WO2017040378A1 - Discovery of enzymes from the alpha-keto acid decarboxylase family - Google Patents

Abstract

2-ketoacid decarboxylase enzymes, compositions encoding for 2 ketoacid decarboxylase enzymes, and host cells comprising such enzymes or compositions are provided

Description

DISCOVERY OF ENZYMES FROM THE ALPHA-KETO ACID

DECARBOXYLASE FAMILY

CROSS-REFERENCES TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 62/21 1 ,439, filed August 28, 2015, the contents of which are hereby incorporated in the entirety for all purposes.

BACKGROUND OF THE INVENTION

[0002] A central goal of the bioeconomy is to reduce our dependence on petroleum through next-generation biomanufacturing. The USD A projects the industrial production of bio-based specialty chemicals to reach ~$340 billion USD by 2025, replacing half of our needs for these products from petroleum. To accomplish this ambitious goal, non-natural biochemical pathways are needed. An example of one such pathway is the "synthetic recursive +1" carbon elongation pathway that has been developed to produce a variety of alcohol products in E. coli. The primary product from the current pathway is 1-butanol, with longer chain alcohols {e.g., pentanol, hexanol, heptanol, and octanol) being either minor products of the pathway, or not produced at all. Yet, there is significant interest in producing long chain alcohols given their use as specialty chemicals, as well as their greater energy density when used as liquid fuels or fuel additives. The present invention fulfills these and other needs.

BRIEF SUMMARY OF THE INVENTION

[0003] In one aspect, the present invention provides an expression cassette comprising a heterologous promoter operably linked to a nucleic acid encoding an enzyme comprising 2- ketoacid decarboxylase activity, wherein the enzyme: comprises a 3 -layer alpha-beta-alpha sandwich; comprises a thiamine pyrophosphate (TPP) binding pocket; has a TM align score of at least about 0.5 {e.g., from about 0.5 to about 1.0, from about 0.5 to about 0.8, or from about 0.5 to about 0.9) as compared against PDB ID:2VBG; and comprises at least about a 10-fold, 25- fold, 50-fold, or 100-fold {e.g., from about 10-fold to about 100,000-fold; from about 10-fold to about 10,000 fold; from about 10-fold to about 200-fold; from about 10-fold to about 100-fold; from about 25-fold to about 100,000-fold; from about 25-fold to about 10,000 fold; from about 25-fold to about 200-fold; from about 25-fold to about 100-fold; from about 50-fold to about 100,000-fold; from about 50-fold to about 10,000 fold; from about 50-fold to about 200-fold; or from about 50-fold to about 100-fold) greater catalytic efficiency for C8 2-ketoacid substrates as compared to C3 or isoC5 2-ketoacid substrates, wherein the catalytic efficiency is measured by kcat/Km.

[0004] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity has a TM align score of at least 0.5 (e.g., from about 0.5 to about 0.8, from about 0.5 to about 0.9, or from about 0.5 to about 1.0), 0.6 (e.g., from about 0.6 to about 0.8, from about 0.6 to about 0.9, or from about 0.6 to about 1.0), 0.7 (e.g., from about 0.7 to about 0.8, from about 0.7 to about 0.9, or from about 0.7 to about 1.0), 0.8 (e.g., from about 0.8 to about 0.9, or from about 0.8 to about 1.0), or 0.9 (e.g., from about 0.9 to about 1.0) as compared against PDB E>:2VBG. In some embodiments, the substrate binding pocket of the enzyme has a solvent accessible surface area of, of about, of at least, or of at least about 5,000; 5,500; 6,000; 6,500; 7,000; 7,500; 8,000; or 8,500; 9,000; 9,500; or 10,000 A² (e.g., from about 5,000 to about 10,000 A²; from about 5,500 to about 9,500 A²; from about 6,000 to about 9,000 A²; from about 6,500 to about 8,500 A²; from about 7,000 to about 8,000 A²; or from about 7,500 to about 8,000 A²).

[0005] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that is, is at least, or is at least about 14%, 15%, 16%, 17%, 18%, 19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%, 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, or 89% identical to SEQ ID NO:l. In some

embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that is less than, or less than about, 90%, 95%, or 99% identical to SEQ ID NO: 1. In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that is from about 14% to about 89% identical to SEQ ID NO:l; from about 20% to about 89% identical to SEQ ID NO:l; from about 25% to about 89% identical to SEQ ID NO: 1 ; from about 30% to about 89% identical to SEQ ID NO: 1 ; from about 35% to about 89% identical to SEQ ID NO: 1 ; from about 40% to about 89% identical to SEQ ID NO: 1 ; from about 45% to about 89% identical to SEQ ID NO:l; from about 50% to about 89% identical to SEQ ID NO: 1; from about 55% to about 89% identical to SEQ ID NO: 1; from about 60% to about 89% identical to SEQ ID NO: 1 ; from about 65% to about 89% identical to SEQ ID NO: 1 ; from about 70% to about 89% identical to SEQ ID NO: 1 ; from about 75% to about 89% identical to SEQ ID NO: 1 ; or from about 80% to about 89% identical to SEQ ID NO: 1.

[0006] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation of a native sequence wherein the mutation is in an amino acid residue within a 2-ketoacid decarboxylase active site or in an amino acid residue having a Ca within 8 A (e.g., having a Ca atom within about 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, or 7.5 A) of a 2-ketoacid decarboxylase active site, wherein the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23- 27, 50, 51, 56, 75, 76, 79, 82, 113, 114, 287, 352, 373-377, 379, 394, 396-399, 424-428, 430, 450-458, 461, and 532 of PDB K>:2VBG.

[0007] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation of a native sequence wherein the mutation is in an amino acid residue within a 2-ketoacid decarboxylase active site or is in an amino acid residue within 8 A (e.g., having a non-hydrogen atom within about 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, or 7.5 A) of a 2-ketoacid decarboxylase active site, wherein the 2- ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 50, 51, 56, 71, 75, 76, 79, 82, 112, 114, 287, 356, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461, and 535 of SEQ ID NO:l.

[0008] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that is at least 90%, 95%, or 99% identical to SEQ ID NO:l (e.g., from 90% to about 99% or more identical to SEQ ID NO: 1, or from 95% to about 99% or more identical to SEQ ID NO: 1). In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains 1, 2, 3, 4, or 5 mutations (e.g., substitutions) as compared to SEQ ID NO:l, 2, or 3.

[0009] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains 1, 2, or 3 mutations (e.g., substitutions) in amino acid residues corresponding to a residue selected from the group consisting of G402, MS38, and F542 of SEQ ID NO: l.

[0010] In some embodiments, the mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1 is from a (e.g., native) sequence to a residue selected from the group consisting of Val, Leu, He, Met, Phe, His, Gly, Ala, Thr, Tyr, and Trp.

[0011] In some embodiments, the mutation corresponds to, or is, a mutation selected from the group consisting of G402V, M538L, and F542V of SEQ ID NO:l.

[0012] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises two mutations (e.g., substitutions) in amino acid residues corresponding to residues selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1. In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises mutations (e.g., substitutions) corresponding to G402V, M538L, and F542V of SEQ ID NO: 1 or comprises SEQ ID NO: 3.

[0013] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence at least 90% or 99% identical (e.g., from about 90% to about 99%), or identical, to SEQ ID NO:2.

[0014] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation, wherein the mutation is in an amino acid residue of a 2-ketoacid decarboxylase active site residue of SEQ ID NO:2 or in an amino acid residue having a Ca within 8 A (e.g., having a non-hydrogen atom within about 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, or 7.5 A)of the 2-ketoacid decarboxylase active site, wherein the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 48, 49, 54, 71, 74, 75, 78, 81, 112, 113, 286, 356, 376-380, 382, 399, 401-404, 427-432, 434, 454-462, 465, and 535 of SEQ ID NO:2.

[0015] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity interacts with a C8 2-ketoacid substrate with a calculated interaction energy of less than -5, -5.5, -6, -6.5, -7, -7.5, -8, -8.5, -9, -9.5, -10, -10.3, -10.4, -10.5, -10.6, -10.7, -10.8, -10.9, -11, -11.5, - 12, -12.5, -13, -13.5, -14, -14.5, -15, -15.5, -16, -16.5, -17, -17.5, -18, -18.5, -19, -19.5, or -20 rosetta energy units. In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity interacts with a C8 2-ketoacid substrate with a calculated interaction energy of from about -20 to about -5 rosetta energy units, from about -15 to about -8 rosetta energy units; or from about -11 to about -9 rosetta energy units.

[0016] In another aspect, the present invention provides a modified enzyme comprising 2- ketoacid decarboxylase activity, wherein the enzyme comprises: a 3 -layer alpha-beta-alpha sandwich; thiamine pyrophosphate; and at least a 10-fold, 25-fold, 50-fold, or 100-fold (e.g., from about 10-fold to about 100,000-fold; from about 10-fold to about 10,000 fold; from about 10-fold to about 200-fold; from about 10-fold to about 100-fold; from about 25-fold to about 100,000-fold; from about 25-fold to about 10,000 fold; from about 25-fold to about 200-fold; from about 25-fold to about 100-fold; from about 50-fold to about 100,000-fold; from about 50- fold to about 10,000 fold; from about 50-fold to about 200-fold; or from about 50-fold to about 100-fold) greater catalytic efficiency for C8 2-ketoacid substrates as compared to C3 or isoC5 2- ketoacid substrates, wherein the catalytic efficiency is measured by kcat/Km, wherein the modification comprises a mutation (e.g., substitution) in the primary amino acid sequence of SEQ ID NO: 1 or SEQ ID NO:2.

[0017] In some embodiments, the enzyme has a TM align score against PDB ID:2VBG of at least 0.5 (e.g., from about 0.5 to about 0.8, from about 0.5 to about 0.9, or from about 0.5 to about 1.0), 0.6 (e.g., from about 0.6 to about 0.8, from about 0.6 to about 0.9, or from about 0.6 to about 1.0), 0.7 (e.g., from about 0.7 to about 0.8, from about 0.7 to about 0.9, or from about 0.7 to about 1.0), 0.8 (e.g., from about 0.8 to about 0.9, or from about 0.8 to about 1.0), or 0.9 (e.g., from about 0.9 to about 1.0). In some embodiments, the enzyme has a Ca root mean squared deviation of less than 1 A (e.g., from about 0.1 A to about 1 A, from about 0.2 A to about 1 A, from about 0.3 A to about 1 A, from about 0.4 A to about 1 A, or from about 0.5 A to about 1 A) over at least 100 (e.g., from about 100 to about 200 or more) structurally aligned residues as compared against PDB ID:2VBG.

[0018] In some embodiments, the modification comprises a mutation (e.g., substitution) in an amino acid residue, wherein the amino acid residue is within a 2-ketoacid decarboxylase active site or within 8 A (e.g., having a Ca atom within about 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, or 7.5 A) of a 2-ketoacid decarboxylase active site, wherein the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, SO, SI, 56, 71, 75, 76, 79, 82, 112, 114, 287, 356, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461, and 535 of SEQ ID NO:l.

[0019] In some embodiments, the modification comprises a mutation (e.g., substitution) in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1. In some embodiments, the mutation (e.g., substitution) in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1 is a mutation from a (e.g., native) sequence to a residue selected from the group consisting of Val, Leu, IIe, Met, Phe, His, Gly, Ala, Thr, Tyr, and Trp.

