Type 2 diabetes T2D is one of the most prevalent chronic diseases and leads to a considerable rate of death and social burden worldwide 1. Patients with T2D with poor glycemic control require insulin therapy in the course of disease progression.

Although good glycemic control can markedly reduce diabetic complications and mortality in hospitalized diabetic patients, it remains challenging and time-consuming to adjust insulin dosages within effective and safe limits 2 , 3. Some treatment regimens may suit some patients better than others or only for some period of time for an individual as their disease condition progresses.

Therefore, personalized and dynamic titration of insulin is of great clinical importance to reduce blood glucose fluctuations and prevent associated comorbidities and mortality in patients with T2D.

Artificial intelligence AI approaches have emerged as potentially powerful tools to aid in disease diagnosis and management 8 , 9 , Existing approaches have used supervised learning SL , in which a list of correct labels must be provided, for disease detection or incidence prediction 11 , However, SL-based methods assume the expert performance to be optimal, which is not always consistent with real-world outcomes due to the complexity of human metabolism and differential responses to drugs among individuals.

Reinforcement learning RL has been proposed as a subfield of machine learning, enabling an agent to learn effective strategies through trial-and-error interactions with a dynamic environment RL could potentially offer an attractive solution for constructing adaptable policies in various healthcare domains, especially in the dynamic treatment regimens DTRs for long-term patient care With the increasing availability of medical record data, RL has been used in sequential medical decision-making systems in various clinical scenarios, including sepsis 15 , coronary heart disease 16 and glycemic regulation by artificial pancreas systems Although several studies have used model-free RL models for treatment recommendation 18 , 19 , 20 , these approaches generally face challenges, such as sample efficiency and potential for unsafe policies when accurate simulation of the environment is lacking 21 , As safety is a primary concern in complex or long-term treatment scenarios, model-based RL might offer potential in simulating diverse scenarios, thereby providing reliable forward planning at decision time Therefore, the incorporation of RL-based methods from development to adoption into the real-world clinical workflow requires comprehensive evaluation Every patient was represented as a temporal sequence of feature vector, including demographics, blood biochemical measurements, medications and insulin usage information.

This model-based RL approach learns the optimal policy by iteratively interacting with the patient model as the environment. Furthermore, we introduced SL to guarantee the safe states by using clinician expertise while optimizing outcomes through trial-and-error interactions with a dynamic environment, which could mimic and potentially augment the physicians in clinical decision-making.

To evaluate our proposed AI system in clinical use 25 , we conducted stepwise clinical evaluations of the AI system in inpatient management from development to deployment, including 1 an internal validation of AI versus physician using both quantitative metrics and qualitative evaluations; 2 an external validation of AI versus physician using qualitative clinical evaluations with test—retest; 3 a prospective deployment study with test—retest; and 4 a final proof-of-concept feasibility clinical trial Fig.

The clinical evaluations indicated that the RL-DITR system could potentially offer benefits in improving glycemic control for inpatients with T2D through dynamic management of subcutaneous insulin injections.

Further investigation in larger, multi-center clinical studies is warranted to demonstrate generalizability of the tool. Left, we constructed a large multi-center EHR dataset consisting of records of long-term continuous clinical observation and medication of hospitalized patients with T2D.

Middle, with the standardized time-series data as input, the patient model generated hidden state transition, status prediction and reward estimation.

Right, the policy model is optimized by interacting with the patient model as an environment. b , Comprehensive evaluation of the AI system step-by-step for integration into the real-world clinical workflow.

Left, we conducted multi-center retrospective studies, including quantitative and qualitative evaluations in the internal and external cohorts. Middle, a prospective study with test—retest was conducted in an academic hospital after AI deployment in the HIS.

Right, a proof-of-concept feasibility trial was conducted to evaluate the glycemic control of and physician satisfaction with the AI system. A total of 12, inpatients with T2D with , treatment days were included in the AI model development phase analysis. The mean age was The demographics and clinical characteristics of patients are presented in Extended Data Table 1.

To represent the patient information into a dynamic evolution process, we processed the patient data into multi-dimensional temporal standardized features.

We used a ClinicalBERT pre-trained model and natural language processing NLP pipeline to extract the clinically relevant sequential features from real-world data Methods.

