The High-Resolution Structure Reveals Remarkable Similarity in PD-1 Binding of Cemiplimab and Dostarlimab, the FDA-Approved Antibodies for Cancer Immunotherapy

Multiple tumors have responded well to immunotherapies, which use monoclonal antibodies to block the immune checkpoint proteins and reactivate the T-cell immune response to cancer cells. Significantly, the anti-PD-1 antibodies pembrolizumab and nivolumab, which were approved in 2014, have revolutionized cancer therapy, demonstrating dramatic improvement and longer duration. The US FDA authorized the third anti-PD-1 medication, cemiplimab, in 2018 for use in patients with cutaneous squamous cell carcinoma. To further understand the molecular mechanism of the antibody drug, we now reveal the intricate structure of PD-1 in complex with the cemiplimab Fab at a resolution of 1.98 Å. The cemiplimab–PD-1 interaction preoccupies the space for PD-L1 binding with a greater binding affinity than the PD-1/PD-L1 interaction, which is the basis for the PD-1 blocking mechanism. The structure reveals that cemiplimab and dostarlimab are significantly similar in PD-1 binding, although the precise interactions differ. A comparative investigation of PD-1 interactions with the four FDA-approved antibodies reveals that the BC, C’D, and FG loops of PD-1 adopt distinct conformations for optimal interaction with the antibodies. The structural characteristics in this work could be helpful information for developing more potent anti-PD-1 biologics against cancer.


Introduction
Antibody-mediated blockade of immune checkpoint proteins has been unprecedentedly successful in treating multiple tumors [1]. Programmed cell death-1 (PD-1), a coinhibitory receptor within the immunological synapse, is mainly expressed in immune cells such as T cells, B cells, dendritic cells, and NK cells [2]. PD-1 is a crucial immune checkpoint for maintaining immune tolerance and exhaustion of T cells in a tumor, restricting active immune responses against cancer cells [3]. The ligands of PD-1, PD-L1 and PD-L2, are highly expressed in tumor cells, enabling the tumor cells to escape from T-cell immune responses by interacting with PD-1 [4]. Inhibition of the interaction of PD-1 with its ligands using monoclonal antibodies can restore T-cell immune responses against cancer cells [5]. The US Food and Drug Administration (FDA) has approved eight monoclonal antibodies against immunological checkpoint proteins, including PD-1, PD-L1, and cytotoxic T-lymphocyte protein 4 (CTLA-4), with four approvals for anti-PD-1, three for anti-PD-L1, and one for anti-CTLA-4. Unfortunately, the overall response rate of these immune checkpoint inhibitors' monotherapies is insufficient, even though they have significantly improved patient outcomes and prolonged patient survival for certain forms of cancer. In order to increase the therapeutic efficiency and overall response rate, various immune checkpoint molecules, such as the lymphocyte activation gene 3 protein (LAG3), T-cell immunoreceptor with Ig and ITIM domains (TIGIT), and T-cell immunoglobulin The details for the expression and purification of the PD-1 protein and cemiplimab Fab are described in the supplementary materials. In short, the DNA sequence encoding the extracellular domain of human PD-1 (aa 26-150) was subcloned into pET-21a. The protein was expressed in E. coli BL21 (DE3) cells as inclusion bodies and refolded by dialysis. The protein was purified by Ni-affinity and gel filtration chromatography. A modified pBAD vector was used for subcloning the DNA sequences for the heavy and light chains of the cemiplimab Fab, which are synthesized after codon optimization for E. coli expression. Ni-affinity and gel filtration chromatography were used to purify the cemiplimab Fab, which was expressed in the periplasm of E. coli.

Crystallization of the Complex
Cemiplimab's Fab fragment was mixed with PD-1 protein at a molar ratio of 1: 1.2, and the mixture was incubated for 30 min on ice. The mixture was loaded using a buffer containing 20 mM Tris, pH 7.5, and 200 mM NaCl on a HiLoad 26/600 Superdex 200 pg column (Cytiva, Marlborough, MA, USA) for gel filtration chromatography. The purified PD-1/cemiplimab complex was concentrated to 10 mg mL −1 and crystallized by hangingdrop vapor diffusion with a well solution containing 0.1 M sodium HEPES pH 7.5 and 5% (w/v) PEG 3350 at 20 • C. Before the X-ray diffraction experiment, crystals were soaked into the crystallization reservoir solution plus 20% (v/v) glycerol and quickly frozen in liquid nitrogen.

