improve our current knowledge and lead to new therapeutic targets. In this direction, we ought to explore the inhibition of TMD-TMD interactions as a target to regulate cell death. The results of this thesis were obtained using some of the more relevant techniques from the biochemistry and molecular biology toolbox as: plasmid cloning, site-directed mutagenesis, SDS-PAGE electrophoresis, western blotting, transcription/translation expression in vitro, apoptosis assays in cell lines (HeLa), and also protein engineering approaches based on bimolecular complementation protocols like BiFC (Bimolecular Fluorescence Complementation) and BLaTM (based on the use of a split β-Lactamase fused to a TMD). These techniques led us to a better description of how polar residues can be inserted into the membrane. They also helped us to improve our understanding of the insertion process, the topology determination of viral proteins and to better comprehend the complexity of TMD-TMD interactions and their role to fine-tune the apoptotic networks, pointing out the importance of the TM region. Finally, we used these interactions to computationally design TM inhibitors as a new possible therapeutic agents to regulate cellular death by modulating TMD-TMD interactions. The research related to this work resulted in four first-author publications and one manuscript that is currently being revised, which constitute the four chapters of this thesis, and two more papers included in the annexed section. The annexed papers (see section 8), whereof I am the first author, include a review depicting some of the most relevant methods used in this thesis to study TMD interactions. Additionally, a patent (P202230029) including part of this work has been filled by the University of Valencia. The following pages are a summary of the results and discussion of these papers organized in four chapters. The objective of this work was to estimate the energetic contribution of intra-helical salt bridges to the insertion of TMDs into biological membranes. The presence of intra-helical salt bridges in TMDs, as well as their impact on insertion, has not been properly studied yet. α-helical TMDs are largely composed of apolar residues because of the hydrophobic nature of the membrane. Nevertheless, in some cases, membrane embedded proteins carry polar amino acids in their TMDs for proper folding or functional purposes (Baeza-Delgado et al., 2013). In fact, the presence of polar and charged amino acids in TMDs is more frequent than what would be expected according to the hydrophobic nature of the environment (Bañó-Polo et al., 2012), especially when these residues are present in pairs. Salt bridges are electrostatic interactions between negatively and positively charged amino acids that play a prevalent role in protein stabilization (Marqusee and Baldwin, 1987). Analysing the TMDs from membrane protein structures we observed that charged pairs of amino acids are especially prevalent at positions i, i+1; i, i+3 and i, i+4. Oppositely paired charges located at these positions could potentially form salt bridges as they are all on the same face of the helix and are close enough, in terms of atomic distances. To assess the contribution of putative salt bridges to the translocon-mediated membrane insertion, we used as a vehicle the leader peptidase (Lep) protein from Escherichia coli. The Lep protein consists of two TMDs (H1 and H2) connected by a cytosolic loop (P1) and a large C-terminal (P2) domain, which inserts into endoplasmic reticulum (ER)-derived rough microsomes (RM) with both termini located in the lumen of the microsomes (von Heijne, 1989). The designed TMDs were inserted into the luminal P2 domain and flanked by two N-linked glycosylation acceptor sites (G1 and G2). Glycosylation occurs exclusively in the lumen of the ER (or the microsomes) because of the location of the oligosaccharyltransferase (OST) active site (a translocon-associated enzyme responsible for the oligosaccharide transfer) (Braunger et al., 2018). Glycosylation of an acceptor site increases the apparent molecular mass of the protein (~2.5 kDa), which facilitates its identification by gel electrophoresis. We first compared the effect of the oppositely charged Lys and Asp residues on the insertion of a 19-residue-long hydrophobic artificial scaffold (L4/A15, 4 leucines and 15 alanines), designed to stably insert into the RM membranes and “insulated” from the surrounding sequence by N- and C-terminal