[0020] In some embodiments, the modification comprises two mutations (e.g., substitutions) in amino acid residues corresponding to residues selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1. In some embodiments,the modification comprises one or two mutations (e.g., substitutions) corresponding to the mutations selected from the group consisting of G402V, M538L, and F542V of SEQ ID NO: 1. In some embodiments, the modification comprises mutations (e.g., substitutions) corresponding to G402V, M538L, and F542V of SEQ ID NO: 1. In some embodiments, the ratio of kcat/Km against C8 2-ketoacid substrates over the kcat/Km against isoC5 2-ketoacid substrates is at least 10-fold, 25-fold, 50-fold, or 100-fold (e.g., from about 10-fold to about 100,000-fold; from about 10-fold to about 10,000 fold; from about 10-fold to about 200-fold; from about 10-fold to about 100-fold; from about 10-fold to about 50-fold; from about 25-fold to about 100,000-fold; from about 25-fold to about 10,000 fold; from about 25-fold to about 200-fold; from about 25-fold to about 100-fold; from about 25- fold to about 50-fold; from about 50-fold to about 100,000-fold; from about 50-fold to about 10,000 fold; from about 50-fold to about 200-fold; or from about 50-fold to about 100-fold) greater than the native enzyme. In some embodiments, the solvent accessible surface area of the substrate binding pocket is, is about, is at least, or is at least about 5,000; 5,500; 6,000; 6,500; 7,000; 7,500; 8,000; 8,500; 9,000; 9,500; or 10,000 A² (e.g., from about 5,000 to about 10,000 A²; from about 5,500 to about 9,500 A²; from about 6,000 to about 9,000 A²; from about 6,500 to about 8,500 A²; from about 7,000 to about 8,000 A²; or from about 7,500 to about 8,000 A²).

[0021] In another aspect, the present invention provides a host cell comprising any of the foregoing expression cassettes or modified enzymes, wherein the host cell produces long chain (>C5) alcohols at a concentration of at least 5 mg/L, 25 mg/L, 50 mg/L, 100 mg/L, 200 mg/L, 250 mg/L, 300 mg/L, 400 mg/L, or 500 mg/L (e.g., from about 5 mg/L to about 300, 400, 500, 600, 700, 800, or more mg/L, from about 10 mg/L to about 500 mg/L, from about 10 mg/L to about 400 mg/L, from about 10 mg/L to about 300 mg/L, from about 25 mg/L to about 500 mg/L, from about 25 mg/L to about 400 mg/L, from about 25 mg/L to about 300 mg/L, from about 50 mg/L to about 500 mg/L, from about 50 mg/L to about 400 mg/L, from about 50 mg/L to about 300 mg/L; from about 100 mg/L to about 500 mg/L, from about 100 mg/L to about 400 mg/L, or from about 100 mg/L to about 300 mg/L).

[0022] In some embodiments, the host cell produces heptanol as the major alcohol product or second-most major alcohol product. In some embodiments, the host cell produces hexanol as the major alcohol product or second-most major alcohol product In some embodimetns, the host cell produces at least 200 mg/L (e.g., from about 200 to about 500, from about 200 to about 400, from about 250 to about 450, from about 250 to about 350, or from about 300 to about 350 mg/L) heptanol or hexanol, or a combination thereof. In some embodiments, the host cell produces octanol. In some embodiments, the host cell does not produce ethanol or propanol, or produces at least 10-fold (e.g., from 10-fold to 100-fold, or more, or from 10-fold to 50-fold, or from 10-fold to 20-fold) more long chain alcohols (>C5) as compared to ethanol and/or propanol.

BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 : Depicts an illustration of the synthetic recursive +1 pathway. This pathway employs enzymes LeuABCD from leucine biosynthesis for carbon chain elongation of 2- ketoacids. The enzyme ketoisovalerate decarboxylase (KIVD) performs the decarboxylation of 2-ketoacids and diverts carbon out from the +1 iterative cycle.

[0024] Figure 2: Depicts an illustration of a computationally directed genomic enzyme mining pipeline. (Left) GEOs are identified based on sequence homology to KIVD. Bioinformatics filters are used to identify a set of protein sequences from genomic databases that are predicted to be KIVD homologs and likely to be decarboxylases. Once identified, homology models are built to obtain a predicted ternary structure of each GEO. Ligand docking and design simulations are subsequently run in the presence of the target ligand (e.g., C8) to evaluate the potential protein- ligand interface energy. (Middle) A phylogenetic tree for the 239 GEOs is depicted with a bar chart above each sequence. The bar height indicates the predicted protein-ligand interface energy, the higher the bar the lower the energy. Bar height is scaled linearly relative to the lowest protein-ligand interface energy. 10 GEOs are selected for experimental characterization. (Right) The pairwise sequence identity of all 239 GEOs to each other has a mode of 20%, indicating the high level of sequence diversity within this set of sequences.

[0025] Figure 3: Depicts an illustration of reaction specificities for various 2 ketoacid decarboxylases, (a) Ketoisovalerate decarboxylase (KTVD) utilizes various ketoacids as substrates to produce the corresponding aldehyde, (b) Ketoacids used for in vitro kinetic constants in addition to 2-ketoisovaleric acid.

[0026] Figure 4: Depicts catalytic efficiency and specificity of characterized ketoacid decarboxylases. Specificity factor is calculated as where the log scaled catalytic

efficiencies for C8 relative to C5 and C3 are compared. The three active GEOs and the naively selected set of decarboxylases are represented as dots and diamonds, respectively. Native K1VD and KTVD VLV are depicted as a square and triangle, respectively. The genomic mining method and computational active site redesign approach both produced enzymes with enhanced specificity relative to native KTVD.

[0027] Figure 5: Depicts a molecular model of a C8 intermediate docked into KIVD (a, c) and GEO 17S (b, d). Cross sections of the space filled active site are represented in a and b where the C8 keto acid and TPP ligand are depicted. A detailed view of the amino acids within the active site are illustrated in c and d. Residues within 5 A of C8 ligand are shown in sticks and their corresponding C -alpha are shown in spheres. Figure was generated using PyMol vl .7.4.

[0028] Figure 6: Depicts results of kinetic characterization of GEO 175 and GEO 175 L376T T240S. Kinetic constants were measured as described herein. The substrate 2- ketooctanoate carbon chain numbering referred to in the text is numbered in white. According to the molecular model of GEO 175, the double mutation L376T T240S is predicted to recede the pocket by 1.4 A and remove interactions between the binding pocket and carbon 8 of the ketoacid alkyl chain. This mutant was observed to decrease catalytic efficiency on C8, but has a negligible effect on shorter chain substrates. [0029] Figure 7: Depicts an illustration of a screening process for identifying an engineered KIVD with altered substrate specificity. The design process can start with identifying active site residues of KIVD (depicted as spheres in panel on left). These amino acids were allowed to either remain native or sample any of eleven relatively hydrophobic amino acids. A total of 10,000 design simulations were run and the amino acids identified in the SO lowest energy designs were used to guide construction of a small library of roughly 400 KIVD mutants from the original combinatorial space of 1011 possible active site mutations. 1200 clonal isolates were screened for activity and specificity. The KTVD VLV mutant was selected and subsequently purified for in vitro kinetic constant characterization.

[0030] Figure 8: Depicts results of in vivo alcohol production of the synthetic recursive +1 pathway with native KIVD, KIVD VLV, and GEO 175. Cells were incubated for forty hours in microaerobic conditions in a defined media as described in materials and methods. Each assay was performed in triplicate and titers reported if all three samples had observed product production above the limit of quantitation (S mg/L).

[0031] Figure 9: Depicts toxicity as a function of alcohol chain length. The optical density of engineered E. coli strains after a 7-hour growth was measured as a function of the alcohol concentration added to growth media at the beginning of incubation.

[0032] Figure 10: Depicts an overlay of active sites between GEO 175 and BFD. Residues that are different between the proteins are shown in sticks and their corresponding alpha carbons are shown in spheres.

[0033] Figure 11: Depicts a screening process for identifying engineered KIVD with altered substrate specificity. Pertinent screening results and candidate mutants are shown (graphs) which contributed directly to finding KIVD VLV. Colorimetric assays (graphs) measured approximate enzyme activity of screened mutants for ketoisovalerate (IsoC5), 2-ketobutyrate (C4), 2-ketovalerate (C5), 2-ketohexanoate (C6), and 2-ketooctanoate (C8). Higher activity is shown by larger downwards slope; greater specificity relative to KTVD was observed in mutants with reduced activity for IsoC5, C4, and C5, and only slightly affected or greater activity for C6 and C8. First round high-throughput colorimetric screening of single mutants identified candidates such as G402V and M538L/F542V. Mutations from each region that showed potential in altering the specificity of KTVD were combined to produce multi-region mutants through site-directed mutagenesis using the plasmids of KIVD single region mutants as templates. These KIVD multi-region mutants, such as G402V M538L F542V (KIVD_VLV) were then screened a second time with the colorimetic assays. Enzyme kinetics was then measured for mutant KIVD VLV.

[0034] Figure 12: Depicts results of Michaelis-Menten curve fitting of tested enzymes with C8 substrate.

[0035] Figure 13: Depicts results of Michaelis-Menton curve fitting of tested enzymes with C5 substrate.

[0036] Figure 14: Depicts results of Michaelis-Menten curve fitting of tested enzymes with C3 substrate.

[0037] Figure 15: Depicts results of Michaelis-Menten curve fitting of tested enzymes with isoCS substrate.

[0038] Figure 16: Depicts pdb codes and accession numbers of GEOs modeled as listed on the phylogenetic tree depicted in Figure 2. Sequences were obtained from homologous sequence search in the non-redundant database using HMMER3.

[0039] Figure 17: Depicts protein sequences of various 2-ketoacid decarboxylases.

[0040] Figure 18: Depicts the native sequence and various engineered substitutions at various amino acid positions of KIVD used to probe for altered specificity for long chain 2-ketoacids. Amino acids in parentheses denote potential mutations due to degenerate codons used for mutagenesis.

[0041] Figure 19: Depicts activity data for various GEOs tested against C3, C5, C8, and isoC5 substrates and calculated Rosetta energy values for bound TPP-C8 ligand.

[0042] Figure 20: Depicts results of a structural alignment of GEOs and KIVD. Active site residues are denoted with a black box under the residue letter.

[0043] Figure 21: Depicts a profile of active site residues and preferred amino acid substitutions at such residues for 2-ketoacid decarboxylases described herein. A 5 A cut-off was utilized to define active-site residues. The active site residues are defined as those protein residues found in the model to have any non-hydrogen atoms within 5 A of any of the atoms in the TPP-C8 ligand. The model of KIVD_VLV was generated by threading the KTVD sequence onto the crystal structure of 2VBG (88% identical in sequence to KIVD), making the three mutations, and subsequently docking in the model of the C8 substrate.

[0044] Figure 22: Depicts a profile of active site residues and preferred amino acid substitutions at such residues for 2-ketoacid decarboxylases described herein. An 8 A cut-off was utilized to define active-site residues. The active site residues are defined as those protein residues found in the model to have any non-hydrogen atoms within 8 A of any of the atoms in the TPP-C8 ligand. The model of KTVD_VLV was generated by threading the KIVD sequence onto the crystal structure of 2VBG (88% identical in sequence to KTVD), making the three mutations, and subsequently docking in the model of the C8 substrate.

[0045] Figure 23: Depicts pairwise identities for 2-ketoacid decarboxylases described herein. [0046] Figure 24: Depicts the contents of a sequence alignment file in aln format.

DEFINITIONS

[0047] As used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural reference unless the context clearly dictates otherwise.

[0048] The term "nucleic acid" or "polynucleotide" refers to deoxyribonucleic acids (DNA) or ribonucleic acids (RNA) and polymers thereof in either single- or double-stranded form. Unless specifically limited, the term encompasses nucleic acids containing known analogues of natural nucleotides that have similar binding properties as the reference nucleic acid and are metabolized in a manner similar to naturally occurring nucleotides. Unless otherwise indicated, a particular nucleic acid sequence also implicitly encompasses conservatively modified variants thereof (e.g., degenerate codon substitutions), alleles, orthologs, SNPs, and complementary sequences as well as the sequence explicitly indicated. Specifically, degenerate codon substitutions may be achieved by generating sequences in which the third position of one or more selected (or all) codons is substituted with mixed-base and/or deoxyinosine residues (Batzer era/., Nucleic Acid Res. 19:5081 (1991); Ohtsviku et al., J. Biol. Chem. 260:2605-2608 (1985); and Rossolini et ah, Mol. Cell. Probes 8:91-98 (1994)). The term nucleic acid is used interchangeably with gene, cDNA, and mRNA encoded by a gene.