Data Collection and Structure Determination
X-ray diffraction data of the crystal of PD-1/cemiplimab complex were acquired at Pohang Accelerator Laboratory (PAL) (Pohang, Korea) beamline 5C and automatically processed and scaled using XDS to a resolution of 1.98 Å [27]. The structures of PD-1 and the variable and constant regions of the Fab fragment in the complex structure of PD-1/pembrolizumab Fab (PDB code 5GGS) were used as search models for molecular replacement (MR) using Phaser software [20,28]. Model building and structure refinement were carried out using PHENIX and Coot [29,30]. Statistics for data collection and structure determination are presented in Table 1. All structural figures were produced using PyMol (Schrödinger, Inc., New York, NY, USA). The atomic coordinates and structure factors for the complex structure of PD-1/cemiplimab were deposited in a Protein Data Bank (http://www.rcsb.org (deposited: 21 September 2022) under the accession code 8GY5. protein was expressed in E. coli BL21 (DE3) cells as inclusion bodies and refolded by dialysis. The protein was purified by Ni-affinity and gel filtration chromatography. A modified pBAD vector was used for subcloning the DNA sequences for the heavy and light chains of the cemiplimab Fab, which are synthesized after codon optimization for E. coli expression. Ni-affinity and gel filtration chromatography were used to purify the cemiplimab Fab, which was expressed in the periplasm of E. coli.

Crystallization of the Complex
Cemiplimab's Fab fragment was mixed with PD-1 protein at a molar ratio of 1: 1.2, and the mixture was incubated for 30 min on ice. The mixture was loaded using a buffer containing 20 mM Tris, pH 7.5, and 200 mM NaCl on a HiLoad 26/600 Superdex 200 pg column (Cytiva, Marlborough, MA, USA) for gel filtration chromatography. The purified PD-1/cemiplimab complex was concentrated to 10 mg mL −1 and crystallized by hangingdrop vapor diffusion with a well solution containing 0.1 M sodium HEPES pH 7.5 and 5% (w/v) PEG 3350 at 20 °C. Before the X-ray diffraction experiment, crystals were soaked into the crystallization reservoir solution plus 20% (v/v) glycerol and quickly frozen in liquid nitrogen.

Data Collection and Structure Determination
X-ray diffraction data of the crystal of PD-1/cemiplimab complex were acquired at Pohang Accelerator Laboratory (PAL) (Pohang, Korea) beamline 5C and automatically processed and scaled using XDS to a resolution of 1.98 Å [27]. The structures of PD-1 and the variable and constant regions of the Fab fragment in the complex structure of PD-1/pembrolizumab Fab (PDB code 5GGS) were used as search models for molecular replacement (MR) using Phaser software [20,28]. Model building and structure refinement were carried out using PHENIX and Coot [29,30]. Statistics for data collection and structure determination are presented in Table 1. All structural figures were produced using PyMol (Schrödinger, Inc., New York, NY, USA). The atomic coordinates and structure factors for the complex structure of PD-1/cemiplimab were deposited in a Protein Data Bank (http://www.rcsb.org (deposited: 2022-09-21) under the accession code 8GY5. protein was expressed in E. coli BL21 (DE3) cells as inclusion bodies and refolded by dialysis. The protein was purified by Ni-affinity and gel filtration chromatography. A modified pBAD vector was used for subcloning the DNA sequences for the heavy and light chains of the cemiplimab Fab, which are synthesized after codon optimization for E. coli expression. Ni-affinity and gel filtration chromatography were used to purify the cemiplimab Fab, which was expressed in the periplasm of E. coli.

Crystallization of the Complex
Cemiplimab's Fab fragment was mixed with PD-1 protein at a molar ratio of 1: 1.2, and the mixture was incubated for 30 min on ice. The mixture was loaded using a buffer containing 20 mM Tris, pH 7.5, and 200 mM NaCl on a HiLoad 26/600 Superdex 200 pg column (Cytiva, Marlborough, MA, USA) for gel filtration chromatography. The purified PD-1/cemiplimab complex was concentrated to 10 mg mL −1 and crystallized by hangingdrop vapor diffusion with a well solution containing 0.1 M sodium HEPES pH 7.5 and 5% (w/v) PEG 3350 at 20 °C. Before the X-ray diffraction experiment, crystals were soaked into the crystallization reservoir solution plus 20% (v/v) glycerol and quickly frozen in liquid nitrogen.

