A Flux-to-AI Discovery Framework Reveals Carbon-Regulated Intermittent Non-Canonical Nitrification in Achromobacter xylosoxidans

Life on Earth depends on a delicate, invisible exchange of nitrogen, an element essential for building proteins and DNA. In the natural world, bacteria act as the primary engineers of this cycle, transforming nitrogen from the air or soil into forms that plants and animals can use. While some bacteria perform this work using sunlight or inorganic chemicals, others, known as heterotrophs, rely on organic carbon—essentially food—to power their nitrogen transformations. For decades, scientists have understood the standard pathways for this process, but a hidden, alternative route has remained elusive in certain bacteria. This mystery is particularly relevant to human health, as some of these bacteria are opportunistic pathogens that thrive in the lungs of people with cystic fibrosis, a genetic disorder that creates thick, sticky mucus. Understanding how these bacteria survive and adapt in such a harsh, nutrient-rich environment could reveal new ways to manage chronic infections.

A team of researchers at San Diego State University has now uncovered a hidden metabolic strategy used by a specific bacterium called Achromobacter xylosoxidans. By combining advanced computer modeling with laboratory experiments and artificial intelligence, they discovered that this bacterium can switch its energy production methods depending on how much food is available. When food is plentiful, the bacteria grow normally. But when food becomes scarce, they activate a unique, intermittent process to generate energy by oxidizing nitrogen, a behavior that had not been clearly characterized before. This discovery not only fills a gap in our understanding of the global nitrogen cycle but also sheds light on how these bacteria persist in the human body, potentially offering new targets for treatment.

The journey to this discovery began with a massive effort to map the genetic blueprint of Achromobacter xylosoxidans. The researchers gathered genomic data from 133 different strains of the bacterium, including 21 new strains they isolated from the sputum of cystic fibrosis patients in the United States. These strains came from diverse environments, including soil, water, animals, and humans. The team found that the human-associated strains, particularly those from cystic fibrosis patients, had slightly smaller genomes than their environmental counterparts, suggesting they had shed unnecessary genetic material to adapt to life inside the human body. Despite these differences, the bacteria shared a core set of genes that hinted at a shared ability to perform a rare type of nitrogen processing.

To understand how this bacterium functions, the researchers built a detailed computer model of its metabolism, essentially creating a digital twin of the cell's chemical reactions. This model, which they named iGC1310, was constructed by mapping the bacterium's genes to known biochemical reactions and then rigorously testing it against real-world data. The team grew the bacteria in the lab on 285 different food sources, ranging from simple sugars to complex amino acids, to see which ones allowed the bacteria to thrive. They then compared these experimental results with the computer model's predictions. Through a process of trial and error, they refined the model until it could accurately predict the bacterium's growth behavior in over 91 percent of the cases. This high level of accuracy gave the researchers confidence to use the model to explore questions that were difficult to answer in the lab.

One of the most significant findings emerged when the researchers simulated how the bacterium handles nitrogen under different food conditions. The model predicted that the bacterium possesses a set of enzymes capable of converting ammonium, a form of nitrogen, into nitrite. This process, known as nitrification, is typically associated with bacteria that do not rely on organic food. However, the model suggested that Achromobacter xylosoxidans performs this task only when its supply of organic carbon is low. To test this, the researchers grew the bacteria in the lab and monitored their chemical output over time. They observed that while the bacteria consumed ammonium during their rapid growth phase, they only began producing nitrite after they had used up most of their food and entered a slower, stationary phase. This confirmed the computer's prediction: the bacterium switches to this nitrogen-processing mode as a survival strategy when its primary food source runs out.

The researchers also discovered that this nitrogen-processing ability is not uniform across all strains of the bacterium. By analyzing the genomes of the 133 strains, they found that the genes required for a related process called denitrification—which allows bacteria to use nitrate as a food source—were completely missing in the strains isolated from human patients. In contrast, these genes were present in strains found in soil and water. This suggests that the human-associated bacteria have lost the ability to use nitrate, likely because it is not a useful resource in the human lung environment. Instead, these human strains have retained and specialized in the ability to process nitrogen in a way that helps them survive when other nutrients are scarce.

To make sense of these complex patterns across so many different strains, the team employed a novel approach they call "Flux-to-AI." They fed the computer model's predictions about how each strain would grow on different nutrients into an artificial intelligence system. The AI analyzed the data to find specific patterns that distinguished bacteria from humans from those found in the environment. The system identified three key nutrients that acted as strong markers for human-associated strains: a small peptide called Gly-Met, a compound called acetoacetate, and the amino acid L-cysteine. These substances are known to be abundant in the lungs of cystic fibrosis patients, where they are released by the breakdown of dead cells and proteins. The AI analysis suggested that human-associated bacteria have evolved to specifically exploit these amino acid-rich environments, using them as a primary energy source while simultaneously managing their nitrogen balance.

The study highlights a sophisticated adaptation strategy where these bacteria have streamlined their genetic toolkit to fit a specific niche. By losing the ability to use certain environmental nutrients like nitrate, they have doubled down on the ability to thrive in the amino acid-rich mucus of the human lung. The intermittent nature of their nitrification process, turning on only when food is low, appears to be a crucial mechanism for maintaining energy levels during periods of stress. This finding challenges the previous view that these bacteria are simply passive inhabitants of the lung, suggesting instead that they are active metabolic engineers constantly adjusting their internal chemistry to survive.

The implications of this work extend beyond a single bacterium. The researchers have created a robust framework that combines genetic data, metabolic modeling, and artificial intelligence to predict how bacteria will behave in different environments. This "Flux-to-AI" approach could be applied to other microbes to uncover hidden metabolic pathways and understand how they adapt to hosts or changing climates. For the medical community, the identification of specific nutrients that human-associated strains rely on opens the door to new therapeutic strategies. By targeting the pathways that allow these bacteria to process amino acids or switch to nitrogen-based energy, it may be possible to disrupt their ability to persist in the lungs of cystic fibrosis patients. The study does not offer an immediate cure, but it provides a clear, detailed map of the metabolic landscape these bacteria inhabit, offering a new perspective on how to navigate it.