All the features were discretized to seven timesteps to obtain multi-dimensional temporal features Fig. Specifically, the patient model characterizes diabetes status via a dynamics function and a prediction function.

Given the input of temporal features of a patient trajectory for example, admission status, hospitalized observation and treatment plan from admission to the timestep T, the dynamics function generated the hidden states of the patient. The hidden state is then updated iteratively and subsequently unrolled recurrently for ahead of K steps.

WTR indicates the blood glucose value within the target range 3. Then, we constructed the policy model to make multi-step planning for long-term care. At each step, the policy model is optimized by interacting with the patient model as an environment.

The policy model was trained through a fusion of SL and patient model-based RL with joint learning. Through patient model-based RL, the policy model can learn individualized treatment trajectories and improve long-term clinical outcome.

At the same time, it learns the treatment practices of clinicians in treating patients with T2D within a reasonable range of dosages by SL. To build a dynamic and individualized AI clinician for managing patients with T2D, we constructed the model-based RL framework.

For the comparison of actual state trajectories and model-based state roll-outs, the predicted glucose values follow the transition tendency accurately in both the internal test and the external test set Fig. For the overall glucose prediction, we aggregated the individual-level prediction to produce population-level results, which were then used for further analysis.

The AI model demonstrated good performance in the internal test set, achieving a Pearson correlation coefficient PCC of 0.

When evaluated on the external test set, the AI model achieved a PCC of 0. As shown in Extended Data Table 2 , the results indicate that our model outperformed the other baseline models with a substantial improvement.

a , b , Comparison of actual patient trajectories and model-based state roll-outs for patients from the internal test set a and the external test set b. The blue curve is measured patient glucose values, and the orange curve is predicted glucose values.

c , d , Correlation analysis of the predicted glucose value versus the actual glucose value generated using the AI glucose model in the internal test set c and the external test set d.

Each predicted value is based on the last timestep of the previous day. Box plots show the median center lines , interquartile range hinges and 1. Each value generated by our RL-DITR system represents an individual-level prediction.

These were then aggregated to produce population-level results. AUROC, area under the receiver operating characteristic; ROC, receiver operating characteristic.

When evaluated using the internal test set, the AI model achieved an area under the curve of 0. The model showed reliable performance validated on the external test set. We further investigated the model performance on predicting daily WTR status glucose values within the target range of 3.

We observed that the model becomes more accurate with more information input about a patient as time goes on. We investigated the correlation between the patient outcome WTR ratio and the cumulative rewards estimated by the patient model. The AI model demonstrated good performance with a Spearman correlation coefficient SCC of 0.

We observed that treatment actions with low cumulative rewards were associated with a low rate of WTR ratio, whereas treatments with high cumulative rewards achieved better glucose outcome with a high rate of WTR ratio.

The results show that the patient model evaluation is highly correlated with the clinical outcome and could be used as the interaction environment for the RL model. Figure 3a,b shows the correlation between the clinician policy and the AI policy in the development phase internal and external test sets.

For daily treatment dosage prediction, the AI policy achieved an MAE of 1. We found that the model becomes more accurate as the observed time window expands due to more trial-and-error interactions with the environment. We aggregated the individual-level predictions to obtain population-level results.

c , d , Comparison of actual treatment regimens and model-based treatment roll-outs of two individual patients from the internal test set c and the external test set d. The blue curve is measured patient glucose values, and the orange curve is predicted glucose values given by the AI model.

e , f , Association analysis of the patient outcome for example, WTR versus the dosage difference in treatment actions between the AI policy and the clinician policy for the internal test set e and the external test set f. The dose excess, referring to the difference between the given and the AI model, suggested dose summed over per day for all patients.

R 2 , coefficient of determination. MAPE, mean absolute percentage error. Our proposed approach was then tested against several SL methods, including convolutional neural network CNN , long-short term memory LSTM , transformer and the standard clinical method. We found that our model-based RL method was able to export an accurate treatment regimen and outperformed other methods in the internal test set and the external test set Extended Data Table 3.

The results presented in Extended Data Table 3 demonstrate that our policy model, guided by our blood glucose model, outperformed other models substantially.