[0049] The term "gene" means the segment of DNA involved in producing a polypeptide chain. It may include regions preceding and following the coding region (leader and trailer) as well as intervening sequences (introns) between individual coding segments (exons).

[0050] A "promoter" is defined as an array of nucleic acid control sequences that direct transcription of a nucleic acid. As used herein, a promoter includes necessary nucleic acid sequences near the start site of transcription, such as, in the case of a polymerase Π type promoter, a TATA element. A promoter also optionally includes distal enhancer or repressor elements, which can be located as much as several thousand base pairs from the start site of transcription.

[0051] An "expression cassette" is a nucleic acid construct, generated recombinantly or synthetically, with a series of specified nucleic acid elements that permit transcription of a particular polynucleotide sequence in a host cell. An expression cassette may be part of a plasmid, viral genome, or nucleic acid fragment. Typically, an expression cassette includes a polynucleotide to be transcribed, operably linked to a promoter. In some cases, the promoter is heterologous to the polynucleotide to be transcribed. In some cases, the expression cassette is heterologous to the host cell in which it resides.

[0052] The term "amino acid" refers to naturally occurring and synthetic amino acids, as well as amino acid analogs and amino acid mimetics that function in a manner similar to the naturally occurring amino acids. Naturally occurring amino acids are those encoded by the genetic code, as well as those amino acids that are later modified, e.g., hydroxyproline, γ-carboxyglutamate, and O-phosphoserine. Amino acid analogs refers to compounds that have the same basic chemical structure as a naturally occurring amino acid, i.e., an a carbon that is bound to a hydrogen, a carboxyl group, an amino group, and an R group, e.g., homoserine, norleucine, methionine sulfoxide, methionine methyl sulfonium. Such analogs have modified R groups (e.g., norleucine) or modified peptide backbones, but retain the same basic chemical structure as a naturally occurring amino acid. "Amino acid mimetics" refers to chemical compounds having a structure that is different from the general chemical structure of an amino acid, but that functions in a manner similar to a naturally occurring amino acid. [0053] There are various known methods in the art that permit the incorporation of an unnatural amino acid derivative or analog into a polypeptide chain in a site-specific manner, see, e.g., WO 02/086075.

[0054] Amino acids may be referred to herein by either the commonly known three letter symbols or by the one-letter symbols recommended by the IUPAC-IUB Biochemical

Nomenclature Commission. Nucleotides, likewise, may be referred to by their commonly accepted single-letter codes.

[0055] "Polypeptide," "peptide," and "protein" are used interchangeably herein to refer to a polymer of amino acid residues. All three terms apply to amino acid polymers in which one or more amino acid residue is an artificial chemical mimetic of a corresponding naturally occurring amino acid, as well as to naturally occurring amino acid polymers and non-naturally occurring amino acid polymers. As used herein, the terms encompass amino acid chains of any length, including full-length proteins, wherein the amino acid residues are linked by covalent peptide bonds.

[0056] The term "heterologous" as used in the context of describing the relative location of two elements, refers to the two elements such as nucleic acids (e.g., promoter and protein encoding sequence), a nucleic acid and a host cell, a protein (e.g., a 2-ketoacid decarboxylase, or portion thereof) and a host cell or reaction mixture, or a protein and a ligand that are not naturally found in the same relative positions. Thus, a "heterologous promoter" of a gene refers to a promoter that is not naturally operably linked to that gene. Similarly a host cell containing a heterologous expression cassette can refer to an expression cassette that is not naturally found in that host cell.

[0057] A "3-layer alpha-beta-alpha sandwich" refers to the 3-layer (aba) Sandwich architecture according to the CATH classification (version 4.0.0). Proteins containing such architecture have a Rossmann fold topology. For enzymes that have not been characterized by X-ray crystallography, NMR spectroscopy, or other experimental means, the presence or absence of a 2-layer alpha beta alpha sandwich can be assessed by homology modeling. Exemplary methods of homology modeling can include the use of one or more of the following programs, program codes, scripts, and/or parameters, e.g. in their default settings,: RosettaCM, SWISS- MODEL, MODELLER, 3D-JIGSAW, ModPipe, Fugue, HHPred, i-Tasser, IntFOLD, M4T, ModWeb, Phyre2, RaptorX, Roberta, HHsuite, and the like, such as derivative works thereof.

[0058] A "thiamine pyrophosphate (TPP) binding pocket" refers to a binding pocket of a enzyme that productively binds TPP, such that the TPP bound by the enzyme can perform a ketoacid decarboxylation reaction. The presence of a TPP binding pocket can be detected by detecting bound TPP, or a radical intermediate thereof (e.g., via mass spectrometry, electron paramagnetic resonance (EPR) spectroscopy, or the like). The presence of a TPP binding pocket can also be inferred on the basis of detection of TPP mediated catalysis by the enzyme. In some cases, the presence of a TPP binding pocket can be determined from an experimentally determined structure of the enzyme (e.g., determined by X-ray crystallography). In some cases, the presence of a TPP binding pocket can be determined by homology modeling using any of the homology modeling methods described herein. In some cases, the presence of a TPP binding pocket can be inferred from the presence of sequence data indicative of a TPP binding domain. For example, a comparison of a candidate sequence to the Pfam database (e.g., Pfam 28.0) can reveal the presence or absence of a TPP binding pocket

[0059] A "TM align score" refers to a normalized structural similarity score comparing two 3- dimensional protein structures (e.g., experimentally determined or homology model structures) calculated by version 2012/05/07 of the TM-align algorithm (TMaligaf). See, Y. Zhang & J. Skolnick, Nucleic Acids Research, 33: 2302-2309 (2005); Y. Zhang & J. Skolnick, Proteins, 57: 702-710 (2004); and J. Xu & Y. Zhang, Bioinformatics, 26, 889-895 (2010). Where one or both structures contain multiple chains of substantially the same structure (e.g., due to non- crystal lographic symmetry or due to the availability of an ensemble of homology models), the TM align score refers to the lowest score between the one or more chains of a first structure and the one or more chains of a second structure.

[0060] "Solvent accessible surface area" or "SASA" in reference to a substrate binding pocket of an enzyme described or claimed herein refers to a SASA calculated using RosettaScripts with Rosetta version C2bf8f674e7b416b5b756630d7ccd5d64c57512c 2015-05-04 14:47:46 -0700 from git@github.com:RosettaCommons/main.git . The SASA can be calculated from a structure file (PDB) with a bound ligand. The structure file can be an experimentally determined structure or a structure determined by any one or more of the homology modeling methods described herein.

[0061] As used herein, "2-ketoacid active site," in the context of the claims can refer to amino acid residues corresponding to amino acid positions: (i) 23-27, 50, 51, 56, 75, 76, 79, 82, 114, 287, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, and 461 of SEQ ID NO:l; (ii) 23-27, 50, 51, 56, 72, 75, 76, 79, 82, 113, 114, 287, 352, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461, and 532 of SEQ ID NO:l; (iii) 23-27, 48, 49, 54, 74, 75, 78, 81, 113, 286, 376- 380, 382, 399, 401-404, 427-432, 434, 454-462, and 465 of SEQ ID NO:2; (iv) 23-27, 48, 49, 54, 71, 74, 75, 78, 81, 112, 113, 286, 356, 376-380, 382, 399, 401-404, 427-432, 434, 454-462, 465, and 535 of SEQ ID NO:2; or (v) 22-29, 32, 43, 46-50, 54, 69-79, 81, 82, 111-113, 240, 285-287, 290, 356, 360, 376-383, 398-407, 426-435, 453-466, 472, 473, 475, 477, 526-528, and 530-535 of SEQ ID NO:2.

[0062] Amino acid positions corresponding to any one or more of the foregoing residues can be determined by structural alignment to a template structure (e.g., a structure of SEQ ID NO: 1 or 2). One or both structures used in the structural alignment can be experimentally determined or determined by homology modeling using any one or more of the homology modelling methods described herein. Exemplary structural alignment methods and/or programs include, but are not limited to, TM-align, LSQMAN, Fr-TM-align, DALI, DaliLite, CE, CE-MC, and the like. Additional structural alignment tools useful in determining active site amino acids include, but are not limited to, those described in

web.archive.org/web/20151221055428/https://en. wikiriedia.org/wiki/Stmctural_alignment_softw are.

[0063] As used herein, "rosetta energy units" in the context of ligand binding interaction energy refers to an interaction energy between a bound ligand and a protein calculated using Rosetta Enzyme Design style constraints. Program code for calculating rosetta energy units is provided herein. DETAILED DESCRIPTION OF THE INVENTION

/. introduction

[0064] In the current synthetic recursive +1 pathway, the enzymes LeuA, LeuB, LeuC, and LeuD (collectively known as LeuABCD) are recruited to recursively elongate 2-ketobutyrate into longer chain 2-ketoacids.³' ⁴ These 2-ketoacids are then converted to aldehydes by ketoisovalerate decarboxylase (KIVD) and subsequently to alcohols by alcohol dehydrogenase (ADH). (Figure. 1) In this engineered pathway, the primary determinant for the final product chain length is the number of cycles through LeuABCD. (Figure 1)

[0065] Previous efforts to engineer the product specificity of this pathway have focused on broadening the specificity of LeuA and KIVD. When the specificity of LeuA and KIVD are broadened, branched Cg products are produced. ³ When only the specificity of LeuA is broadened, additional cycles through the pathway are carried out, resulting in a mixture of C₂-Cg linear alcohols; ⁴ however, the primary product is still 1-butanol. In order to enhance long chain alcohol production, the inventors have focused on discovering a ketoacid decarboxylase specific for long chain ketoacids. By replacing KIVD with an enzyme specific for long chain ketoacids, the short chain substrates are able to reenter the +1 iteration cycle until conversion to long chain ketoacids, which are then decarboxylated and reduced into the corresponding long chain alcohols.

II. Methods

[0066] Provided herein are methods of identifying 2-ketoacid decarboxylase enzymes having an altered specificity, or having a desired endogenous specificity. In some cases, the 2-ketoacid decarboxylases have or are altered to have improved specificity for or catalytic efficiency against long chain (>C5) ketoacids and/or provide or are altered to provide an increased amount or concentration of long chain (>C5) alcohols when present in a synthetic recursive +1 pathway in a host cell or reaction mixture, e.g. , as compared to SEQ ID NO: 1.

[0067] The methods can include mining known protein sequences to identify enzymes having homology to a known 2-ketoacid decarboxylase template, at least a subset of which can be predicted to have a desired endogenous specificity or can be predicted to be able to be redesigned to have an altered specificity. Protein sequences having homology to a template enzyme sequence (e.g., SEQ ID N0:1) can, e.g., be identified using a hidden markov model algorithm such as that implemented in HMMER3 or using a BLAST-type sequence similarity search. Sequences can be filtered by removing sequences that are highly similar in sequence, e.g., using CD-HIT. Exemplary cut-offs for highly similar sequences can include a cut-off which removes sequences having at least about 50%, 60%, 70%, 80%, 90%, 95%, or 99% identity to the template sequence.

[0068] Structural features of sequences identified as having sequence homology to the template sequence can then be assessed. For example, a homology model of candidate hits can be generated, e.g., using RosettaCM, SWISS-MODEL, MODELLER, 3D- JIGSAW, ModPipe, Fugue, HHPred, i-Tasser, IntFOLD, M4T, ModWeb, Phyre2, RaptorX, Roberta, HHsuite, and the like. Homology models so generated can be scored for three dimensional structural similarity to the structure of the template sequence. For example, a structural alignment program or methods as described herein, including but not limited to one or more of TM align, the DALI server (e.g., DaliLite v. 3 available at ekhidna.biocenter.helsinki.fi/dali_server), or the cealign algorithm implemented in PyMOL (e.g., version 1.7.6 available at www.pymol.org) can be used to assess structural similarity by calculating a TM align score, Z-score, or an RMSD value respectively.

