The antidiabetic drug metformin aids bacteria in hijacking vitamin B12 from the environment through RcdA

Years of use of the antidiabetic drug metformin has long been associated with the risk of vitamin B12 (B12) deficiency in type 2 diabetes (T2D) patients, although the underlying mechanisms are unclear. Accumulating evidence has shown that metformin may exert beneficial effects by altering the metabolism of the gut microbiota, but whether it induces human B12 deficiency via modulation of bacterial activity remains poorly understood. Here, we show that both metformin and the other biguanide drug phenformin markedly elevate the accumulation of B12 in E. coli. By functional and genomic analysis, we demonstrate that both biguanides can significantly increase the expression of B12 transporter genes, and depletions of vital ones, such as tonB, nearly completely abolish the drugs’ effect on bacterial B12 accumulation. Via high-throughput screens in E. coli and C. elegans, we reveal that the TetR-type transcription factor RcdA is required for biguanide-mediated promotion of B12 accumulation and the expressions of B12 transporter genes in bacteria. Together, our study unveils that the antidiabetic drug metformin helps bacteria gather B12 from the environment by increasing the expressions of B12 transporter genes in an RcdA-dependent manner, which may theoretically reduce the B12 supply to T2D patients taking the drug over time.

The discussion section is a major area of improvement for this manuscript. The current discussion is short and largely repeats the contents in the introduction and results. The authors should try to discuss their findings in the context of previous literature. Specifically, the transcription factor RcdA and its potential link to the B12 transporter system should be elaborated.
Other concerns: -P5 L89: it is the tonB gene which was knocked out but not the pathways -P8 L158: in fact, only 4 out of 7 of the B12 transporter genes showed significant results -P8 L161: I think the authors mean "although not for the other tested genes from the B12 transport system". -P9 L178: although not statistically significant, there was actually some noticeable increase in the delta tonB group with B12 and Phen when compared to the WT group. Is it possible that the insignificant result was actually due to a small sample size (N = 4) used? -P11 L225: Figure 2d instead of 2e -P11 L226: is there any possible explanation for the reduction? Also, what does the reduction imply? -P29 Figure 1d: provide the full names of the two B12 forms in the legend -P35 Figure 4b: how about the other 25 targets? Was there any reason why delta yqjC and ymcC were chosen here? -P39 Figure S3b: this figure should be moved to the main text since it shows the effect of metformin, the main target of the study, on the expression of B12 transporter genes.
Reviewer #2 (Remarks to the Author): Using the C. elegans model, Yao et al. present a study describing their finding that the biguanide drug helps E. coli gather vitamin B12 from the environment by increasing the expression of VB12 transporter genes in a RcdA-dependent manner. The result is attention-grabbing; it may help to explain VB12 deficiency in T2D patients with long-term use of the antidiabetic drug metformin. The authors did a very good job of presenting their findings, making the manuscript easy to follow. The study was well-done, but some experiments could be articulated more clearly.
Here are my concerns -1. Before talking about the results in Fig 1b, it would be helpful if the authors could provide some background information about the VB12-sensing GFP worms.
2. The commonly used E. coli strain for C. elegans is OP50, and the authors used OP50 in their first experiment - Fig 1a. But in Fig 1c, they switched to another E. coli strain, BW25113. Different E. coli strains were discussed later, but some explanation is needed here.
3. It would be helpful if the authors could draw simple sketches to elucidate the different treatments of the worms and add them along with Fig 1b and 1c. Based on the description in M&M, my understanding is that in Fig 1b, worms were synchronized in the VB12-deficient CeHR medium, and then L1 stage worms were dropped onto CeHR agar plates which contain 0 or 15 nM VB12, and Veh or Phen; while in Fig 1c, E. coli were incubated with 0 or 1 nM VB12, and with Veh or Phen, for 12 hr, and then dropped onto VB12-deficient NGM plates, and then L1 stage worms were transferred onto the plates. It may be hard to follow for those who have never worked with C. elegans.
4. Could the authors justify the concentrations of metformin and phenformin used in the study? In the fecundity experiment, they used 50 mM metformin and 5 mM phenformin, but in Fig 1c, they used 200 mM metformin and 4 mM phenformin. And how is the concentration used in this study compared to the actual physiological drug level in T2D patients who have been using the drug for an extended period?
5. In Fig 2d, qPCR was performed to measure the expression level of VB12 transport genes +/-4 mM Phen. It's not clear to me how long the bacteria were treated with Phen, was Phen added to the growth medium from the beginning? The authors later showed that 5 mM Phen dramatically inhibited bacteria growth (Fig 4b). Would 4mM Phen cause any growth inhibition? 6. In the growth inhibition assay, 5 mM Phen was added to the medium during inoculation. What is the mode of action of phenformin, bacteriostatic or bactericidal? If 5 mM Phen were added to a log phase or an overnight E. coli culture, would you see any growth inhibition or cell lysis? And, is the growth inhibition strain-dependent (what about OP50)? This is also related to my question 4 -"how is the concentration used in this study compared to the actual physiological drug level in T2D patients who have been using the drug for an extended period ?" If the drug is bactericidal, even if it can stimulate the uptake of VB12 by bacteria, VB12 will eventually be released into the environment due to cell lysis. If the drug is bacteriostatic, the bacterial number will not increase, and the amount of VB12 uptake will reach a plateau, which would not cause a long-term VB12 deficiency.
Reviewer #3 (Remarks to the Author): The manuscript by Luxia Yao et al. estimates an effect of biguanides on the ability of E. coli to increase the accumulation of vitamin B12. This is a carefully designed and well performed study that has led to the identification of the mechanism responsible for the accumulation of B12 in E.coli. The paper is well-written and displays a set of elegant experiments containing the necessary controls to validate the conclusions. Is there a reason why BW25113 is not added to the comparative gene expression in Figure 2c ?