Figure 3c,d shows the dynamic treatment strategies generated by clinicians and model-based RL for two individual patients on different hospital days. We further investigated whether the patient outcome WTR ratio varied with the difference of the dose actually administered and the dose suggested by the RL method by correlation analysis Fig.

When the dose actually administered differed from the dose suggested by the AI algorithm, the average outcome got worse. For the internal validation cohort, we compared the performance between our AI system and human physicians in giving insulin dosage recommendation using 40 patients with T2D with insulin data points Extended Data Fig.

RL-generated and physician-generated dosage titrations were evaluated by an expert panel, including quantitative metrics and qualitative metrics from clinical experience. Taking the dosage recommended by the expert panel as references, the MAE of the AI system was 1.

Evaluation was based on the expert panel review including effectiveness f , safety g and overall acceptability h. Orange dashed line represents the average performance of AI; blue dashed line represents the average performance of treating physicians. G, group.

These results suggest that our AI model is superior to junior physicians and similar to experienced physicians in the overall treatment regimen acceptability, hyperglycemia and hypoglycemia control.

Furthermore, we performed an external validation in 45 patients with T2D to compare the performance of AI plans and treating physician plans under a blinded review by an expert panel and by another blinded review for retesting at 2-week intervals at least Extended Data Fig.

The results demonstrated that the acceptability, effectiveness and safety of the AI plans were similar to the treating physicians who were board-certified endocrinologists, evaluated by subjective measurements made by an expert panel Fig.

The percentage of selected superior AI plans was These results demonstrate consistently superior performance of the AI model compared to its physician counterparts. We used adoption rate to evaluate the percentage of the AI regimens adopted by endocrinologists for patient treatment.

Our proposed RL model demonstrated stable performance of effectiveness, safety and acceptability over time, even better in the retest review Fig. The score scale of effectiveness and safety is 1—5.

The adoption rate refers to the percentage of the AI regimens adopted by endocrinologists at the bedside for patient treatment. Intriguingly, a higher adoption rate of Although the adoption rate of the AI plan was relatively low at the initial test review, we found an increase of These results suggested a step-by-step increase of trust of the AI treatment regimen by physicians through human—machine interaction, and the AI system was gradually adopted by physicians into routine clinical practice.

A proof-of-concept feasibility trial was performed to investigate the clinical utility and safety of AI in hospitalized patients with T2D for glycemic control.

Sixteen inpatients with T2D were enrolled in the trial Extended Data Fig. Their mean HbA1c was 8. Over the trial, b , The capillary blood glucose of a patient with T2D during the treatment period.

II Mean daily capillary blood glucose. III Mean preprandial capillary blood glucose. IV Mean postprandial capillary blood glucose during the treatment period.

The preprandial blood glucose target was 5. c , Average percentage of continuous glucose monitoring data within glycemic ranges throughout the treatment period. The satisfaction agreement was scored from a scale of 1—5.

IQR, interquartile range. of At the end of the trial, A patient example of the seven-point capillary blood glucose during the AI intervention is shown in Extended Data Fig. We also used continuous glucose monitoring CGM for the evaluation of the algorithm-directed glycemic control for the secondary outcomes.

The percentage of glucose concentration in time in range TIR 3. TIR 3. Time spent above Time spent below 3. In addition, glycemic variability was slightly decreased during the treatment period coefficient of variation CV of No episodes of severe hypoglycemia that is, requiring clinical intervention or hyperglycemia with ketosis occurred during the trial.

Most physicians stated that the AI interface is understandable 4. In this study, we developed an RL-based AI system, called RL-DITR, for personalized and dynamic insulin dosing for patients with T2D.

We performed development phase validation and clinical validations, including internal validation, comparing AI to physicians using quantitative and qualitative metrics, external validation with test—retest, prospective deployment with test—retest and a proof-of-concept feasibility study with clinical trial.

Taken together, our findings demonstrate that our RL-DITR system has potential as a feasible approach for the optimized management of glycemic control in inpatients with T2D. The management of blood glucose in diabetes remains challenging due to the complexity of human metabolism, which calls for the development of more adaptive and dynamic algorithms for blood glucose regulation.

To address the challenge of personalized insulin titration algorithm for glycemic control, our RL-based architecture is tailored to achieve precise treatment for individual patients, with clinical supervision.