[0069] In some embodiments, a TM align score is used to assess structural similarity. For example, a TM align score cut-off of at least about 0.5 to a known 2-ketoacid decarboxylase (e.g., SEQ ID NO: 1, SEQ ID NO:2 or any other 2-ketoacid decarboxylases described herein) can be used to filter out candidate 2-ketoacid decarboxylases from homology models of the initial sequence search results or from homology models of the sequence search results after removing highly similar sequences. Alternatively, a more stringent cut-off can be applied. For example, a TM align score of at least about 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.81, 0.815, 0.817, 0.85, or 0.9 can be utilized.

[0070] In some embodiments, an RMSD value as calculated by, e.g., cealign or other similar methods or programs can be utilized to assess structural similarity and filter out candidate 2- ketoacid decarboxylases from homology models of the initial sequence search results or from homology models of the sequence search results after removing highly similar sequences. In exemplary embodiments, RMSD cutoffs calculated as described herein of less than about 6.7 A, 6 A, 5.5 A, 5 A, 4 A, or 3.5 A can be used to filter out candidate 2-ketoacid decarboxylases from homology models of the initial sequence search results or from homology models of the sequence search results after removing highly similar sequences. Alternatively, a more stringent cut-off can be applied. For example, an RMSD value of less than about 3 A, 2.5 A, 2 A, or 1.6 A can be utilized.

[0071] In some embodiments, a Z-score calculated by the DALI server is used to assess structural similarity. For example, a Z-score cut-off of at least about 2 to a known 2-ketoacid decarboxylase (e.g., SEQ ID NO: 1, SEQ ID NO:2 or any other 2-ketoacid decarboxylases described herein) can be used to filter out candidate 2-ketoacid decarboxylases from homology models of the initial sequence search results or from homology models of the sequence search results after removing highly similar sequences. Alternatively, a more stringent cut-off can be applied. For example, a Z-score of at least about 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, or 62 can be utilized.

[0072] The homology models (e.g. , after removing models that lack significant structural similarity as assessed by, e.g., TM align score, Z-score, and/or RMSD) can be assessed to identify sequences that bind a preferred substrate. For example, a covalent intermediate of a 2- ketoacid decarboxylase reaction can be modeled in the active site of the generated homology models, the complex can be relaxed computationally, e.g., using Rosetta Enzyme Design, and a ligand binding energy calculated. Candidate sequences having a low ligand binding energy can be selected for in vitro and/or in vivo analysis or selected as a preferred enzyme for catalyzing decarboxylation of a substrate to produce a product that corresponds to the modeled catalytic intermediate ligand. In some embodiments, candidate sequences having a ligand binding energy of less than, or less than about, -5, -5.5, -6, -6.5, -7, -7.5, -8, -8.5, -9, -9.5, -10, -10.3, -10.4, -10.5, -10.6, -10.7, -10.8, -10.9, -11, -11.5, -12, -12.5, -13, -13.5, -14, -14.5, -15, -15.5, -16, -16.5, -17, - 17.5, -18, -18.5, -19, -19.5, or -20 Rosetta energy units can be selected for in vitro and/or in vivo analysis or selected as a preferred enzyme for catalyzing decarboxylation of a substrate to produce a product that corresponds to the modeled catalytic intermediate ligand.

[0073] Methods of identifying 2-ketoacid decarboxylase enzymes having a desired altered specificity (e.g., greater catalytic efficiency against long chain (>C5) ketoacids) can additionally or alternatively include reprogramming substrate specificity of a known 2-ketoacid decarboxylase (e.g., SEQ ID NO:l or any of the GEOs described herein) using computation enzyme design methods. For example, the Rosetta Molecular Modeling Suite, or another computational design platform, can be used to design such enzymes. For example,

experimentally determined structures and/or homology models of known or suspected 2-ketoacid decarboxylases can be used to explore potential active-site sequence space that would accommodate long chain (>C5) 2-ketoacids, e.g., 2-ketooctanoate as substrates. In some embodiments, a modeled reaction intermediate of, e.g., a C8 substrate, is docked within a predicted active site of an experimentally determined structure or homology model structure, the complex optionally minimized, and a ligand binding energy calculated. In some embodiments, ligand binding energies are sampled among a number of computationally generated enzyme mutations or ensemble conformations by allowing the identity and conformation of non-catalytic residues within the active site to change, and/or allowing backbone distances and/or angles to move during minimization.

[0074] Candidate 2-ketoacid decarboxylases can be cloned into an expression cassette, e.g., under the control of a promoter such as a heterologous promoter, expressed and assayed for activity. For example 2-ketoacid decarboxylase enzymes identified by one or more of sequence mining, homology modeling, structural similarity scoring, and ligand binding analysis can be assayed for activity against one or more 2-ketoacid substrates. As another example, 2-ketoacid decarboxylase enzymes identified by virtual screening of mutants of a known 2-ketoacid decarboxylase can be assayed for activity against one or more 2-ketoacid substrates.

//. Compositions

[0075] Described herein are 2-ketoacid decarboxylases, expression cassettes encoding such 2- ketoacid decarboxylases, and host-cells and/or reaction mixtures containing such 2-ketoacid decarboxylases or expression cassettes. In some embodiments, the expression cassette is heterologous to the host cell or reaction mixture in which it resides. In some embodiments, the expression cassette contains a heterologous promoter operably linked to a polynucleotide encoding a 2-ketoacid decarboxylase. The expression cassette can, e.g., comprise a heterologous promoter operably linked to a nucleic acid encoding an enzyme comprising 2-ketoacid decarboxylase activity, wherein the enzyme: comprises a 3-layer alpha-beta-alpha sandwich; comprises a thiamine pyrophosphate (TPP) binding pocket; has a TM align score of at least 0.S as compared against PDB ID:2VBG; and comprises at least a 100-fold greater catalytic efficiency for C8 2-ketoacid substrates as compared to C3 or isoCS 2-ketoacid substrates, wherein the catalytic efficiency is measured by kcat/Km.

[0076] In some cases, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that is at least 10%, 12%, 14%, 15%, 20%, 25%, 30%, 35%, 40%, 50%, 60%, 70%, 80%, 90% or 99% identical to SEQ ID NO: 1. Sequence identity can be determined using methods known in the art. For example, using BLAST, HMMER3, or the like, e.g., using the default settings.

[0077] In some cases, the enzyme comprising 2-ketoacid decarboxylase activity has a TM align score of at least 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.81, 0.815, 0.817, 0.85, or 0.9; a Z-score of or of at least about 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, or 62; or an RMSD of or of less than about 6.7 A, 6 A, 5.5 A, 5 A, 4 A, 3.5 A 3 A, 2.5 A, 2 A, or 1.6 A as compared against PDB ID:2VBG. In some cases, the substrate binding pocket of the enzyme has a solvent accessible surface area of, or of at least about 7,500; 8,000; 8,500; 9,000; 9,500; or 10,000 A².

[0078] In some cases, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation of a native sequence wherein the mutation is in an amino acid residue with a Ca within 8 A of a 2-ketoacid decarboxylase active site, wherein the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to one or more of (e.g., corresponding to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, or 46 of) the following amino acid positions: (i) 23-27, 50, 51, 56, 75, 76, 79, 82, 114, 287, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, and 461 of SEQ ID NO: 1 ; (ii) 23-27, 50, 51, 56, 72, 75, 76, 79, 82, 113, 114, 287, 352, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461 , and 532 of SEQ ID NO:l; (iii) 23-27, 48, 49, 54, 74, 75, 78, 81, 113, 286, 376-380, 382, 399, 401- 404, 427-432, 434, 454-462, and 465 of SEQ ID NO:2; (iv) 23-27, 48, 49, 54, 71, 74, 75, 78, 81, 112, 1 13, 286, 356, 376-380, 382, 399, 401-404, 427-432, 434, 454-462, 465, and 535 of SEQ ID NO:2; or (v) 22-29, 32, 43, 46-50, 54, 69-79, 81, 82, 111-113, 240, 285-287, 290, 356, 360, 376-383, 398-407, 426-435, 453-466, 472, 473, 475, 477, 526-528, and 530-535.

[0079] Amino acid positions corresponding to any one or more of the foregoing residues can be determined by structural alignment to a template structure (e.g., a structure of SEQ ID NO: 1 or 2). One or both structures used in the structural alignment can be experimentally determined or determined by homology modeling using any one or more of the homology modelling methods described herein. Exemplary structural alignment methods and/or programs include, but are not limited to, TM-align, LSQMAN, Fr-TM-align, DALI, DaliLite, CE, CE-MC, and the like. Additional structural alignment tools useful in determining active site amino acids include, but are not limited to, those described in en.wikipedia.org/wiki/Strucrtural_aligriment_sofhvare .

[0080] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation in an amino acid residue

corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1. In some embodiments, the mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1, and wherein the mutation is to a residue selected from the group consisting of Val, Leu, lie, Met, Phe, His, Gly, Ala, Thr, Tyr, and Trp. In some embodiments, the mutation in an amino acid residue corresponding to a mutation selected from the group consisting of G402V, M538L, and F542V of SEQ ID NO: 1. In some embodiments, the enzyme comprising 2-ketoacid

decarboxylase activity comprises two or three mutations in amino acid residues corresponding to residues selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1. In some cases, the enzyme comprising 2-ketoacid decarboxylase activity comprises one, two, or three of the mutations corresponding to G402V, M538L, and F542V of SEQ ID NO: 1.

[0081] In some embodiments, the enzyme comprising 2-ketoacid decarboxylase activity comprises SEQ ID NO:2 or SEQ ID NO:3. In some embodiments, the enzyme is, or is at least about, 80%, 90% or 99% identical to SEQ ID NO:2 or 3. In some embodiments, the enzyme comprises 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more insertions, substitutions, or deletions, relative to SEQ ID NO: 1 , 2 or 3, or relative to any of the GEOs described herein. In some cases, the insertion, substitution, or deletion is independently an insertion, substitution, or deletion of 1, 2, 3, or 4 amino acids at a given amino acid position. [0082] In some embodiments, , the enzyme comprising 2-ketoacid decarboxylase activity interacts with a C82-ketoacid substrate with a calculated interaction energy of less than, or less than about, -10, -10.3, -10.4, -10.5, -10.6, -10.7, -10.8, -10.9, -11, -11.5, -12, -12.5, -13, -13.5, - 14, -14.5, -15, -15.5, -16, -16.5, -17, -17.5, -18, -18.5, -19, -19.5, or -20 Rosetta energy units.

[0083] All patents, patent applications, and other publications, including GenBank Accession Numbers, cited in this application are incorporated by reference in the entirety for all purposes.

EXAMPLES

[0084] The following examples are provided by way of illustration only and not by way of limitation. Those of skill in the art will readily recognize a variety of non-critical parameters that could be changed or modified to yield essentially the same or similar results.

Example 1:

Introduction

[0085] To rationally engineer the product profile of the synthetic recursive +1 pathway two computationally directed approaches were employed to discover a decarboxylase with the desired functional profile. The first is a new approach to mine the rapidly growing sequence databases derived from genomic sequencing. Since the vast majority of these proteins are derived from the genomic sequencing of organisms (i.e., genomic enzyme orthologs: GEOs), most of them have not been experimentally characterized and neither function nor specificity is known. Previous efforts to mine GEOs for function have relied on random sampling or sequence-based bioinformatics over an entire enzyme family. ^{8,9 10} While enzymes with a desired function have been found using these methods, it requires a large number of genes to be produced and experimentally characterized. To enable a more efficient sampling method, a novel

computational pipeline was developed that integrates the use of bioinformatics and molecular modeling to carry out forward predictions of a GEO's ability to carry out a function of interest Using this integrative genomic mining approach a highly diverse set of ketoacid decarboxylases capable of utilizing 2-ketooctanoate (C8) as a substrate was identified. The median activity of GEOs selected using the integrative genomic mining approach is 75-fold greater than a set of naively selected proteins from the enzyme family. [0086] The second approach focuses on reprogramming the substrate specificity of KIVD by using computational enzyme design methods. The Rosetta Molecular Modeling Suite was used to direct the generation and screening of over 400 mutants.¹¹ This effort lead to the discovery of a mutant with >600-fold increase in specificity for C8. Finally, the most specific and active enzyme from each approach is incorporated into the synthetic recursive +1 pathway to evaluate its ability to modulate in vivo alcohol production. Both enzymes result in a complete switch of the alcohol production profile towards long chain alcohols, demonstrating the ability to rationally control biosynthetic product specificity for chemicals not commonly found in Nature.