GGPG- and -GPGG (G, glycin; P, prolin) tetrapeptides. Single Lys and Asp residues were placed at positions 8 and 12 respectively, and pairs of Lys-Asp residues were designed to cover positions 7–12 (more than one helical turn). When pairs of charged residues were present, our results showed a tendency to insert more efficiently when oppositely paired charges were placed at positions that are permissive distances for salt bridge formation (i, i+1; i, i+3; i, i+4), an effect not observed in the current prediction algorithms. Similar results were obtained on a different Leu/Ala background with a slightly higher insertion efficiency (L5/A14, 5 leucines and 14 alanines). After testing the effect of oppositely charged residues in the insertion of model sequences, we decided to look for salt bridges in natural proteins. We analysed the TMDs from high-resolution membrane protein structures. Then, we generated a list of potential candidates for further in vitro and whole cell studies. We selected the helix G from Natronomonas pharaonis halorhodopsin (PDB code: 3QBG) and the helix A from the Oryctolagus cuniculus calcium ATPase (PDB code: 1SU4). In both cases we studied the insertion of these helices using in vitro assays (Lep in microsomes) and eukaryotic cells (Lep and CLTM systems). Halorhodopsin from Natronomonas pharaonis is a protein made up of seven TMDs (A to G) and a retinal chromophore bound via a protonated Schiff base to the ε-amino group of the Lys258, located in the middle of helix G (Kanada et al., 2011). In silico analysis of the apoprotein 3QBG structure showed a Lys-Arg pair, involving i, i+4 Lys258 and Asp254 residues from helix G. The distance between charges in the crystal structure of the apoprotein was about 3.5 Å, a permissive distance for a salt bridge formation. Then, we designed three mutants that were supposed to perturb salt bridge interaction in different ways: the K258D mutant places two charged residues with the same polarity at positions i, i+4; the K258A mutant replaces one of the charged residues by a non-polar amino acid, and the K258Y/Y259K double mutant places the two different charges at a non-permissive position for salt bridge formation (i, i+5) while preserving amino acid composition. The results of the Lep-based glycosylation assay indicated that wild type and K258A mutant are inserted properly, but when the salt bridge is disrupted, insertion efficiency decreased substantially (~0.5 kcal/mol). These results were replicated in HEK-293T cells. Cells transfected with the chimera containing helix G native sequence rendered almost full insertion. In contrast, cells transfected with the construct harbouring i, i+5 sequence rendered almost exclusively membrane translocation. These results emphasized the relevance of intra-helical salt bridges in translocon-mediated TM insertion, especially in the cellular environment. Our second studied protein was the calcium ATPase from Oryctolagus cuniculus. This protein contains a bundle of 10 TM helices (A to J) (Toyoshima et al., 2000). In silico analysis of 1SU4 highlighted an Asp-Arg pair, involving Asp59 and Arg63, in the centre of helix A. The distance between these charges in the crystal structure was about 3.0 Å, clearly within the permissive range for salt bridge formation. Again, the results of the Lep-based glycosylation assay demonstrated the efficient insertion of the native helix A sequence. Importantly, when the paired residues were placed at the non-permissive i, i+5 positions the insertion efficiency was decreased, with a ΔGapp estimated in ~0.7 kcal/mol. Similar results were also observed in HEK-293T cell experiments. In summary, charged residues found in α-helices can be important for function (Lin and Lin, 2018) but also provide a stabilizing effect (Armstrong and Baldwin, 1993). Pairs of oppositely charged residues can form salt bridges that could allow them to facilitate TMD insertion when facing the hydrophobic environment (Walther and Ulrich, 2014). Salt bridges might also be important for shielding the charges during translocon-mediated TMD insertion (Whitley et al., 2021). In TMDs, oppositely charged residue pairs are more prevalent at positions i, i+1; i, i+3 and i, i+4, compatible with salt-bridge formation. By analysing two native helices containing intra-helical salt bridges we found that the free energy of insertion (ΔGapp) is significantly reduced when both oppositely charged residues are spaced at a permissive distance. These results