Data Collection and Structure Determination
X-ray diffraction data of the crystal of PD-1/cemiplimab complex were acquired at Pohang Accelerator Laboratory (PAL) (Pohang, Korea) beamline 5C and automatically processed and scaled using XDS to a resolution of 1.98 Å [27]. The structures of PD-1 and the variable and constant regions of the Fab fragment in the complex structure of PD-1/pembrolizumab Fab (PDB code 5GGS) were used as search models for molecular replacement (MR) using Phaser software [20,28]. Model building and structure refinement were carried out using PHENIX and Coot [29,30]. Statistics for data collection and structure determination are presented in Table 1. All structural figures were produced using PyMol (Schrödinger, Inc., New York, NY, USA). The atomic coordinates and structure factors for the complex structure of PD-1/cemiplimab were deposited in a Protein Data Bank (http://www.rcsb.org (deposited: 2022-09-21) under the accession code 8GY5.

Crystal Structure of PD-1 in Complex with Cemiplimab
The ectodomain of PD-1 was expressed as inclusion bodies in E. coli, and its soluble form was produced using a refolding procedure. A cemiplimab Fab fragment was expressed in the periplasm of E. coli. The 1:1 complex of PD-1 and cemiplimab Fab was present in the solution as a monomer, according to size exclusion chromatography (Supplementary Figure S1). The crystal structure of PD-1 in complex with the Fab fragment of cemiplimab was determined with a resolution of 1.98 Å and R/R free of 0.187/0.235. The crystal's space group is P1, and an asymmetric unit contains two copies of the complex with a pseudo-translational symmetric relationship ( Figure 1A). Superposition of the structures of PD-1 from the two copies in an asymmetric unit showed that the PD-1 protein and all the six complementarity determining regions (CDRs) of cemiplimab displayed minimal structural differences from one another, with clear electron density on almost all of the complex's residues ( Figure 1B,D). One copy of the complex had a somewhat disordered PD-1 C'D loop, implying that its contribution to cemiplimab binding was likely to be minor ( Figure 1B). Due to the inherent flexibility of the elbow region within an antibody, the constant region of the bound Fab was slightly skewed ( Figure 1B). Cemiplimab's PD-1 binding was mediated by five complementarity determining regions (CDRs), leaving LCDR2 uninvolved ( Figure 1C). When the complex structure of PD-1 and cemiplimab was overlaid onto the complex structure of PD-1 and PD-L1 with the shared fixed PD-1 ( Figure 1E), the underlying mechanism by which cemiplimab blocks the PD-1 from the binding of PD-L1 became clear. The variable region of cemiplimab's light chain sterically collided with PD-L1 prior to establishing the PD-1/PD-L1 complex when cemiplimab binds to PD-1.

Specific Cemiplimab-PD-1 Interactions
The high-resolution structure of the PD-1/cemiplimab complex made it possible to identify this antibody's precise epitope and paratope. Although the light chain of cemiplimab contributes mostly to the steric blockage of PD-L1 binding ( Figure 1E), a buried solvent-accessible area of 1727 Å 2 within the binding interface is primarily caused by interactions with the heavy chain (71%). The cemiplimab binding involves 16 residues of PD-1 for the interaction via hydrogen bonds, a salt bridge, and van der Waals interactions ( Figure 2). The cemiplimab residues heavy S52, heavy R56, heavy D57, heavy Y59, light S92, and light T94 make hydrogen bonds with the residues of PD-1, particularly PD-1 E61 and PD-1 S62 of the BC loop and PD-1 P130 and PD-1 A132 of the FG loop ( Figure 2A). Furthermore, the side chain of the PD-1 residue PD-1 E61 also creates a salt bridge with the side chain of heavy R56 in cemiplimab heavy chain, implying its crucial contribution to the binding energy of the complex (Figure 2A). The side chains of the cemiplimab residues, including heavy T28, heavy N31, heavy Y32, heavy Y59, heavy W99, heavy N101, heavy I102, light F32, light N93, light T94, and light F96, make extensive van der Waals contacts with the side chains of the PD-1 residues, including PD-1 S60 and PD-1 S62 within the BC loop, PD-1 P83 and PD-1 R86 within the C'D loop, PD-1 L128, PD-1 A129, PD-1 P130, PD-1 K131 and PD-1 A132 within the FG loop, and PD-1 V64, PD-1 N66, PD-1 F82, PD-1 I126, PD-1 Q133, and PD-1 I134 within the front β-sheet consisting of the CC'FG strands. Thanks to the high resolution of the complex structure, the apparent electron density could visualize the water molecules mediating the hydrogen bonds between the residues of PD-1 and cemiplimab ( Figure 2B). The residues of PD-1, including PD-1 E61, PD-1 F63, PD-1 N66, PD-1 E84, PD-1 L128, and PD-1 A132, establish hydrogen bonds via water molecules with heavy G53, heavy G100, heavy N101, heavy T28, light S91, and light S92 of cemiplimab, respectively. A surface plasmon resonance (SPR) study revealed that cemiplimab's binding affinity (K D ) to PD-1 is 0.60 nM, whereas the PD-L1 binds to PD-1 with a binding affinity of 8.2 µM [19,31,32]. Taken together, all of the interactions mentioned above would support cemiplimab's strong affinity to outcompete PD-L1 binding to PD-1.