Results

Integrative Genomic Mining for Enzyme Function

[0087] KIVD belongs to the thiamine pyrophosphate enzyme family that is composed of more than 17,000 sequences.¹² To identify GEOs with the desired function within this vast sequence space, bioinformatics¹³ and molecular modeling¹⁴ with functional constraints were combined to guide our exploration. (Figure 2) First, ketoacid decarboxylase GEOs were identified based on sequence similarity to the KIVD used in the current synthetic recursive +1 pathway. At the time the search was performed, 2082 sequences were found in the non-redundant sequence database as significant matches. To broadly search KIVD sequence space, redundant sequences were filtered out with a sequence identity cut-off of 90%. ¹⁵ Sequences derived from eukaryotic organisms were removed in order to increase the likelihood of producing soluble proteins in E. coli where the synthetic recursive +1 pathway has been implemented.

[0088] The approach assumed that all potential scaffolds with the desired function should be structurally homologous to KIVD and the sequence set was further refined accordingly. Since crystal structures for most of these predicted proteins are not available for analysis, homology models were produced to obtain a predicted ternary structure. Using RosettaCM one hundred models were generated for each of the 239 GEOs.¹⁴ The lowest energy model was selected as the representative for each GEO and evaluated for structural similarity to KIVD. The TMalign algorithm¹⁶ was used to overlay the models with the native KIVD crystal structure and only those with a TMalign score of >0.5 were kept These bioinformatics and structural filters resulted in 239 GEOs as candidates for the targeted function. (Figure 16) The pairwise identity of every sequence to each other was calculated and the mode of these values lies near 20%. (Figure 2) Thus the selected set of 239 GEOs represents a diverse sampling of sequence space for this fold family.

[0089] To further enrich the set for GEOs likely to function on C8, a modeled reaction intermediate of the C8 substrate was docked within the predicted active site. The lowest energy model for each GEO was used as a template for docking a modeled C8 intermediate with functional constraints, which ensured a productive geometric orientation between the predicted catalytic residues in the protein and intermediate. The C8 reaction intermediate used for docking is based on the synthetic thiamine pyrophosphate (TPP) analogue from a crystal structure of KIVD (PDB ID: 2VBG), which mimics the initial nucleophilic attack by the thiazolium.¹⁷ Docking calculations were carried out using Rosetta Design in which both the identity and conformation of non-catalytic residues within the active site were allowed to change, and the backbone was allowed to move during minimization. The calculated interface energy after the dock and design simulations was used to evaluate each GEO's ability to accommodate C8.

[0090] A phylogenetic tree based on sequence similarity was built for the 239 GEOs in which the lowest protein-ligand interface energy from the Rosetta Design simulation for each GEO are depicted. (Figure 2) The bar height above each GEO leaf is proportional to the lowest interface energy from all 239 GEOs, the higher the bar the lower the energy. Based on the calculated energies it is immediately apparent that a cluster of GEOs distant in sequence to KTVD (-15% sequence identity) is predicted to be capable of utilizing C8. The enzyme with the lowest energy, GEO 175, is a predicted protein in the genome database with no known function. The closest enzyme to GEO 175 (-50% sequence identity) for which significant biophysical characterization has been performed is benzoylformate decarboxylase from bacteria Pseudomonas putida

(ppBFD). ¹⁸ A structural analysis of the active site of ppBFD and GEO 175 reveals that the active sites are only 50% identical in sequence (Figure 10), and the activity of ppBFD on C8 has not been previously studied.

[0091] From the design simulations, ten GEOs were chosen for experimental characterization based on the predicted protein-ligand interface energy as well as sequence diversity (Figure 2, Figure 16). Before ordering these GEOs, each protein was manually analyzed and mutations introduced through the Rosetta Design simulations were reverted if they were not predicted to significantly enhance the calculated Rosetta interface energy. Synthetic genes encoding each enzyme was obtained, expressed in E. coli and purified for in vitro kinetic characterization using a panel of 2-ketoacid substrates of different chain lengths including C8, CS, C3 and the natural substrate for KIVD ketoisovaleric acid (Figure 3). Out of the 10 selected GEOs, six express and are able to be purified in a soluble form using our standardized method. Of these six, three exhibit detectable activities on at least one of the four ketoacids used for kinetic characterization.

[0092] Among the active GEOs, the enzyme with the highest efficiency on C8 is the computationally top ranked GEO 175, with a k_cat/K_Mof 17,000 NT's^"1. (Table 1, Figure 4) This is only 2-fold lower than the efficiency of native KIVD on C8. However, GEO 175's catalytic efficiency on C8 is 33,000-fold higher than on C3 and 354-fold higher than on C5. When assayed under the same conditions, native KTVD's catalytic efficiency on C8 is 762-fold higher than on C3 and 3.4-fold higher than on C5. This corresponds to over a 100-fold improvement in specificity (C8 versus C5) relative to the native KIVD.

[0093] To evaluate if the integrative genomic mining approach enriches the mining for C8 activity, an additional set of previously characterized ketoacid decarboxylases diverse in sequence were experimentally characterized in an equivalent manner. Five enzymes were selected (PDB: 10VM, 2VBI, 3FZN, 1ZPD, 10ZF) and all were found to have detectable activity on at least one of the four ketoacids used for GEO characterization. The median C8 activity from the GEOs selected using the integrative genomic mining approach is 75-fold higher than the naively picked decarboxylases (Table 1, Figure 4). This indicates the integrative genomic mining approach is effective in identifying functional enzymes that perform the desired reaction on C8.

[0094] The protein sequence for GEO 175 is not predicted to require any mutations to be capable of utilizing the C8 substrate, and therefore represents the native enzyme from Streptomyces sp. C. Comparing the active site pocket of GEO 175 with native KIVD (16.8% identity to each other) reveal significant differences which gave implications to the high specificity of this newly identified enzyme. (Figure 5) Based on molecular modeling, GEO 175 is predicted to have an extended active site that is significantly more open to solvent relative to native KIVD. This results in the predicted binding mode for C8 to be extended in GEO 175 as opposed to a "wrapped" conformation in KTVD. In GE0175, the third through fifth carbons of the C8 ketoacid ligand are predicted to occupy a solvent-exposed, open pocket with limited molecular interactions. However the sixth through eighth carbons are predicted to interact with a narrow and hydrophobic pocket Contrary to GEO 175, KIVD has an enclosed hydrophobic binding pocket and is predicted to make molecular interactions throughout the ketoacid alkyl chain (Figure 5).

[0095] From these observations it was hypothesized that the specificity of GEO 175 could be attributed to a combination of hydrophobic interactions with the sixth through eighth carbons of the alkyl chain, while lacking interactions with the third through fifth carbons. (Figure 6) Therefore, an engineered a mutant of GEO 175 containing two amino acid changes (L376T, T240S) predicted to remove the direct molecular interactions with seventh and eighth carbons of the ketoacid alkyl chain, while maintaining overall protein stability was generated. Upon kinetic characterization, it was found that the mutant protein's catalytic efficiency on C8 is 15-fold lower than GEO 175, but the decrease in efficiency for C5, C3, and isoC5 assayed under the same conditions is less than 2-fold. The ability to rationally modulate activity based on structure supports the accuracy of the GEO 175 molecular model.

Computationally Directed KTVD Active Site Redesign

[0096] In the second effort to reprogram the specificity of KIVD, Rosetta Design methods were used to explore the potential active site sequence space that would accommodate 2- ketooctanoate as a substrate.¹¹ Design simulations were carried out with the same reaction intermediate and functional constraints utilized in the genomic mining pipeline. Ten residues in the active site were targeted for mutagenesis. At each site, one of eleven relatively hydrophobic amino acids (Val, Leu, He, Met, Phe, His, Gly, Ala, Thr, Tyr, Trp) was systematically introduced, and the remaining nine sites were allowed to be redesigned. The identities of amino acids at all other residues were kept constant (Figure 7) The sequence profile from the 50 lowest energy designs was used to guide construction of a small library of roughly 400 KIVD mutants from the original combinatorial space of over 10¹¹ possible active site mutations.

[0097] Since KIVD can already efficiently utilize C8 as a substrate, enzyme specificity is of primary importance. Therefore each mutant was screened against a panel of 2-ketoacid substrates in order to assess the specificity and activity of each mutant (Figure 11) This assay was performed using a high-throughput pH-dependent colorimetric assay of enzyme activity in crude cell extracts. Mutations that increased specificity for C8 were then combined to produce combinatorial mutants and screened for specificity with the colorimetric assays. The mutant that exhibits the highest specificity and activity is G402V/M538L/F542V (KIVD VLV). The engineered enzyme KIVD_VLV has a 600-fold improvement in specificity between C8 and C5, in terms of catalytic efficiency, relative to the native KIVD.

In Vivo Testing of Engineered and Genomic Enzymes for Long-chain Alcohol Production

[0098] The measured specificity and activity for each enzyme was used to select two candidates for further characterization of their ability to modulate alcohol production in vivo. The two enzymes with the highest activity and specificity (Figure 4), KIVD VLV and GEO 175, were tested for their ability to reprogram the synthetic recursive +1 pathway to produce long- chain alcohol products in vivo. The alcohol-producing biochemical pathway was adapted from a previous study used to synthesize alcohols of various lengths.³' ⁴ Starting from glucose, the aforementioned +1 pathway, involving LeuABCD, recursively elongates 2-ketoacids, starting with 2-ketobutyrate, into longer chained 2-ketoacids. These 2-ketoacids are then converted to aldehydes by ketoacid decarboxylases (native KIVD, KIVD VLV, and GEO 175) and subsequently to alcohols by an alcohol dehydrogenase. Carbon flux through this alcohol synthesizing biochemical pathway is enhanced by overexpressing thrABC, ilvA, LeuABCD, and ADH6 on inducible plasmids transformed intoE. coli? The effects of overexpression were furthermore supported and maintained by knocking out the threonine exporter enzyme RhtA and the primary DNA recombination degradation enzyme RecA. The enzyme AdhE was also knocked out, which helps raise long chain alcohol production by eliminating a competing pathway where acetyl-CoA is directly converted to ethanol.¹⁹

[0099] Consistent with previous reports, the primary products (82% of total alcohol production) with native KTVD are short alcohols (C₂ - C₄) (Figure 8, Table 2). The KIVD_VLV triple mutant completely switches the product profile of this pathway to predominantly long chain (≥Cs) alcohols at 728 mg Γ¹. Hexanol (Ce) is the major alcohol detected at a titer of 341 mg Γ¹ (47%), with titers of 269 mg Γ¹ (37%) heptanol (C₇), and 118 mg Γ¹ pentanol (16%) also observed. GEO 175 also switches the product profile of the pathway so that long chain alcohols are the primary products and produced at a level of 522 mg Γ¹. For GE0175 the product profile is further shifted, resulting in heptanol (C₇) as the major product at a titer of 274 mg Γ¹ (50%), with hexanol production at 160 mg Γ¹ (29%), and 88 mg Γ¹ pentanol (16%). Similar to the results for KIVD VLV, no ethanol or propanol production is observed, and only 19 mg Γ¹ butanol is produced. In addition, 10 mg Γ¹ of octanol was also observed, a product not observed when either native KIVD, or KIVD VLV were tested under equivalent conditions. The in vivo butanol and octanol titers for GEOl 75 compared to KIVD VLV are consistent with the in vitro enzyme kinetics in which GEO 175 has a significantly higher activity on C8, but lower specificity relative to C5, than KTVD_VLV. Total long chain alcohol titers for both GEO 175 and

KIVD VLV are increased significantly in comparison to native KIVD, with a >10-fold increase in heptanol production in both cases, and a >95% yield of long chain alcohols.