indicate that intra-helix salt bridges could form during translocon-assisted insertion or even earlier, since TM helices can be compacted inside the ribosome exit tunnel (Bañó-Polo et al., 2018). The reduction of ΔGapp in these natural proteins is between 0.5-0.7 kcal/mol. As found in the case of the halorhodopsin helix G, this reduction might be higher in the cell context, since some auxiliary components of the membrane insertion machinery (Chitwood and Hegde, 2020; Shurtleff et al., 2018; Tamborero et al., 2011) might be not represented in the microsomal vesicle preparations. Current prediction algorithms for membrane protein insertion tend to overestimate the free energy penalty of charged residues in TMDs. Incorporating the effect of potential salt bridges in the reduction of ΔGapp during membrane integration could help to improve future prediction tools. As soon as the genome of the SARS-CoV-2, the virus responsible of the COVID-19 pandemic, was released, we started a project aimed to study the envelope (E) protein. The E protein is the smallest and has the lowest copy number among the membrane proteins found in the lipid envelope of mature virus particles (Bar-On et al., 2020). However, it is critical for pathogenesis of the SARS-CoV-2 and other human coronaviruses (Almazán et al., 2013; Ruch and Machamer, 2012; Xia et al., 2021; Zheng et al., 2021) and has been described as a viroporin. Interestingly, the sgRNA encoding E protein is one of the most abundantly expressed transcripts despite the protein having a low copy number in mature viruses (Wu et al., 2020). This sgRNA encodes a 75 residues long polypeptide with a predicted molecular weight of approximately 8 kDa. Comparative sequence analysis of the E protein of SARS-CoV-2 and the other six known human coronaviruses do not reveal any large homologous/identical regions, with only the initial Met, Leu39, Cys40 and Pro54 being ubiquitously conserved. Regarding overall sequence similarity SARS-CoV-2 E protein has the highest similarity to SARS-CoV (94.74%) with only minor differences, followed by MERS-CoV (36.00%). Interestingly, sequence similarities are significantly lower for the other four human coronaviruses, which usually cause mild to moderate upper-respiratory tract illness typical for common cold. To determine its membrane topology, we assayed E protein insertion in microsomal membranes using in vitro transcription/translation experiments in the presence of [35S]-labelled amino acids but also in eukaryotic membranes using HEK-293T cells. Using a glycosylation based assay as a molecular reporter, we determined that the SARS-CoV-2 E protein integrates into the membrane co-translationally as a single-spanning membrane protein with an Ntlum/Ctcyt orientation in in vitro and in vivo systems. This topology is compatible with the ion channel capacity described previously (Verdiá-Báguena et al., 2012). Furthermore, this topology is reinforced by different topological determinants present in the SARS-CoV-2 E protein sequence. In all seven human coronaviruses there is a strongly conserved positively charged residue placed after the hydrophobic region. It is worth to mention that this residue is an arginine (Arg38) in MERS-CoV, SARS-CoV and SARS-CoV-2, while in the other human coronaviruses is a lysine. Also, the alignment of MERS-CoV, SARS-CoV and SARS-CoV-2 E proteins unveils a tendency to accumulate a net positive charge balance C-terminally to the TMD, which correlates with the positive-inside rule, suggesting an increasing robustness in the topology determination from MERS-CoV to SARS-CoV-2. We experimentally confirmed this increasing robustness by modifying the full protein charge balance for all three pathogenic E proteins. In all three cases, the conserved Arg38 residue plays a limited role in the topology determination. Our data also suggested that the Arg to Glu mutation present in both SARS-CoVs’ N-terminus compared with MERS-CoV sequence, is most likely one of the mechanisms contributing to the proved topology robustness of the SARS-CoVs by converting the net charge of 0 at N-terminal region of MERS-CoV into a -2 in both SARS-CoVs, in good agreement with the so-called “negative outside enrichment” rule (Baker et al., 2017). Programmed cell death is a fundamental process in the development of multicellular organisms contributing to the balance among cell death, proliferation, and differentiation, that is crucial for tissue development and homeostasis (Kerr