Remarkably Similar Binding of Cemiplimab and Dostarlimab
When the complex structures of PD-1/cemiplimab and PD-1/dostarlimab were superimposed (PDB code 7WSL) [24], it was discovered that their binding modes and epitopes are very similar ( Figure 3B). Even though the detailed interactions are somewhat different from each other, the loop conformations of the HCDR1 and HCDR2 remain conserved, most likely as a result of the amino acid sequence similarities within the regions of the two antibodies ( Figure 3A). The side chains of heavy Y32 and heavy D33 within the dostarlimab HCDR1 form hydrogen bonds with PD-1. In contrast, cemiplimab's corresponding residues, heavy F32 and heavy G33, are unable to make any hydrogen bonds (Figures 2A and 3D). With the PD-1 E61 of PD-1, the side chain of heavy R56 in the cemiplimab HCDR2 creates a salt bridge, whereas the corresponding residue heavy S56 of dostarlimab forms a hydrogen bond. The side chain of heavy D57 in cemiplimab HCDR2 forms a hydrogen bond with the backbone nitrogen atom of PD-1 E61 in PD-1, unlike the analogous residue heavy Y57 in dostarlimab. In particular, the HCDR3 was entirely different from one another, with cemiplimab's HCDR3 being one residue longer than dostarlimab's ( Figure 3A,D). While the residue light W50 within the dostarlimab LCDR2 aids in the hydrophobic contact with PD-1's FG loop area, cemiplimab's equivalent residue, light A50, is significantly smaller than tryptophan and makes little contribution to the interaction ( Figure 3E). Instead, the residue light F32 within the cemiplimab LCDR1 is important in the interaction with the FG loop, with the dostarlimab residue light A32 making a negligible contribution (Figures 2A and 3E). The LCDR3 of both antibodies also participate, albeit in distinct ways, in the interaction with the FG loop. In Biomedicines 2022, 10, 3154 7 of 13 cemiplimab, hydrogen bonds play a significant role in mediating the interaction between the LCDR3 and the FG loop ( Figure 3E). In comparison, the dostarlimab LCDR3 interacts with the FG loop through hydrophobic contacts with the residues light Y91 and light Y94 [24].