Table 2: In vivo Alcohol Production

Cells were incubated for forty hours in microaerobic conditions in a defined media. The best enzyme in shifting the product profile was GEO 175 with 1 -heptanol being the major product. [0100] Replacing KIVD with GEO 175 or KIVD_VLV results in a significant increase of long chain alcohol production, however the overall alcohol production drops ~10-fold from the 4.0 g/L obtained with KIVD. We hypothesized that a potential reason for this could be due to the toxicity of long chain alcohols. To explore the potential of toxicity being a limiting factor in alcohol production, E. coli growth was monitored over a 7-hour incubation with 10-1000 mg Γ¹ Cs-Cg alcohols supplemented at the beginning of growth. (Figure 9) Long chain alcohols are found to be toxic as the supplemented concentration of C₆-Cg alcohols reach 250 mg Γ¹. The final OD of E. coli cultures drops from 1.9 to approximately 1.6 for and C₇ alcohols, and 1.0 for Ce. At 500 mg Γ¹ the OD drops to 1.0 and 0.1 for Ce and C₇ alcohols, respectively. No significant growth is observed for Cg. At 1000 mg Γ¹ no significant growth is observed for heptanol, and only an OD of 0.2 is achieved in the presence of hexanol. However, growth up to an OD of 1.5 is observed for pentanol. Significant toxicity in the range of 250 to 1000 mg Γ¹ for long chain alcohols is consistent with the level of long chain alcohol titer, roughly 500mg/L, produced in the engineered E. coli strains. This result indicates that in order to achieve higher total alcohol titer with this pathway, a strain of E coli with high tolerance toward long chain alcohols is likely necessary. ' However, similar issues with product toxicity have been addressed through the design of continuous extraction methods and could be applied to the system here to increase levels of long chain alcohols titers.

Discussion

[0101] In this study an integrative genomic mining approach was introduced to enable discovery of enzymes for a targeted function from sequence databases. The function of interest here was an enzyme with high activity for the decarboxylation of long chain ketoacids in order to modulate the carbon flux of a synthetic pathway and increase titers of long-chain alcohols. The integrative genomic mining approach led to the discovery of a previously uncharacterized protein that we demonstrate has the targeted functional properties, and only required experimental characterization often new proteins. In parallel, a more traditional computationally directed library screening approach was utilized to reengineer the activity of a well-established enzyme. This required two successive rounds of screening over four hundred mutants against a panel of substrates in order to identify an enzyme with the desired functional properties. When evaluated for their ability to modulate carbon flux in vivo both are able to completely shift the product profile towards long chain alcohols. [0102] The enzyme GEO 175 represents the product of a general and rapid approach for obtaining enzymes with a desired function from the rapidly growing sequence databases. This approach provides a viable alternative to often expensive and laborious enzyme engineering efforts that require screening libraries of mutants. By combining bioinformatics and molecular modeling this approach enables the identification of proteins likely to carry out a targeted function regardless of their native or putatively annotated activity. This overcomes issues with misannotation of protein function²³ or biased assumptions based on the closest, but often distal in sequence, characterized protein. The enzyme identified in this study, GE017S, is a clear illustration of this as it is roughly 15% identical to KTVD and its closest characterized sequence homolog (ppBFD, roughly 50% sequence identity) has a significantly different functional profile.

[0103] While GEOl 75 and KTVD_VLV both increased long chain alcohol titer >10-fold, we identified that toxicity needs to be addressed in order to further increase total long-chain alcohol titer. The titer of long chain alcohols produced is on the order of 0.5 g/L, which we then demonstrate is highly toxic for the cell lines used in this study. For industrial applications, bio- alcohol production often requires yields on the order of grams per liter.²⁴ To avoid the need of a continuous extraction fermentation system, future efforts for increasing long chain alcohol production should focus on engineering or finding strains that are tolerant to this level of product formation. Efforts to integrate pumps and reengineer the cell wall to be resistant to alcohols could potentially mitigate toxicity effects and concurrently enable higher titers of long-chain alcohols to be produced through this pathway. ²⁰'²¹

[0104] In summary, a new genomic mining approach and computationally directed library design efforts were both successfully implemented to obtain enzymes that enabled specific production of long chain alcohols in vivo. These results demonstrate the modularity of the synthetic +1 recursive pathway and provided a clear path forward to engineer industrial level production of long-chain alcohols. Finally, the integrative genomic mining approach introduced here is highly general, and with the rapid growth of sequence databases it has the potential to revolutionize the development and discovery of enzyme catalysts. Methods

Integrative Genomic Mining

[0105] To obtain the GEO sequences, the native KTVD sequence was input for a homologous sequence search using HMMER3's* online server.¹³ The resulting sequences were filtered using the CD-HIT* online server with a 90% identity cutoff.^{15, 25} A homology model of each sequence was made using Rosetta Comparative Modeling.¹⁴ At this point, the intermediate was placed into the active site and 1000 simulations were ran to relax the intermediate according to the constraints. For each model, the lowest 100 in overall protein energy models were selected and then from that subset, the lowest protein-ligand interface was chosen as the energy for the GEO.

[0106] An intermediate of the C8 product-yielding reaction was modeled using Spartan*.' Different conformations of the alkyl chain were included in the modeling and a conformational library was made using OpenEye Omega*.²⁷ For the enzyme design of KTVD, this intermediate was placed into the active site using distance and angle constraints. Rosetta Enzyme Design was run with default settings, an example of which is provided in the Rosetta Molecular Modeling Suite demos.

[0107] From the design simulations for each GEO the lowest protein-ligand interface Rosetta energy was used to select a tractable number of GEOs for experimental characterization. During these simulations, any amino acids with a C_awithin 8 A of the active site could be mutated to any of the 20 amino acids. An exemplary mapping of GEO 175 active site residues to 2VBG active site residues is provided as Table 3. Ten GEOs of significant interest were chosen with the following criteria: Five GEOs were chosen because they had the lowest predicted energies; the second five were chosen with the purpose of maximizing sequence space diversity. The 234 GEOs (less the five lowest in energy) were filtered with a sequence identity cutoff of 40% and the five sequences from this filtered list with the lowest energy were picked as the second five GEOs in our final list. Each model was evaluated in the Foldit interface and mutations made during the design simulations were reverted to the native amino acid if not predicted to improve the interface energy by more than 2 Rosetta energy units. Table 3: Mapping of GEO 175 active site residues to active site residues in 2VBG crystal structure

[0108] Phylogenetic tree of GEOs was generated using Geneious software⁴¹ using a Muscle sequence alignment.²⁸* The resulting tree was visualized using iTOL online tool in circular tree mode and rooted at native KIVD.²⁹'³⁰

[0109] Synthetic genes coding for each GEO were synthesized as a DNA String by Life Technologies or the Joint Genome Institute. Genes were codon optimized for Escherichia coli and the amino acid sequences are provided herein. See, Figure 17 and Sequence Listing. The string was cloned into the pET-29b(+) plasmid vector using the Gibson assembly between the Ndel and Xhol restriction sites which added a C-terminal 6x-His tag in-frame to the gene.

[0110] All programs referenced in this section and denoted with an asterisk (*) were run using the default settings unless otherwise specified.

KIVD Active Site Redesign

[0111] In the design simulations, ten residues in the proposed active site pocket were allowed to either remain native or sample any of eleven relatively hydrophobic amino acids (Val, Leu, He, Met, Phe, His, Gly, Ala, Thr, Tyr, Trp). The identities of amino acids at all other positions were kept constant. Residues within 12 A of the ligand were allowed to undergo conformational sampling during simulation. A total of 10,000 design simulations were ran, from which the 50 designs lowest in ligand-protein interface energy, and non-redundant in terms of sequence, were selected to represent the potential sequence space predicted to accommodate the C8 substrate. The profile was used to construct a small library of -400 KIVD mutants. Each amino acid in the library was sampled as a single mutation, with the exception of residues that were within five residues from one another. These were sampled in a combinatorial fashion as both single and double mutants in order to evaluate synergistic effects given their proximity to one another. . A complete list of the amino acids allowed for each site is provided in Figure 18.

Construction and Selection of KI VP Libraries

[0112] Partially degenerate oligonucleotides were ordered from Integrated DNA Technologies (San Diego, CA) and were used to generate libraries of kivd mutants (with N-terminal his tags) by PCR. Library DNA was purified, inserted into the PCR-amplified pQE9 vector by isothermal Gibson assembly, ³¹ and transformed into XLl-Blue cells. Individual ampicillin-resistant colonies were picked and grown in 96- well blocks at 37 °C overnight. The number of colonies picked was three times the theoretical library size to ensure approximately 95% probability of all possible mutation combinations occurring. Glycerol was added to 25% (w/v) and libraries were stored at -80 °C until enzyme activity and specificity were assayed. KIVD mutant libraries were screened using a pH-monitored enzyme assay (method explained below) to measure the rate of tT consumption resulting from the decarboxylation of 2-ketoacids.

[0113] KIVD mutant libraries were grown overnight in culture blocks, diluted 1 : 100 into fresh media (Luria Broth), grown at 37 °C for 3 hours to ODeoo -0.6, induced with 0. lmM IPTG (GoldBio, Saint Louis, MO), and grown for 2 additional hours at 37 °C. Cells from the well blocks were transferred to 96-well assay plates (Costar, Corning, NY) and cell density was measured at 600nm. An equal volume of permeabilization solution (8.7mM potassium phosphate, 43.4mMKCl, 0.87mMMgSC>4, pH 7.1±0.1, 8.7% (v/v) chloroform, 0.0043% (w/v) SDS, 0.26% (v/v) 2-mercaptoethanol) was then added to break the cell membranes.

Bromothymol blue, TPP (Sigma-Aldrich, St. Louis, MO), and substrate were added to 0.008% (w/v), 0.5mM, and lOmM, respectively. All substrates (2-ketobutyrate, 2-ketovalerate, 2- ketoisovalerate, 2-ketocaproate, and 2-ketooctanoate; Sigma-Aldrich, St Louis, MO) were dissolved in MilliQ (JH2O and pH was adjusted to 7.1±0.1. Absorbance was measured at 615nm in a spectrophotometer (TEK Powerwave XS, BioTek, Winooski, VT). Data were acquired for 15 min at 30 °C. Cells containing pQE_hiskivd_wt and pQE9 were used as positive and negative controls, respectively. Enzyme activity data were corrected for cell density. Mutants chosen from this screening had a 25% higher activity than wild-type KIVD for any of the 2-ketoacids.

[0114] The library derived from the pH-coupled assay was further screened using a second enzyme kinetics assay. This assay involved a second enzyme, alcohol dehydrogenase from S. Cerevisiae (ADH6), which is a key enzyme in the alcohol production pathway in reducing the aldehyde produced by KIVD into an alcohol. Alcohol dehydrogenase couples oxidation of NADPH to NADP⁺ to the reduction of aldehyde into an alcohol. Stoichiometrically, conversion of one mole of 2-ketoacid to n-alcohol depletes exactly one mole of NADPH. Therefore, activity of KIVD could be measured directly by measuring depletion of NADPH through ultra-violet spectrophotometry.

[0115] Overnight cultures of XL IB with pQE9 containing KIVD mutations were grown at 37 °C in a 96 well block. Cultures were diluted 1 : 100 and then grown 3 hours at 37 °C to OD«x) of 0.6. Cultures were induced with 0.1 mM IPTG (GoldBio, Saint Louis, MO) for 3 hours at 37 °C. Cultures were then centrifuged, and pellets were lysed with BugBuster (Novagen, Madison, WI). 5 μL of cell lysate from each culture and 175 μΐ_^ of a buffer mix containing coenzyme 1.5 mM TPP (Sigma), 0.2 mM NADPH (Fisher, Waltham, MA), 0.045 U ADH6, 100 mM NaP0₄ , 100 mM NaCl, and 10 mM MgCh, pH of 7 were added together into a 96 well plate. 20 μL of 100 mM substrate was added to dilute to make a final concentration of 10 mM. The substrates tested were IsoC5, C4, C5, C6, C8 and ¾0. (Figure 11) Absorbances were measured with a plate reader (TEK Powerwave XS, BioTek, Winooski, VT) at 340 nm at 30 °C for 15 minutes. The protein concentration in each sample was determined with a BCA assay (Thermo Scientific, Waltham, MA) and used to normalize slope values. Candidate KIVD mutants were chosen for greater activity on longer chained 2-ketoacids and decreased activity on shorter 2-ketoacids. This was determined from graphing absorbance vs. time for each substrate and qualitatively comparing slope values vs. wild-type (Figure 11).

