Through our proteomic approach, proteins identified in the Ab ATCC 19606 strain were classified into five major categories (Class A, C, D β-lactamases, RNAases, PBP; Fig. 2), revealing that Class C β-lactamases were the most abundantly expressed. The other major categories, RNAases and PBP, were not considered further since these may be recognized as an artifact of the experimental design. For example, the concentration dependent increase in PBP is likely indicative of the greater killing of bacteria by carbapenems at higher concentrations. Since PBPs are retained within the cell membrane and not secreted into the extracellular environment like β-lactamases, the increase in PBP identification may result from the release of PBPs into the extracellular space following bacterial death, rather than upregulation. However, this alternative explanation cannot be entirely excluded. We consider the same arguments to be possible explanations for the observation of RNAases in our samples. Instead, by focusing only on the β-lactamase expression, the relative amounts of various β-lactamase classes could be investigated (Fig. 3).
The putative Class A β-lactamase (Accession number: A0A1V2V229) was identified, and some concentration dependency was demonstrated. However, since the genome of Ab 19606 does not contain genes encoding for such proteins, this protein may be the result of misidentification. It is, at the current time, difficult to analyze this protein further as all similar records in Uniprot are proteins that have been predicted or inferred from homology and may be subject to change. At the same time, this identified protein is minor and therefore likely plays a limited role in any observed carbapenem resistance.
Surprisingly, the Class D β-lactamases (e.g., OXA) encoded in the Ab 19606 genome were minor contributors, and their expression was found to decrease upon exposure to biapenem in a dose-dependent manner. The reason why the Class D Oxa beta-lactamase, known to be significantly associated with carbapenemase activity, is hardly or not observed in the Ab ATCC 196060 strain has not been elucidated. Although these proteins are similar in molecular weight to both Class C and A proteins, samples were prepared with a minimum amount of polyacrylamide. It is, therefore, possible that a section in which OXA was present, but only in minute amounts, and therefore invisible to the naked eye, may have been missed; however, they would have remained minor contributors to the overall β-lactamase production in the ATCC 196060 Ab strain.
As the primary contributors to identified β-lactamases belonged to Class C, a closer look at possible trends was taken. Here, the major isoforms (those representing 20% or more of Class C) did not exhibit any clear concentration dependence (Fig. 4b) although they could be identified in all examined strains (WT and mutant) of Ab ATCC 19606. Instead, pronounced expression trends were mainly observed in minor isoforms (Fig. 4c). We observed a tendency for meropenem to induce the greatest relative increase in these minor isoforms as its concentration increases, while imipenem was most frequently associated with a dependent expression trend. Conversely, treatments with doripenem and biapenem resulted in isoform expression rate increases that were either equivalent or relatively lower in correlation.
At this time, none of the physical properties that are readily obtainable from PubChem could be correlated with the observed trends (Supplementary Fig. S1). Rather than search for a single needle-in-the-haystack property that could be correlated to our observations, we sought to understand the connection between the various observed protein isoforms. To do so, we consider that these minor isoforms are most likely mutations of the observed major counterparts (Q0VTS5 ADC-23, A0A5C0PFX8, A0A858S429). This was investigated by examining the sequences of each identified isoform and comparing their similarities to each other (Fig. 5).
From this, it supports the idea that several of the minor isoforms that follow monotonic trends, are mutants of the major, identified isoforms (Q0VTS5 ADC-23, A0A5C0PFX8, A0A858S429). For example, Q5UCE5 (ADC-1), which was found to increase with imipenem concentration (Fig. 4c), has 99% similarity to A0A55C0PXF8, suggesting the two are related by the mutation of several amino acids as well as the possibility that imipenem is in some way potentiating this mutation. Another example of this can be observed with J7GS97 and A0A482GTJ0 (Fig. 4c), which both increase in response to all four treatments and are highly similar to ADC-7. However, ADC-7 was not identified by LC-MS/MS in any sample, suggesting that our lab strain may have already had mutations associated with this protein or that numerous mutations took place or that these isoforms are instead associated with the highly similar ADC-30 (Fig. 5b).
Again, both sets of observations may suggest that there is a previously unknown, off-site interaction between carbapenems and other regulatory, transcriptional, or translational mechanisms promoting specific mutation patterns. In some cases, such as A0A143J2U6 (Fig. 4c), which is a distant mutant from all other Class C isoforms that were identified, trends are observed with all carbapenems utilized in this study. This again seems to point towards a pattern in mutation and expression that cannot be explained by the previous model of completely random mutation resulting in greater fitness and thus increased antibiotic resistance. Instead, we consider that there is the possibility that some mutations are more likely than others due to unknown off-target interactions between antibiotics and bacteria that may increase the likelihood of some phenotypes over others.
Our current model of antibiotic resistance is driven, in large part, by random mutation, whereas our proteomic analysis of independent bacterial samples exposed to various carbapenems at three different concentrations demonstrates that some β-lactamase isoforms may be expressed in a dose-dependent manner, calling into question the stochasticity of such mutations. This may suggest that our current model needs revision to include the possibility of off-target interactions or other potentiating factors that favor some mutations over others. Future studies will therefore focus on identifying possible features or properties of these carbapenems that may be associated or correlated with our observed differences in Class C isoform expression as these may also help us determine which drug features may promote specific mutation patterns. This concept, if elucidated, may be of significant benefit to those in drug development as it would offer a new rational tool with which the risk of developing certain phenotypes may be determined.