RESPONSE:
We sincerely appreciate the reviewer for finding our work interesting and well conveyed.

RESPONSE:
We thank the reviewer for this important criticism and the recognition that metformin has been included in all crucial experiments. Metformin is costly for high throughput screens, especially when high doses of the drug are needed. Therefore, we used the stronger version of biguanides phenformin rather than metformin in our screens. To address this specific concern as much as possible, we have added metformin

RESPONSE:
The reviewer is correct. We have revised the statement accordingly (Page RESPONSE: This is an outstanding question. Our study applied high doses of biguanides but for a relatively much shorter time in the in vitro experiments, compared to the drug treatment in T2D patients. According to a previous study, metformin level in T2D patients is estimated as 500 ug/mg tissue (3 mM) in the gut (Bailey et al., 2008;Mccreight et al., 2016). Notably, B12 deficiency is commonly observed in patients with long-term metformin treatment, occasionally post years of usage of the drug (de Jager et al., 2010;Chapman et al., 2016). A meaningful note has been added in the revised discussion section (Page 14 Lines 293-295). Fig 2d, qPCR was

RESPONSE:
We apologize for the confusion. Exactly as the reviewer predicted, phenformin was added to the growth medium at the beginning of the culture. The detailed information has been added in the method section (Page 19 Lines 390-395).

Reviewer #2 comment 8: The authors later showed that 5 mM Phen dramatically
inhibited bacteria growth (Fig 4b). Would 4mM Phen cause any growth inhibition? RESPONSE: Phenformin at 4 mM also restricted the growth of bacteria but not as strong as it does at 5 mM, which was used in the screen for the strongest phenforminresistant strains (Main text Fig 4b and 4c, and also see Response to Reviewer #2 comment 11). We have emphasized this point in the revised method section (Page 16 Lines 323-325).

Reviewer #2 comment 9: In the growth inhibition assay, 5 mM Phen was added to the medium during inoculation. What is the mode of action of phenformin, bacteriostatic or bactericidal?
RESPONSE: To answer this specific question, we performed the colony-forming assay with bacteria treated with 5 mM Phen or not according to a modified protocol from the previous study by Balouiri et al (Balouiri et al., 2016). The results showed no significant differences in colony-forming units between the two groups despite a reduction trend present in the phenformin-treated group (Response Fig. 1), indicating that the action of the drug at 5 mM should be bacteriostatic.
Response Fig. 1 Colony-forming assay with BW25113 bacteria treated by 5 mM Phen or not.