et al., 1972). Also, protection and defence against many disorders, including cancer and pathogen-related diseases, rely on apoptosis to eliminate the affected cells (Häcker, 2018; Hua et al., 2019). One of the primary modulators of apoptosis is the B-cell lymphoma 2 (Bcl2) protein family (Kim et al., 2006). The proteins in this family can be divided in anti-apoptotic (e.g., Bcl2 and BclxL) (Boise et al., 1993), pro-apoptotic (e.g., Bax and Bak)(Oltvai et al., 1993), and BH3-only apoptosis activators (e.g., Bid and Bmf) (Wang et al., 1996). Most pro-apoptotic and anti-apoptotic proteins in this family share up to four main Bcl2 sequence homology domains, known as BH1, BH2, BH3, and BH4, while BH3-only members have solely the BH3 domain. In addition, many Bcl2 family members have a TMD in the carboxyl-terminal end that effectively allows for insertion of the protein into the target lipid bilayer (Delbridge et al., 2016). Cellular Bcl2 (cBcl2) proteins can physically interact with each other, forming homo-and hetero-oligomers that are crucial for programmed cell death regulation (Cosentino and García-Sáez, 2017; Kelekar et al., 1997; Oltvai et al., 1993; O’Neill et al., 2006; Wang et al., 1996; Xie et al., 1998). To prevent the premature death of host cells, viruses have developed functional homologues of cBcl2, known as viral Bcl2 (vBcl2), as a strategy to modulate cell death (Kvansakul et al., 2017; Polčic et al., 2017). Although there is low sequence homology between vBcl2 and cBcl2, crystal structures reveal a structural homology in key domains (Galluzzi et al., 2008; Kvansakul and Hinds, 2013). In this work, we firstly proved that vBcl2 contain a functional TMD in the Ct end, like their cellular counterparts. We selected six vBcl2 proteins from two distinct viral families (3 hepesviruses and 3 poxviruses). BHRF1 (Human gammaherpesvirus 4 – Epstein–Barr virus, HHV4) (Pearson et al., 1987), ORF16 (Human gammaherpesvirus 8 – Kaposi’s sarcoma–associated herpesvirus, HHV8) (Cheng et al., 1997), ORF16 (Bovine gammaherpesvirus 4, BoHV4) (Bellows et al., 2000), F1L (Vaccinia virus, VacV) (Nichols et al., 2017; Wasilenko et al., 2003), M11L (Myxoma virus, MyxV) (Douglas et al., 2007; Nichols et al., 2017), and ORFV125 (Orf virus, OrfV) (Westphal et al., 2007). To avoid confusion, here we use the viral acronym to refer to the vBcl2 protein. In silico analyses suggested the presence of TMDs in the selected vBcl2. Then, we aimed to explore the membrane insertion capacity of the predicted segments using an in vitro assay based on the E. coli leader peptidase (Lep) described before (see section 3.1). Using this glycosylation-based system we determined that all the studied vBcl2 regions were efficiently inserted into microsome membranes. After determining that the vBcl2s included a Ct end anchoring TMD, we sought to study whether their role went beyond that of a structural anchor. As interactions between cBcl2 TMDs have been reported in biological membranes (Andreu-Fernández et al., 2017), we decided to study the vBcl2 TMD capability of homo and hetero-oligomerization in biological membranes. For this purpose, we used two different systems of bimolecular complementation based on different reporters and in different model organisms. We used BiFC (bimolecular fluorescent complementation) approach (Kerppola, 2006) adapted for the study of intramembrane interactions in eukaryotic cells (Andreu-Fernández et al., 2017; Grau et al., 2017). This technique is based on a split venus fluorescent protein (VFP). Each of the two non-fluorescent fragments of the VFP is fused to one of the studied TMDs and expressed in eukaryotic cells. The two fragments of the VFP have no affinity for each other. The VFP will be reconstituted (and its fluorescence) only when TMD-TMD interaction is reported. The TMDs of HHV4, HHV8, VacV, MyxV, and OrfV showed a homo-interaction capability above the controls and similar to that observed with cBcl2 TMD. However, BoHV did not show VFP-associated ﬂuorescence significantly higher than the negative controls. Western blot analysis showed comparable expression levels of all chimeras. To investigate the potential TMD-TMD heteromeric interactions between vBcl2 and cBcl2 we used the previously described BiFC approach. We investigated the potential TMD-TMD interactions between vBcl2 and the anti-apoptotic cellular Bcl2 and BclxL proteins. Interestingly, all viral TMDs included in the assay could interact with the TMD of Bcl2. However, although