[0116] Site-directed mutagenesis. Oligonucleotides encoding specific kivd mutations were ordered from Integrated DNA Technologies (Coralville, Iowa) and were used to mutate kivd in pZE_LeuA*BCDKA6 and pQE_hiskivd_wt by PCR. Amplified DNA fragments were purified, inserted into either the PCR-amplified pZE vector containing leuA*BCD and adh6 or the pQE vector by isothermal Gibson assembly, and transformed into XLl-Blue cells. Plasmid DNA was purified (Qiagen, Hilden, Germany) from overnight cultures of antibiotic-resistant colonies and the plasmid sequences were verified (Laragen, Culver City, CA).

[0117] Chromosomal gene knockout. Genes were removed from the ATCC 98082 ArhtA strain genome using PI transduction from the Keio collection as previously. The aldehyde- alcohol dehydrogenase gene (adhE) was knocked out to eliminate ethanol production from acetyl-coA. In all strains, recA was knocked out to prevent recombination between the genome and plasmids, thereby stabilizing the transformants. Primers used to target recA for knockout were designed based on the Keio collection (Genobase, ecoli.aist-nara.acjp) and ordered from Integrated DNA Technologies.

[0118] Fermentation procedure and analysis. For n-alcohol production, strains of ST128 were transformed with pZS thrO, pZAlacJlvAasleuA, and pZE_LeuA*BCDK* A6 containing various kivd mutations. Fermentation conditions were adapted from those in previous works, ^3-4 with the following changes: 20 mL of medium was used, with 100 μg/πlL ampicillin, 50 μg/uL kanamycin, and 100 μg/mL spectinomycin added. Cells were grown to an optical density at 600 nm of ~0.6, followed by induction with 0.1 mM isopropyl-b-D-thiogalactoside (IPTG). After fermentation, cells were centrifuged for 15 minutes at 4000 x g and 4 °C. The supernatant was split into two fractions for analysis, 5 mL for short chain alcohols (ethanol, 1-propanol, and 1- butanol), and 15 mL for long chain alcohols (1-pentanol, 1-hexanol, 1-heptanol, and 1-octanol). Long chain alcohols were extracted from the 15 mL fraction by 3 mL n-hexane prior to analysis. GC-FID analysis was performed as previously described.⁴

[0119] Protein expression, purification, and enzymatic assay of KIVD. For the native KIVD and KIVD VLV mutant, 2 ml overnight cultures of XL IB cells were transformed with pQE9 containing N-terminal his-tagged KIVD enzymes and grown in Terrific Broth (BP biomedical, Cat# 3046-042) with 50 μg/πlL of carbenicillin. (Fisher scientific, Cat# BP2648-5) For the GEOs, 2 ml overnight cultures of BLR cells were transformed with pet29b+ plasmid containing N-terminal his-tagged GEOs, and grown in Terrific Broth with 50 μg/mL of kanamycin. (Fisher scientific, Cat# BP906-5) These cultures were diluted 1 : 1000 in 500 mL of Terrific Broth with 1 mM MgSC>4, 1% glucose, and 50ug/mL of corresponding antibiotics, then grown at 37 °C for 24 hours. Cultures were pelleted down at 4700 rpm for 10 mins and resuspended in auto-induction media (LB broth, 1 mM MgS0₄, 0.1 mM TPP, IX NPS, and IX 5052) for induction at 18 °C for 34 hours. At the end of induction, cells were centrifuged (4700 rpm, 4 °C, 20 minutes), supernatant was removed, and cells were resuspended in 40 mL lysis buffer (100 mM Hepes pH 7.5, 100 mM NaCl, 10% glycerol, 0.1 mM TPP, 1 mM MgS0₄, 10 mM Imidazole, 1 mM TCEP) and 1 mM PMSF) and sonicated for 2 minutes. Lysed cells were centrifuged at 4700 rpm at 4 °C for 60 mins to remove cell debris. Supernatant was loaded on gravity flow column with 700 μΐ of cobalt slurry (Fisher scientific, CAT# PI-90091) washed with 10 ml of wash buffer (100 mM Hepes pH 7.5, 100 mM NaCl, 10% glycerol, 0.1 mM TPP, 1 mM MgSC>4, 10 mM Imidazole, and 1 mM TCEP). Cobalt bead bed was washed with 15 ml of wash buffer 5 times and proteins were eluted with 1000 μΐ of elution buffer (100 mM Hepes pH 7.5, 100 mM NaCl, 10% glycerol, 0.1 mM TPP, 1 mM MgS0₄, 200 mM Imidazole, and 1 mM TCEP). Protein samples were immediately buffer exchanged with spin concentrators (Satorius, CAT# VS0112) into storage buffer (100 mM Hepes pH 7.5, 100 mM NaCl, 10% glycerol, 0.1 mM TPP, 1 mM MgS0₄, and 1 mM TCEP) and stored at 4 °C until kinetics characterization. Protein concentrations were determined using a Synergy HI spectrophotometer (Biotek) by measuring absorbance at 280 nm using their calculated extinction coefficients with ExPASy ProtParam Tool.³³ All other buffers and salts were purchased from Fisher Scientific unless otherwise specified.

[0120] The kcat and KM values of selective KIVD mutants were measured for the substrates: C3, C5, isoCs, and C%. All substrates were dissolved in MilliQ H₂0 and pH was adjusted to 7.5 as necessary. Activity was measured at 0.005 mM to 10 mM substrates. The assay was performed in a 96- well half-area plate. Each reaction contains a final concentration of 0.5 mM NADH, 1 mMDTT, 0.1 mM TPP, 1 mM MgS0₄, reaction buffer (100 mM Hepes, 100 mM NaCl, 10% glycerol, pH 7.5) and ADH (Sigma Aldrich, A7011, 10 U/ml for C₃, C₅ and C₈ reactions, 500 U/ml for isoCs reactions). A wide range of ketoacid decarboxylase concentrations, 4.5 nM - 15 μΜ, were used according to the activity of each enzyme toward different substrates in order to perform steady state kinetics measurement over a period of an hour. Absorbance readings were taken every 1 -minute at OD340 at 21 °C for 60 minutes using the Synergy HI. Kinetic parameters (A½_t and KM) were determined by fitting initial velocity versus substrate concentration data to the Michaelis-Menten equation.

[0121] Alcohol Toxicity. Alcohol tolerance of the in vivo alcohol production strain was evaluated by supplementing the fermentation media with specified quantities of long chain alcohols, and measuring growth over 7 hours. Both the E. coli strain (ATCC 98082 pZS_thrO, pZAlac_ilvA_LeuA, pZE12LeuA*BCDKA6_KIVD_wt) and fermentation media (IX M9 metals + IX trace metal mix + 0.5% yeast extract + 2% glucose + antibiotics) and conditions are the same as described above (Fermentation procedure and analysis). Cells were grown in media without IPTG-induction to a starting ODeoo ~ 0.02 and then supplemented with either 1-petanol, 1-hexanol, 1-heptanol, or 1-octanol at specified concentrations of 10 mg/L, 50 mg/L, 250 mg/L, 500 mg/L, or 1000 mg/L. ODeoo readings were taken hourly up to 7 hours. Figure 9 shows the final ODeoo recorded at 7 hours.

REFERENCES

1. House, US White. "National bioeconomy blueprint" Washington DC, The White House, April (2012).

2. USD A, US. "Bio-based products: market potential and projections through 2025." US Department of Agriculture (2008).

3. Zhang, K., Sawaya, M. R., Eisenberg, D. S., & Liao, J. C. "Expanding metabolism for biosynthesis of non-natural alcohols." Proceedings of the National Academy of Sciences, 20653-20658 (2008).

4. Marcheschi, R. J., et al. "A synthetic recursive "+ 1" pathway for carbon chain

elongation." ACS chemical biology 7.4, 689-697 (2012).

5. Erickson, B., and Winters P. "Perspective on opportunities in industrial biotechnology in renewable chemicals." Biotechnology journal 7.2, 176-185 (2012).

6. Plaza, M., et al. "Biochemical and molecular characterization of a-ketoisovalerate

decarboxylase, an enzyme involved in the formation of aldehydes from amino acids by Lactococcus lactis " FEMS microbiology letters 238.2, 367-374 (2004).

7. Felnagle, E. A., et al. "Engineering synthetic recursive pathways to generate non-natural small molecules." Nature chemical biology 8.6, 518-526 (2012).

8. Bayer, Travis S., et al. "Synthesis of methyl halides from biomass using engineered microbes." Journal of the American Chemical Society. 6508-6515 (2009). 9. Heins, Richard A., et al. "Phylogenomically guided identification of industrially relevant GH1 β-glucosidases through DNA synthesis and nanostructure-initiator mass spectrometry. " ACS chemical biology. 2082-2091 (2014).

10. The National Academies. "Industrialization of Biology: A Roadmap to Accelerate

Advanced Manufacturing of Chemicals." The National Academis Press. (2015)

11. Richter, F., et al. "De novo enzyme design using Rosetta3." PLoS One 6.5, el9230

(2011).

12. Finn, Robert D., et al. "Pfam: the protein families database." Nucleic acids research (2013).

13. Finn, R.D., Clements J., and Eddy S.R "HMMER web server: interactive sequence similarity searching." Nucleic acids research, gkr367 (2011).

14. Song, Yifan, et al. "High-resolution comparative modeling with RosettaCM." Structure 21.10 (2013): 1735-1742.

15. Huang, Y., etal. "CD-HIT Suite: a web server for clustering and comparing biological sequences." Bioinformatics 26.5, 680-682 (2010).

16. Zhang, Y., and Skolnick J. "TM-align: a protein structure alignment algorithm based on the TM-score." Nucleic acids research 33.7, 2302-2309 (2005).

17. Berthold, Catrine L., et al. "Structure of the branched-chain keto acid decarboxylase (KdcA) from Lacto coccus lactis provides insights into the structural basis for the chemoselective and enantioselective carboligation reaction." Acta Crystallographica Section D: Biological Crystallography. 1217-1224 (2007).

18. Bruning, M, et al. "Structural and Kinetic Studies on Native Intermediates and an

Intermediate Analogue in Benzoylformate Decarboxylase Reveal a Least Motion Mechanism with an Unprecedented Short-Li ved Predecarboxylation Intermediate¹." Biochemistry 48.15, 3258-3268 (2009). 19. Shen, C.R., and Liao J.C. "Metabolic engineering of Escherichia coli for 1 -butanol and 1- propanol production via the ketoacid pathways." Metabolic engineering 10.6, 312-320 (2008).

20. Kell, Douglas B., et al. "Membrane transporter engineering in industrial biotechnology and whole cell biocatalysis." Trends in biotechnology (2015).

21. Foo, Jee Loon, et al. "Improving Microbial Biogasoline Production in Escherichia coli Using Tolerance Engineering." mBio 5.6, e01932-14 (2014).

22. Brennan, Timothy CR, et al. "Alleviating monoterpene toxicity using a two-phase

extractive fermentation for the bioproduction of jet fuel mixtures in Saccharomyces cerevisiae." Biotechnology and bioengineering 2513-2522 (2012).

23. Schnoes, Alexandra M., et al. "Annotation error in public databases: misannotation of molecular function in enzyme superfamilies." PLoS computational biology 5.12 (2009).

24. Parekh S. Strain improvement Desk encyclopedia of microbiology. Academic Press, Waltham, (2010).

25. Kearse, M, et al. "Geneious Basic: an integrated and extendable desktop software

platform for the organization and analysis of sequence data." Bioinformatics 28.12, 1647- 1649 (2012).