Our proposed patient model-based RL model can make multi-step planning to improve prescription consistency. In addition, because the multi-step plan can be interpreted as the intent of the model from now to a span of time period into the future, it offers a more informative and intuitive signal for interpretation Additionally, our RL-based system delivers continuous and real-time insulin dosage recommendation for patients with T2D who are receiving subcutaneous insulin injection, combining optimal policies for clinical decision-making and the mimicking of experienced physicians Another strength of our study is that we conducted a comprehensive early clinical validation of the AI-based clinical decision-making system across various clinical scenarios.

In clinical deployment, our AI framework offers potential benefits, including automated reading of a large number of inputs from the EHRs, integration of complex data and accessible insulin dosing interface. Although some algorithms have been developed to assist physicians in insulin titration, only a few have been validated in clinical trials 31 , We conducted a proof-of-concept feasibility trial demonstrating the viability of the RL-DITR system in inpatients with T2D.

Notably, the use of the RL-DITR system resulted in a considerable improvement in blood glucose control, meeting our pre-determined feasibility goal.

The percentage of well-controlled blood glucose levels of TIR also demonstrated a substantial increase. Managing hypoglycemia risk is a key consideration for real-world deployment of the AI system.

While achieving improved control of blood glucose levels, the system did not increase the risk of hypoglycemia. Additionally, physicians using the RL-DITR system have reported an increased level of satisfaction, including aspects such as efficiency in clinical practice and perceived effectiveness and safety in glycemic control.

These results suggest that our RL-DITR system has the potential to offer feasible insulin dosing to inpatients with T2D.

A large and multi-center randomized controlled trial would help to determine the efficacy and benefits of this clinical AI solution. Our RL-DITR system was designed as a closed-loop intelligent tool that could use real-time patient data to track blood glucose trajectories and modify treatment regimens accordingly.

Furthermore, the RL-DITR system was developed using EHRs of inpatients with T2D, but its generalizability to other populations, such as outpatients, needs further investigation. We conducted simulated experiments using Gaussian noise to mimic low data quality and dropout 33 to simulate missing data scenarios before deployment Supplementary Fig.

Therefore, although the RL-DITR workflow was implemented and tested for inpatients with T2D, there exists the possibility to extend its application to a wider range of healthcare settings, such as outpatient management, given appropriate integration and continued development.

Although our RL-DITR system has achieved good performance in insulin dosage titration, some challenges remain. The generalization of the AI to other ethnicities needs to be further investigated.

Second, the variety of diet during the hospitalized period was uniformly supplied in the EHRs to build our model. For patients out of hospital, dietary variation and physical activity should be taken into account and explored by our RL model. We have opened an interface to accumulate dietary information for late updated model.

In conclusion, we developed an RL-based clinical decision-making system for dynamic recommendation of dosing that demonstrated feasibility for glycemic control in patients with T2D.

The RL-DITR system is a model-based RL architecture that could enable multi-step planning for patients with long-term care. With the integration of RL structure and supervised knowledge, the RL-DITR system could learn the optimal policy based on non-optimized data while retaining the safe states by balancing exploitation and exploration.

Furthermore, we performed a stepwise validation of the AI system from simulation to deployment and a proof-of-concept feasibility trial. These demonstrate the RL approaches as a potential tool to assist clinicians, especially junior physicians and non-endocrine specialists, with diabetes management in hospitalized patients with T2D.

To train and validate a computational clinical decision support model, we constructed a large multi-center dataset using EHRs of hospitalized patients with T2D who received insulin therapy from January to April in the Department of Endocrinology and Metabolism, Zhongshan Hospital and Qingpu Hospital, in Shanghai, China.

The demographics and clinical characteristics of patients are presented in Extended Data Table 1 , demonstrating a typical T2D population.

We conducted stepwise studies to evaluate the performance of our RL-DITR model version 1. In addition, we performed a proof-of-concept feasibility trial of the RL-DITR system in clinical practice with inpatients with T2D who were admitted for optimization of glycemic control at Zhongshan Hospital ClinicalTrials.

gov: NCT details of proof-of-concept trial protocol provided in Supplementary Information. The retrospective study obtained the following institutional review board IRB approval: Zhongshan Hospital, Shanghai, China R ; XuHui Central Hospital, Shanghai, China and Qingpu Branch of Zhongshan Hospital, Shanghai, China Patient informed consent was waived by the Ethics Committee.