the majority of vBcl2 TMDs also interacted with BclxL TMD, BoHV and MyxV TMDs did not. We also analysed the interactions between vBcl2 TMDs and the TMDs of the cellular pro-apoptotic Bax and Bak proteins. In the absence of apoptotic stimulus and independent of any cytosolic (soluble) domain contacts, all three poxviral TMDs (VacV, MyxV, and OrfV) could interact with Bax and Bak TMDs. On the contrary, herpesviruses HHV4, HHV8, and BoHV TMDs showed no interaction with Bax or Bak TMDs. Finally, we included in the BiFC-based screening the TMDs from BH3-only apoptotic modulators Bik and Bmf (Andreu-Fernández et al., 2016). The TMDs of HHV4, HHV8, VacV, and OrfV could interact with the TMD of Bik. However, viral interactions with the TMD of Bmf were limited, only the HHV8 TMD interacts with Bmf TMD. Also, we further explored some of these interactions (e.g., HHV8-Bcl2 and MyxV-Bax) using the BLaTM assay (Schanzenbach et al., 2017) and computational modelling. BLaTM is a genetic tool designed to study TMD–TMD interactions in bacterial membranes. This tool is based on the use of a split β-Lactamase fused to the TMDs of study and expressed in E. coli cells. The TMD-TMD interaction aims the reconstitution of the β-Lactamase activity and is reported by ampicillin resistance. In this assay, the LD50 of the antibiotic served as an indicator of the interaction strength. Of note, bacteria can grow only in the presence of ampicillin when the β-lactamase is reconstituted in the periplasm. Therefore, the BLaTM assay also reports the insertion of the tested regions. With this system we double checked some of the previously studied interactions in a more quantitative manner. Overall, our findings point out an intricate network of interactions between the TMDs of viral and cellular origin. Finally, we wanted to determine if the TMD-TMD interaction network described played a role in apoptosis modulation. To that end, we transfected HeLa cells with Bcl2, HHV8, or MyxV either with or without the TMD (FL and ΔTMD, respectively). We also included chimeras in which the TMD of each of the previously described proteins was replaced by the TMD of the non-apoptotic mitochondrial protein TOMM20 (Bcl2-T20, HHV8-T20, and MyxV-T20, respectively), which our earlier studies suggested could not interact with any viral or cellular Bcl2 TMD. Cells were transfected with the appropriate constructs and then treated with doxorubicin to induce apoptosis (Rooswinkel et al., 2014). As expected, the FL proteins prevented apoptosis. When the TMD was removed, however, none of the proteins retained their anti-apoptotic effect. Similarly, the chimeras carrying the TMD of T20 could not control the doxorubicin-induced apoptosis, although localizing in the same membranes as the FL protein. These experiments were also replicated using different apoptotic stimulus as viral-induced apoptosis or Bax-induced apoptosis (only for Bcl2 and MyxV; FL and T20 variants). In summary, we have identified the Ct hydrophobic region of the vBcl2 as a genuine TMD capable of interacting with cBcl2 TMDs. We have also demonstrated that these intramembrane interactions are critical for viral cell fate control. This study provides a deeper understanding about how viruses control cellular death for their own benefit and contribute to our better understanding of how viruses interact with their hosts. In this chapter, we aimed to explore the intramembrane protein–protein interactions (PPIs) of BclxL protein and their role in its anti-apoptotic function. We used the information obtained for computationally design an inhibitor capable of selectively sequester the TMD of BclxL to turn back apoptosis resistance. As explained in section 2.3, the members of the Bcl2 protein family can interact with each other, forming homo- and hetero-oligomers (Andreu-Fernández et al., 2017; Cosentino and García-Sáez, 2017; Kelekar et al., 1997; Oltvai et al., 1993; Wang et al., 1996). These PPIs are part of an important regulatory network of mitochondrial outer membrane (MOM) permeabilization driving programmed cell death. In healthy cells, anti-apoptotic Bcl2 members inhibits activation of pro-apoptotic proteins through direct interaction or by sequestering BH3-only proteins (Kim et al., 2006). Upon an apoptotic stimulus, BH3-only and pro-apoptotic proteins are released and free to induce MOM permeabilization. Interactions among Bcl2 family members have been thought to occur through soluble domains, especially BH