26. Spartan'08, Wavefunction, Inc. Irvine, CA. Y. Shao et al. Phys. Chem. Chem. Phys., 3172-3191 (2006).

27. OEChem, version 1.7.7, OpenEye Scientific Software, Inc., Santa Fe, NM, USA,

www.eyesopen.com, (2010).

28. Edgar, Robert C. "MUSCLE: multiple sequence alignment with high accuracy and high throughput." Nucleic acids research 32.5, 1792-1797 (2004).

29. Letunic, Ivica, and Peer Bork. "Interactive Tree Of Life (iTOL): an online tool for

phylogenetic tree display and annotation.", 127-128 Bioinformatics 23.1 (2007).

30. Letunic, Ivica, and Peer Bork. "Interactive Tree Of Life v2: online annotation and display of phylogenetic trees made easy." Nucleic acids research (2011). 31. Gibson, G.D., et al. "Enzymatic assembly of DNA molecules up to several hundred kilobases." Nature methods 6.5, 343-345 (2009).

32. Datsenko, A.K., and Wanner, B.L. "One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products." Proceedings of the National Academy of Sciences 97.12, 6640-6645 (2000).

33. Gasteiger E., et al. ExPASy. the proteomics server for in-depth protein knowledge and analysis. Nucleic Acids Res. 31 :3784-3788 (2003).

34. The PyMOL Molecular Graphics System, Version 1.7.4 Schrodinger, LLC.

INFORMAL SEQUENCE LISTING

PROGRAM CODE AND INPUT

[0122] Described herein are various program codes and input that can be utilized in the methods of the present invention for identifying, designing, scoring, and/or characterizing 2- ketoacid decarboxylases of the present invention. Ligand.pdb

[0123] The following exemplary ligandpdb file contents can be used for calculating Rosetta energy units for ligand:enzyme interactions. The ligand.pdb file can also be used to calculate a ligand binding pocket (e.g., active site) solvent accessible surface area.

Ligand.params

[0124] The following exemplary ligand.params file contents can be used for calculating Rosetta energy units for ligand: enzyme interactions. The ligandpdb file can also be used to calculate a ligand binding pocket (e.g., active site) solvent accessible surface area.

ATOM H20 Hapo X 0.08

ATOM HI 5 Hapo X 0.08

ATOM HI 6 Hapo X 0.08

ATOM H3 Hapo X 0.08

ATOM HI 7 Hapo X 0.08

ATOM H4 Hapo X 0.08

ATOM H5 Hapo X 0.08

ATOM H6 Hapo X 0.08

ATOM HI 4 Hapo X 0.08

ATOM HI Hapo X 0.08

ATOM H2 Hapo X 0.08

ATOM OIO OOC X -0.77

BOND Nl H8

BOND Nl H22

BOND N2 C9

BOND N3 CIO

BOND N4 C6

BOND N4 Cll

BOND CI C19

BOND CI H6

BOND CI H14

BOND C2 H7

BOND C2 H21

BOND C2 H28

BOND C3 C5

BOND C3 HI 5

BOND C3 HI 6

BOND C4 H9

BOND C4 H23

BOND C4 H27

BOND C5 HI 8

BOND C5 HI 9

BOND C5 H20

BOND C6 C8

BOND C6 H10

BOND C6 H24

BOND C7 N3

BOND C7 Hll

BOND C8 C7

BOND C8 C9

BOND C9 Nl

BOND CIO N2

BOND CIO C2

BOND Cll C17

BOND Cll SI

BOND C12 N4

SASA.xml

[0125] The following exemplary SASA.xml file contents can be used to calculate solvent accessible surface area (S AS A) of a 2-ketoacid decarboxylase active site.

Ligand Energies

[0126] Ligand binding energies described herein refer to the transition state substrate that is docked into the active site of a 2-ketoacid decarboxylase in a catalytically relevant conformation and calculated using Rosetta Enzyme Design style constraints. The base energy function is the talaris2013_csts.wts file which contains the following terms: fa_atr fa_rep fa_sol fa_intra_rep fa_elec pro_close hbond_sr_bb hbond_lr_bb hbond_bb_sc hbond_sc dslf_fal3 rama omega fa_dun p_aa__pp ref

[0127] This base score function can be invoked with the flag: -score: weights

RosetWmain/database/scoring/weights/talaris2013_cstwts

[0128] The flag -ligand: :old_estat turns the fa_elec term into the hack elec term.

[0129] The flag -enzdes : : favor native res 2 turns on the res type constraint term with a weight of 2.

[0130] This results in the final score function of : fa_atr fa_rep fa_sol fa_intra_rep hack elec pro_close hbond_sr_bb hbond_lr_bb hbond bb sc hbond_sc dslf_fal3 rama omega fa dun p_aa_pp ref res_type_constraint with weights of .8 .44 .75 .004 .7 1 1.17 1.17 1.17 1.1 1 .2 .5 .56 .32 1 2

[0131] The ligand score is weighted sum of this modified talaris 2013 score function. Solvent Accessible Surface Area (SASA) Calculation

[0132] The SAS A can be calculated using Rosetta with an input PDB file (the protein structure to score that has a ligand bound in it). The ligand and associated params file in the Rosetta format can be used to instruct the Rosetta program in how to treat and score the ligand. The binding pocket is auto detected. RosettaScripts can be used to run the SASA filter on the input PDB. The score is output in the log file/trace.

[0133] In an exemplary embodiment, SASA values are calculated using Rosetta version C2bf8f674e7b416b5b756630d7ccd5d64c57512c 2015-05-04 14:47:46 -0700 from

git@github.com:RosettaCommons/main.git . This version can be used for any of the Rosetta functions described herein (e.g., enzyme design, homology modeling, docking, energy function evaluation, etc.).

[0134] The SASA score can be calculated using the following command line input:

~/Rosetta/main/sourc^in/rosetta_scripts.default.macosclangrelease -database

~/Rosetta/main/database -parsenprotocol sasa.xml -s input pdb -extra res fa ligandparams .

Claims

WHAT IS CLAIMED IS: 1. An expression cassette comprising a heterologous promoter operably linked to a nucleic acid encoding an enzyme comprising 2-ketoacid decarboxylase activity, wherein the enzyme:

comprises a 3 -layer alpha-beta-alpha sandwich;

comprises a thiamine pyrophosphate (TPP) binding pocket;

has a TM align score of at least 0.5 as compared against PDB ID:2VBG; and comprises at least a 100-fold greater catalytic efficiency for C82-ketoacid substrates as compared to C3 or isoCS 2-ketoacid substrates, wherein the catalytic efficiency is measured by kcat/Km.

2. The expression cassette of claim 1, wherein the enzyme comprising 2- ketoacid decarboxylase activity has a TM align score of at least 0.7 as compared against PDB ID::2VBG.

3. The expression cassette of claim 1 , wherein the substrate binding pocket of the enzyme has a solvent accessible surface area of at least about 7S00 A².

4. The expression cassette of claim 1 , wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that is at least 14% identical to SEQ ID NO:l.

5. The expression cassette of claim 1, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation of a native sequence wherein

the mutation is in an amino acid residue having a Ca within 8 A of a 2-ketoacid decarboxylase active site, wherein

the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 50, 51, 56, 75, 76, 79, 82, 113, 114, 287, 352, 373-377, 379, 394, 396-399, 424-428, 430, 450-458, 461, and 532 of PDB K>:2VBG.

6. The expression cassette of claim 1, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation of a native sequence wherein

the mutation is in an amino acid residue within 8 A of a 2-ketoacid decarboxylase active site, wherein

the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 50, 51, 56, 71, 75, 76, 79, 82, 112, 114, 287, 356, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461, and 535 of SEQ ID NO: 1.

7. The expression cassette of claim 1, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that is at least 90% or 99% identical to SEQ ID NO: 1.

8. The expression cassette of claim 7, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1.

9. The expression cassette of claim 8, wherein the mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1 is to a residue selected from the group consisting of Val, Leu, He, Met, Phe, His, Gly, Ala, Thr, Tyr, and Trp.

10. The expression cassette of claim 8, wherein the mutation corresponds to a mutation selected from the group consisting of G402V, M538L, and F542V of SEQ ID NO: 1.

11. The expression cassette of claim 7, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises two mutations in amino acid residues corresponding to residues selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1.

12. The expression cassette of claim 7, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises mutations corresponding to G402V, M538L, and F542V of SEQ ID NO:l.

13. The expression cassette of claim 1, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence at least 90% or 99% identical, or identical, to SEQ ID NO:2.

14. The expression cassette of claim 13, wherein the enzyme comprising 2- ketoacid decarboxylase activity comprises an amino acid sequence that contains a mutation, wherein

the mutation is in an amino acid residue of SEQ ID NO:2 having a Ca within 8 A of a 2-ketoacid decarboxylase active site, wherein

the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 48, 49, 54, 71, 74, 75, 78, 81, 112, 113, 286, 356, 376-380, 382, 399, 401-404, 427-432, 434, 454-462, 465, and 535 of SEQ ID NO:2.

15. The expression cassette of claim 1 , wherein the enzyme comprising 2- ketoacid decarboxylase activity interacts with a C8 2-ketoacid substrate with a calculated interaction energy of less than -10.4 rosetta energy units.

16. A modified enzyme comprising 2-ketoacid decarboxylase activity, wherein the enzyme comprises:

a 3 -layer alpha-beta-alpha sandwich;

thiamine pyrophosphate; and

at least a 100-fold greater catalytic efficiency for C8 2-ketoacid substrates as compared to C3 or isoC5 2-ketoacid substrates, wherein the catalytic efficiency is measured by kcat/Km,

wherein the modification comprises a mutation in the primary amino acid sequence of SEQ ID NO:l or SEQ ID NO:2.

17. The modified enzyme of claim 16, wherein the enzyme has a TM align score against PDB ff>:2VBG of at least 0.5.

18. The modified enzyme of claim 16, wherein the enzyme has a Ca root mean squared deviation of less than 1 A over at least 100 structurally aligned residues as compared against PDB ID:2VBG.

19. The modified enzyme of claim 16, wherein the modification comprises a mutation in an amino acid residue, wherein

the amino acid residue is within 8 A of a 2-ketoacid decarboxylase active site, wherein

the 2-ketoacid decarboxylase active site comprises the amino acids corresponding to amino acid positions 23-27, 50, 51, 56, 71, 75, 76, 79, 82, 112, 114, 287, 356, 373-377, 379, 394, 396-399, 423-428, 430, 450-458, 461, and 535 of SEQ Π> ΝΟ:1.

20. The modified enzyme of claim 16, wherein the modification comprises a mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1.

21. The modified enzyme of claim 20, wherein the mutation in an amino acid residue corresponding to a residue selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1, and wherein the mutation is to a residue selected from the group consisting of Val, Leu, He, Met, Phe, His, Gly, Ala, Thr, Tyr, and Trp.

22. The modified enzyme of claim 20, wherein the modification comprises two mutations in amino acid residues corresponding to residues selected from the group consisting of G402, M538, and F542 of SEQ ID NO: 1.

23. The modified enzyme of claim 22, wherein the modification comprises one or two mutations corresponding to the mutations selected from the group consisting of G402V, M538L, and F542V of SEQ ID NO: 1.

24. The modified enzyme of claim 23, wherein the modification comprises mutations corresponding to G402V, M538L, and F542V of SEQ ID NO: 1.

25. The modified enzyme of claim 16, wherein the ratio of kcat/Km against C8 2-ketoacid substrates over the kcat/Km against isoC5 2-ketoacid substrates is at least 50-fold greater than the native enzyme.

26. The modified enzyme of claim 16, wherein the solvent accessible surface of the substrate binding pocket is at least about 7500 A².

27. A host cell comprising the expression cassette of claim 1 , or the modified enzyme of claim 16, wherein the host cell produces long chain (>CS) alcohols at a concentration of at least 500 mg/L.

28. The host cell of claim 27, wherein the host cell produces heptanol as the major alcohol product 29. The host cell of claim 27, wherein the host cell produces octanol. 30. The host cell of claim 27, wherein the host cell does not produce ethanol or propanol.