The prospective study and proof-of-concept feasibility trial were approved by the Ethics Committee of Zhongshan Hospital, Fudan University. Each participant provided written informed consent for the prospective study and the proof-of-concept feasibility trial. For time-series data representation, every patient in the dataset was represented as a temporal sequence of feature vectors.

Specifically, each day was broken into seven time periods, including pre-breakfast, post-breakfast, pre-lunch, post-lunch, pre-dinner, post-dinner and pre-bedtime. All records that occurred within the same period were grouped together and formed a feature set to feed into the RL model as input detailed list of the input features provided in Supplementary Table 2.

For structured data, we aligned and normalized them. For free-text notes, we applied a pre-trained language model, ClinicalBERT. Specifically, we first trained the ClinicalBERT on a large corpus of EHR data.

ClinicalBERT is a masked medical domain language model that predicts randomly masked words in a sequence and, hence, can be transformed into downstream tasks. Then, the ClinicalBERT was fine-tuned for information extraction from free text. We further automatically extracted temporal features from patient clinical records, including clinical observations blood glucose records , a sequence of decision rules to determine the course of actions for example, treatment type and insulin dosage titration and clinical assessment of patients.

The numerical values were extracted from demographics, laboratory reports, blood glucose and medications and further translated with standard units according to the LOINC database.

Then, each numerical value was normalized to a standard normal distribution. In terms of discrete values, all the diagnoses of a patient were mapped onto the International Classification of Diseases-9 ICD-9 and used as discrete features, encoded as binary presence features.

We constructed a large multi-center dataset with a large corpus of 1. ClinicalBERT was fine-tuned on a multi-label dataset to extract 40 symptom labels from medical notes.

Phenotype data were extracted from free-text notes of chief history of present illness and physical examination by ClinicalBERT. Validated on 1, annotated samples from the training set, the results showed that ClinicalBERT could accurately identify the symptom information with an average F1 score of Each extracted symptom label was encoded as a binary presence feature.

The process of patient trajectory and treatment decision-making could be formulated as a Markov decision process MDP. An MDP 34 is a tuple S , A , P , G , γ , where S and A are sets containing the states and actions, respectively; P is a transition function; G is a reward function; and γ is a discount factor.

The patient model was learned from historical trajectories, approximating the transition function P and the reward function G and providing support for policy model learning and planning.

The policy model iteratively interacted with the patient model as an environment. At each step, the patient model generated state transition, status prediction and reward estimation based on observed patient trajectories. The policy model, taking the state as input, generated an action that was fed to the patient model.

The patient model updated the states recurrently by an iterative process, enabling the policy model to plan for sequences of actions and find optimal solutions across generated trajectories.

The hidden state would be used as input for patient model and policy model. For patient trajectory tracking, we trained a patient model. When conducting correlation analysis with daily outcome, Magni risk values were summed for each day.

Both of the dynamics function f T and the prediction function f P shared the representation encoder f R when training and inference. f R was optimized together through backpropagation with the loss to capture meaningful patient representations and dynamics.

Each node indicates the states of a patient. The state distribution demonstrated a good cluster hierarchy that individuals in the same cluster are associated with their observable properties diabetes outcome, such as glucose level. Recent evidence suggests that these distinct tissue-specific IR phenotypes may also respond differentially to dietary macronutrient composition with respect to improvements in glucose metabolism.

Objective: The main objective of the PERSON study is to investigate the effects of an optimal vs. suboptimal dietary macronutrient intervention according to tissue-specific IR phenotype on glucose metabolism and other health outcomes. Extensive measurements in a controlled laboratory setting as well as phenotyping in daily life are performed before and after the intervention.

The primary study outcome is the difference in change in disposition index, which is the product of insulin sensitivity and first-phase insulin secretion, between participants who received their hypothesized optimal or suboptimal diet. Discussion: The PERSON study is one of the first randomized clinical trials in the field of precision nutrition to test effects of a more personalized dietary intervention based on IR phenotype.