domains (Dadsena et al., 2021). However, recent findings demonstrate that their TMDs also participate in these PPIs (Andreu-Fernández et al., 2017; Lucendo et al., 2020) and, as proved in Chapter 3, these intramembrane interactions are crucial for apoptotic control. Among the anti-apoptotic proteins, the Bcl-2–like protein 1, better known as BclxL, displays relevant functions in several forms of cancer. In melanoma, BclxL participates preventing cells from executing apoptosis, and inducing drug resistance, cell migration and invasion, and angiogenesis (Lucianò et al., 2021). Because of the relevance of BclxL in the progression of cancer, different strategies have been considered to inhibit it (Lucianò et al., 2021) but new and less adverse BclxL inhibitors are needed. First, we sought to determine if the BclxL TMD can interact with other pro- and anti-apoptotic Bcl2 members. The potential intramembrane contacts were assessed using the genetic tool BLaTM (see section 3.3). Using this assay, we tested the homo-oligomerization of BclxL TMD and its putative hetero-oligomerization with the TMD of anti-apoptotic Bcl2, and the TMDs of pro-apoptotic (Bax and Bak) and BH3-only (Bik and Bmf) Bcl2 members. According to our findings, BclxL TMD forms weak homo-oligomers. Furthermore, we identified TM hetero-oligomers with Bcl2, Bax, and Bak. To corroborate the interaction capabilities of the BclxL TMD in eukaryotic cells, we used the BiFC assay (Kerppola, 2006), adapted for the study of intramembrane interactions (Andreu-Fernández et al., 2017; Grau et al., 2017) (see section 3.3). Our results indicated that the BclxL TMD can homo-oligomerize, and hetero-oligomerize with the TMDs of Bcl2, Bak, and Bik in eukaryotic membranes. Next, to investigate whether these TMD–TMD interactions are necessary for BclxL cellular apoptosis regulation, we transfected HeLa cells with full-length BclxL with or without the TMD (BclxL-FL and BclxL-ΔTMD, respectively). Additionally, we included a chimera in which the TMD of BclxL was replaced by the TMD of TOMM20 (BclxL-T20). Our results indicated that when doxorubicin is used as a cell death stimulus, BclxL requires the TMD to block apoptosis. As BclxL localizes primarily in the mitochondria (Fang et al., 1994; González-García et al., 1994; Zamzami et al., 1998), and given the importance of TMDs for membrane protein sorting (Martínez-Gil et al., 2011), we explored how deletions or substitutions in the Ct hydrophobic region of BclxL affected cellular localization. To analyze the subcellular location, BclxL (BclxL-FL), the BclxL-T20, and BclxL-ΔTMD variants were expressed in HeLa cells alongside a fluorescent mitochondrial marker. The fluorescence micrographs revealed that both BclxL-FL and BclxL-T20 moieties were located at the mitochondria while BclxL-ΔTMD, as expected, showed a cytosolic distribution. Once established that BclxL TMD–TMD interactions were crucial for the anti-apoptotic function, we aimed to design a specific inhibitor for these intramembrane PPIs. Inhibitor design started with the modeling of the BclxL TMD homo-interaction using TMHOP (Trans-membrane Homo Oligomer Predictor) (Weinstein et al., 2019). TMHOP uses Rosetta symmetric all-atom ab initio fold-and-dock simulations in an implicit membrane environment to predict thousands of low-energy conformations based on the energy function that relies on the empirical measurement of amino acid insertion propensities (Elazar et al., 2016; Weinstein et al., 2019). Based on structural characteristics and associated Rosetta energy, we selected a TMHOP model that forms a tightly packed parallel dimer. The selected model was then fed into the FuncLib design algorithm (Khersonsky et al., 2018) to generate higher affinity binders that could serve as inhibitors. The FuncLib algorithm uses Rosetta design calculations to enumerate combinations of tolerated amino acid substitutions at specific positions. It then relaxes each combination using whole-protein minimization (based on the Rossetta membrane energy function) (Weinstein et al., 2019) and ranks these combinations by energy. Since single-span TMDs are known to self-associate (Grau et al., 2017), we designed sequences with minimal self-association potential by including the following rules: i) we intend to design a sequence able to interact with the BclxL wt TMD. For that, we used positive selection for a heterodimer (Non-symmetric FuncLib; ΔΔG<+1 Rosetta energy units; R.e.u.). ii) we use negative selection