Comprehensive Comparison of the PD-1 Binding by the Antibodies
Previous studies have reported the complex structures of PD-1 with the FDA-approved antibodies, nivolumab, pembrolizumab, and dostarlimab, describing the structural basis of PD-1 blockades by antibody drugs [20][21][22][23][24]. The molecular mechanism underlying the clinical efficacy of the PD-1 antibodies can be better understood by comparing the binding interactions of PD-1 with PD-L1 and the antibodies in extensive detail. The antibodies and PD-L1 bind to the immunoglobulin-like extracellular domain of PD-1 from various directions and on distinct binding sites when the PD-1 structures extracted from the complex structures with PD-L1, cemiplimab, pembrolizumab nivolumab, and dostarlimab were superimposed (Figure 4). The binding of PD-1 to PD-L1 is mediated by the interaction involving the residues within the front β-sheet of the ectodomain of PD-1, while the antibody drugs predominantly interact with the loop regions of PD-1 protein [23,29]. Although the area of the binding interface between PD-1 and PD-L1 is comparable to the interfaces of antibody-binding, the antibody drugs bind to PD-1 with considerably increased binding affinity than PD-L1, implying the loop regions would be hotspots on PD-1 for effective interaction between PD-1 and the antibody drugs. Although there is no head-to-head comparison of the antibodies' effect in vitro, the binding affinities (Kd) of the antibodies to PD-1 vary; 27 pM for pembrolizumab, 4.06 nM for nivolumab, 0.6 nM for cemiplimab, and 0.3 nM for dostarlimab [17,19,21,22].The loops of the PD-1 ectodomain adopt distinct structures when they engage to various binding partners, such as PD-L1 and the antibody drugs, implying that they are intrinsically flexible and their optimal conformations are induced to increase binding affinity ( Figure 5). The binding of nivolumab is largely dependent on its ability to interact with the N-terminal region of PD-1, whereas pembrolizumab and dostarlimab bind primarily to the C'D loop of distinct conformation [20][21][22][23][24]. The nivolumab-bound form of PD-1 was the only crystal structure in which the N-terminal extension of PD-1 was visible, suggesting its inherent flexibility. The C'D loop encroaches upon a groove created by the paratope of pembrolizumab, and the majority of the loop's residues participates in the interaction with the drug. The C'D loop would also be the primary contributor to the dostarlimab-binding of PD-1 because the residue R86 in the C'D loop of PD-1 occupies a concavity within the paratope of dostarlimab through hydrogen bonds, ionic interactions, and pi-cation interactions. Additionally, the BC and FG loops of PD-1 also undergo conformational changes upon binding to these antibodies, providing extra interactions with them [20,24]. Additionally, these loops' specific conformations are incompatible with PD-L1 binding because they participate in the PD-L1 interaction in distinct conformations. Even though cemiplimab and dostarlimab similarly bind to PD-1, the differences in the specific interaction mentioned above caused the loops' conformation to vary for the best interaction with the two antibodies. The binding of cemiplimab to PD-1 also involves the interaction with the flexible BC, C'D, and FG loops of PD-1. Given that the PD-1 residue E61 is thought to play a significant role in the interaction with cemiplimab through a salt bridge and multiple hydrogen bonds, the BC loop of PD-1 would be the dominant hotspot for binding cemiplimab.

Discussion
Blocking the interaction between PD-1 and PD-L1 by monoclonal antibodies removes the ability of tumors to evade the immune system, revolutionizing cancer immunotherapy. Among the FDA-approved antibodies, the most redundant therapeutic approach is blocking the PD-1/PD-L1 interaction, with four anti-PD-1 antibodies, including pembrolizumab, nivolumab, cemiplimab, and dostarlimab, and three anti-PD-L1 antibodies, including atezolizumab, avelumab, and durvalumab [33].
In this study, we describe the high-resolution crystal structure of the PD-1/cemiplimab complex. Overall, the binding of cemiplimab is very distinct from pembrolizumab and nivolumab but strikingly resembles that of dostarlimab. Cemiplimab is the third anti-PD-1 antibody approved by the US FDA for the treatment of metastatic cutaneous squamous cell carcinoma in 2018. In 2021, the FDA approved this antibody as the first immunotherapy for locally advanced basal cell carcinoma. In addition, ceimplimab monotherapy was also approved as a first-line treatment for patients of advanced non-small cell lung cancer with high PD-L1 expression. The fourth PD-1 monoclonal antibody, dostarlimab was approved by the FDA for the treatment of mismatch repair deficient (dMMR) endometrial cancer and dMMR solid tumors in 2021. Our study reveals that the binding modes of cemiplimab and dostarlimab are strikingly similar except for some details. Besides, these two antibodies are both IgG4 kappa antibodies, expecting their negligible effector functions, including ADCC and CDC activity. However, their approved indications are not redundant, so additional approvals of their indications and combination therapies with these antibodies are anticipated. Recently, relatlimab, an anti-LAG-3 antibody, was approved to treat melanoma in combination with nivolumab [34]. This approval is especially notable since it validates a completely novel combination of two immunotherapies that may work in tandem to enhance anti-tumor response by targeting LAG-3 and PD-1, two separate immune checkpoints. Cemiplimab is also under clinical

