Nitrogenase resurrection and the evolution of a singular enzymatic mechanism

Abstract

The planetary biosphere is powered by a suite of key metabolic innovations that emerged early in the history of life. However, it is unknown whether life has always followed the same set of strategies for performing these critical tasks. Today, microbes access atmospheric sources of bioessential nitrogen through the activities of just one family of enzymes, nitrogenases. Here, we show that the only dinitrogen reduction mechanism known to date is an ancient feature conserved from nitrogenase ancestors. We designed a paleomolecular engineering approach wherein ancestral nitrogenase genes were phylogenetically reconstructed and inserted into the genome of the diazotrophic bacterial model, Azotobacter vinelandii, enabling an integrated assessment of both in vivo functionality and purified nitrogenase biochemistry. Nitrogenase ancestors are active and robust to variable incorporation of one or more ancestral protein subunits. Further, we find that all ancestors exhibit the reversible enzymatic mechanism for dinitrogen reduction, specifically evidenced by hydrogen inhibition, which is also exhibited by extant A. vinelandii nitrogenase isozymes. Our results suggest that life may have been constrained in its sampling of protein sequence space to catalyze one of the most energetically challenging biochemical reactions in nature. The experimental framework established here is essential for probing how nitrogenase functionality has been shaped within a dynamic, cellular context to sustain a globally consequential metabolism.

Keywords: Azotobacter vinelandii; ancestral sequence reconstruction; azotobacter vinelandii; biochemistry; chemical biology; early life; evolution; evolutionary biochemistry; evolutionary biology; metabolic engineering; nitrogen fixation; nitrogenase.

Figure 1.. Engineering strategy for ancestral nitrogenase resurrection.

(A) Experimental pipeline for nitrogenase resurrection in A. vinelandii and subsequent characterization, as described in the main text. (B) Structural overview of ancestral nitrogenases reconstructed in this study. Homology models (template PDB 1M34) of Anc1B NifH and NifDK proteins are shown with ancestral substitutions (relative to WT) highlighted in red. Select substitutions at relatively conserved sites in proximity to FeMoco (NifD, I355V) and the NifD:NifK interface (NifD, F429Y; NifK, R108K) are displayed in the insets. (C) Parallel genome engineering strategies were executed in this study, involving both ancestral replacement of only nifD (Anc1A and Anc2) and replacement of nifHDK (Anc1B). ‘P_nifH’: nifH promoter, ‘KanR’: kanamycin resistance cassette. *Anc1A and Anc1B were each reconstructed from equivalent nodes of alternate phylogenies (see Materials and methods).

Figure 2.. Phylogenetic and genomic context of resurrected ancestral nitrogenases.

(A) Maximum-likelihood phylogenetic tree from which ancestral nitrogenases were inferred. Extant nodes are colored by microbial host taxonomic diversity. Red box highlights the clade targeted in this study and depicted in (B). Tree shown was used to infer Anc1A and Anc2 sequences (an alternate tree was used for Anc1B inference; see Materials and methods). (B) nif gene cluster complexity within the targeted nitrogenase clade. Presence and absence of each nif gene are indicated by gray and white colors, respectively. Because some homologs for phylogenetic analysis were obtained from organisms lacking fully assembled genomes, the absence of accessory nif genes may result from missing genomic information. (C) Amino acid sequence identity matrix of nitrogenase protein subunits harbored by WT and engineered A. vinelandii strains. (D) Extant and ancestral nitrogenase protein sequence space visualized by machine-learning embeddings, with the resulting dimensionality reduced to two-dimensional space. UMAP dimension axes are in arbitrary units. The field demarcated by dashed lines in the left plot is expanded on the right plot.

Figure 2—figure supplement 1.. Maximum-likelihood phylogenies built from nitrogenase NifHDK homologs.

Anc1A and Anc2 sequences were inferred from the left tree, and the Anc1B sequence was inferred from the right tree (targeting an equivalent node to Anc1A). Both trees were reconstructed from the same extant sequence dataset (see Materials and methods for a description of phylogenetic reconstruction and ancestral sequence inference methods). Branch length scale indicates amino acid substitutions per site and applies to both trees.

Figure 2—figure supplement 2.. Alignment of WT and ancestral NifHDK nitrogenase proteins.

Figure 2—figure supplement 3.. WT and ancestral NifD alignment with sitewise ConSurf conservation scores (see Materials and Methods).

Conserved sites are defined by a ConSurf conservation score >7.

Figure 3.. Cellular-level characterization of ancestral nitrogenase activity and expression.

(A) Diazotrophic growth curves of A. vinelandii strains measured by the optical density at 600 nm (‘OD600’). A smoothed curve is shown alongside individual data points obtained from five biological replicates per strain. The non-diazotrophic DJ2278 (ΔnifD) strain was used as a negative control. (B) Mean doubling and midpoint times of A. vinelandii strains, calculated from data in (A). (C) In vivo acetylene (C₂H₂) reduction rates quantified by the production of ethylene (C₂H₄). Bars represent the mean of biological replicates (n=3) per strain. (D) Immunodetection and protein quantification of Strep-II-tagged WT (‘WT-Strep,’ strain DJ2102) and ancestral NifD. Top gel image shows Strep-II-tagged NifD proteins detected by anti-Strep antibody and bottom gel image shows total protein stain. Plot displays relative immunodetected NifD signal intensity normalized to total protein intensity and expressed relative to WT. Bars in the plot represent the mean of biological replicates (n=3) per strain. (B–D) Error bars indicate ±1 SD and asterisks indicate p<.01 (one-way ANOVA, post-hoc Tukey HSD) compared to WT or WT-Strep.

Figure 4.. In vitro analyses of ancestral nitrogenase activity profiles and mechanism.

All measurements were obtained from assays using purified NifDK assayed with WT NifH. (A) Specific activities were measured for H⁺, N₂, and C₂H₂ substrates. (B) Partial schematic of the reductive-elimination N₂-reduction mechanism of nitrogenase is shown above, centering on the N₂-binding E₄(4 H) state of FeMoco (see main text for discussion) (Harris et al., 2019). (C) Inhibition of N₂ reduction by H₂, evidencing the mechanism illustrated in (B). (D) Catalytic efficiencies of ancestral nitrogenases, described by the ratio of formed H₂ to reduced N₂ (H₂/N₂), mapped across the targeted phylogenetic clade. NifD homology models (PDB 1M34 template) are displayed with ancestral substitutions highlighted in red. (A,C) Bars represent the mean of independent experiments (n=2) with individual data points shown as black circles.

Figure 4—figure supplement 1.. SDS-PAGE of purified WT and ancestral NifDK proteins.

*NifD and NifK were not separately resolved for Anc2 and Anc1B. Qualitative assessment of band density suggests both subunits migrate together. The presence of both NifD and NifK is inferred based on the observed N₂ reduction activity of these fractions together with WT NifH (see Figure 4).