for a new homodimer (symmetric FuncLib; ΔΔG>+5 R.e.u.). As a consequence, the algorithm will only include diversification in the sequence, which will allow the new sequences to efficiently interact with the native BclxL TMD sequence but will not allow the inhibitor to interact with itself. This process resulted in three designed potential TMD inhibitor sequences (named D1, D2, and D3). We verified that these designed segments could be inserted into ER-derived microsomes using an in vitro transcription/translation assay. Next, using BLaTM, we analyzed the interactions between the TMD of BclxL and the computationally designed inhibitors D1, D2, and D3. The results of these experiments revealed that D1 can efficiently bind to the BclxL TMD but does not form homo-oligomers, as we intended in our design. Of note, the interaction between the TMD of BclxL and D1 was stronger than the homo-oligomerization of the TMD of BclxL. Although D2 and D3 did not form homo-oligomers, they did not interact with the TMD of BclxL. Also, we investigated the specificity of the observed interaction by challenging D1 with the TMD of Bcl2, another anti-apoptotic protein. We detected no interaction between D1 and the Bcl2 TMD. Thus, any effect of D1 on cell survival would most likely arise from its interaction with the BclxL TMD. Additionally, we used the BiFC assay to ensure that the interaction between D1 and the BclxL TMD was maintained in eukaryotic membranes. The results indicated that D1 could efficiently bind to the TMD of BclxL and that did not form homo-oligomers in eukaryotic cells. To inhibit the anti-apoptotic effect of BclxL, the designed sequences must be in the same cellular compartment where BclxL is found. To test their location, we fused D1 and D2 sequences to the Ct of the enhanced green fluorescent protein eGFP (eGFP-D1 and eGFP-D2) and expressed these constructs in HeLa cells together with BclxL attached to the fluorescent protein mCherry (mCherry-BclxL). Next, we analyzed the subcellular distribution of these chimeras by confocal fluorescence microscopy. Both eGFP-D1 and eGFP-D2 showed a strong co-localization with mCherry-BclxL. To ensure that D1 and BclxL coexist in the same cellular compartment we performed a second localization assay based on organelle differential ultracentrifugation (Geladaki et al., 2019). We then analyzed the protein content in each fraction by sequential window acquisition of all theoretical mass spectra (SWATH‐MS) (Rotello and Veenstra, 2021; Zhang et al., 2020). BclxL and eGFP-D1 had a similar distribution profile, suggesting a similar subcellular localization. Finally, we tested the anti-apoptotic effect of D1 and D2. HeLa cells were transfected with BclxL alongside the eGFP-T20, eGFP-D1, eGFP-D2, or eGFP-xL chimeras. As a control, we used cells that did not receive BclxL or any of the chimeras and transfected them with an empty plasmid to keep the amount of transfected DNA constant across all samples. After transfection, cells were treated with doxorubicin to induce apoptosis. The cells that received eGFP-T20 or eGFP-xL plus BclxL could block doxorubicin-induced apoptosis. Remarkably, transfection of eGFP-D1 eliminated the anti-apoptotic effect of BclxL. D2 also reduced cell viability but less drastically than D1. Of note, no significant differences were found between the samples transfected with eGFP-D1 or eGFP-D2 (plus BclxL), or cells transfected with an empty plasmid treated in all cases with doxorubicin, indicating that D1 and D2 are capable of inhibiting BclxL function. We also tested cell viability after the transfection, proving that neither D1 nor D2 was toxic to HeLa cells, a vital characteristic when designing non-toxic inhibitors. In summary, these results have provided evidence of the importance of TMD–TMD interactions in apoptosis control, particularly in the case of BclxL. We successfully designed sequences capable of specifically inhibit the anti-apoptotic action of BclxL. Our work shows a path to design effective inhibitors based only on the sequences of the target receptor. The fact that two of three designs exhibited the desired hetero- and no homo-interactions highlights the accuracy of the TMHOP modelling strategy and the FuncLib design algorithm, which has already been applied to a wide range of soluble protein design tasks. This work significantly advances our understanding of sequence-specific recognition in membranes and opens the way for a new generation of anti-cancer drugs.