Discussion
Blocking the interaction between PD-1 and PD-L1 by monoclonal antibodies removes the ability of tumors to evade the immune system, revolutionizing cancer immunotherapy. Among the FDA-approved antibodies, the most redundant therapeutic approach is blocking the PD-1/PD-L1 interaction, with four anti-PD-1 antibodies, including pembrolizumab, nivolumab, cemiplimab, and dostarlimab, and three anti-PD-L1 antibodies, including atezolizumab, avelumab, and durvalumab [33].
In this study, we describe the high-resolution crystal structure of the PD-1/cemiplimab complex. Overall, the binding of cemiplimab is very distinct from pembrolizumab and nivolumab but strikingly resembles that of dostarlimab. Cemiplimab is the third anti-PD-1 antibody approved by the US FDA for the treatment of metastatic cutaneous squamous cell carcinoma in 2018. In 2021, the FDA approved this antibody as the first immunotherapy for locally advanced basal cell carcinoma. In addition, ceimplimab monotherapy was also approved as a first-line treatment for patients of advanced non-small cell lung cancer with high PD-L1 expression. The fourth PD-1 monoclonal antibody, dostarlimab was approved by the FDA for the treatment of mismatch repair deficient (dMMR) endometrial cancer and dMMR solid tumors in 2021. Our study reveals that the binding modes of cemiplimab and dostarlimab are strikingly similar except for some details. Besides, these two antibodies are both IgG4 kappa antibodies, expecting their negligible effector functions, including ADCC and CDC activity. However, their approved indications are not redundant, so additional approvals of their indications and combination therapies with these antibodies are anticipated. Recently, relatlimab, an anti-LAG-3 antibody, was approved to treat melanoma in combination with nivolumab [34]. This approval is especially notable since it validates a completely novel combination of two immunotherapies that may work in tandem to enhance anti-tumor response by targeting LAG-3 and PD-1, two separate immune checkpoints. Cemiplimab is also under clinical investigation in combination with a LAG-3 blocking antibody fianlimab, demonstrating a greater than 60% response rate in melanoma patients [35]. To date, no structural information for the antibody-mediated blockade of LAG-3 is available. Further structural studies of LAG-3 antibodies can provide more details on the molecular mechanism of the therapeutic synergy of the dual blockade of PD-1 and LAG-3.
Due to their exceptional specificity and affinity to their targets, monoclonal antibodies have become a major therapeutic tool. Different epitopes might result in varying therapeutic efficacies for antibodies, so the epitope inside a target molecule can be a key component of a therapeutic antibody. The authorized antibodies against PD-1 and PD-L1 detect different antigenic epitopes from other antibodies, even though they share a similar blocking mechanism [25]. In this study, cemiplimab has a different binding mode from other antibodies and mainly interacts with the flexible loops of PD-1, including the BC, C'D, and FG loops. These structural investigations into the hotspots on a target protein where antibodies bind may help develop more effective biologics against PD-1 for cancer immunotherapy. The observation and study of off-target effects caused by antibodies on huge cohorts have revealed that they are not as specific as previously believed, raising new concerns about the polyreactivity and polyspecificity of antibodies [36,37]. Accumulating structural information for the recognition of various targets by antibodies could also contribute to enhancing the target specificity of therapeutic antibodies.

Conclusions
In conclusion, the high-resolution structure of PD-1 in complex with the Fab fragment of cemiplimab sheds light on the molecular mechanism underlying the therapeutic activity of this antibody drug. Additionally, the specific epitope and interactions revealed by this structural study may provide significant information for the improvement of the current biological agents against PD-1 for cancer immune checkpoint therapy. Funding: This research was funded by a grant from the National Research Foundation of Korea (NRF-2020R1A2C2008865 and NRF-2022R1F1A1073879). This work was supported by MSIT and PAL, Korea. This paper was written as part of Konkuk University's research support program for its faculty on sabbatical leave in 2020.

Informed Consent Statement: Not applicable.
Data Availability Statement: The atomic coordinates and structure factors for the complex structure of PD-1/cemiplimab were deposited in Protein Data Bank (http://www.rcsb.org (deposited: 21 September 2022)) under the accession code 8GY5.