The results of the PERSON study will contribute knowledge on the effectiveness of targeted nutritional strategies to the emerging field of precision nutrition, and improve our understanding of the complex pathophysiology of whole body and tissue-specific IR.

gov as NCT The prevalence of overweight and related metabolic disturbances, including impaired glucose homeostasis, is rising at an alarming rate, thereby increasing the risk for type 2 diabetes mellitus T2DM and cardiovascular disease CVD 1.

Dietary modulation can effectively lower blood glucose levels and reduce the risk of chronic metabolic diseases, independent of weight loss 2 , 3. Interestingly, there is great heterogeneity in individuals' metabolic response to dietary interventions 4 , 5.

Part of this heterogeneity may be attributed to differences in adherence, but recent findings of large inter-individual variation in postprandial responses to standardized meals indicate that individuals actually respond differently to food 6 , 7.

This inter-individual variation in response to food has complex underpinnings that include biological including genetic , environmental, and lifestyle factors, and may partly explain the differential metabolic impact of dietary interventions 4 — 9.

Whole-body insulin resistance IR reflects defective insulin action in tissues such as skeletal muscle, liver, adipose tissue, gut and brain, and is a major risk factor for T2DM and CVD. IR can develop concurrently in different tissues, but the severity of IR may vary between tissues 10 , Individuals may, for example, have IR predominantly in the liver or skeletal muscle Liver insulin resistance LIR is manifested by impaired insulin-mediated suppression of hepatic glucose production HGP , while muscle insulin resistance MIR is characterized by decreased insulin-mediated glucose disposal The gold-standard method to quantify LIR and MIR is the two-step hyperinsulinemic-euglycemic clamp Tissue-specific IR can also be modeled based on glucose and insulin responses during an oral glucose tolerance test OGTT , which has been validated against the clamp technique 10 , These tissue-specific IR phenotypes have previously been linked to distinct metabolic profiles, representing different etiologies toward T2DM and CVD 11 , 13 — More specifically, greater disturbances in circulating lipidome 13 and metabolome profiles 14 have been found in individuals with more pronounced LIR as compared to individuals with more pronounced MIR.

Additionally, in individuals with LIR, abdominal subcutaneous adipose tissue scAT has been characterized by higher expression of genes related to extracellular modeling, whilst MIR has been associated with higher expression of genes related to inflammation in scAT, as well as higher levels of circulating plasma markers of systemic low-grade inflammation Recent findings indicate that these distinct metabolic phenotypes may respond differently to dietary macronutrient manipulation with regard to outcomes of glucose homeostasis, ectopic fat deposition, and tissue-specific lipid metabolism amongst others 15 , Therefore, further characterization of these IR phenotypes as well as studying these metabolic phenotypes in relation to dietary intervention outcomes may be a promising strategy to develop more personalized dietary interventions.

In addition, improvement of glycemic control by more personalized dietary interventions may enhance mood, self-control, and cognitive function 1 , 19 — Such short-term benefits may in turn increase adherence to a healthy diet. Importantly, prospective randomized controlled trials with a pre-specified hypothesis on differential metabolic responses to diets based on metabolic phenotype are largely lacking in the emerging field of precision nutrition.

The PERSonalized glucose Optimization through Nutritional intervention PERSON study was designed to investigate the effects of an optimal compared to a suboptimal dietary intervention according to tissue-specific IR phenotype on glucose metabolism and other metabolic health outcomes.

This two-center, week dietary intervention study with a randomized, double-blind, parallel design, aims to enroll a total of individuals with either LIR or MIR. Individuals are randomized to follow one of two diets that are hypothesized to target one of the two tissue-specific IR phenotypes.

Before and after the week dietary intervention, individuals are extensively phenotyped both in laboratory settings and in daily life.

The extensive phenotyping performed in this unique clinical trial allows for a comprehensive study of both the complex metabolic and lifestyle determinants of glucose homeostasis, as well as the dietary intervention effects on metabolic health and its metabolic underpinnings.

In the present article, we describe the study design and measurements in detail, and present preliminary results of the screening population. gov identifier NCT The study is conducted according to the principles of the Declaration of Helsinki revised version, , Fortaleza, Brazil , and all subjects provide written informed consent before the start of the study.

Figure 1. Study design of the PERSON study. Tissue-specific insulin resistance MIR, muscle insulin resistance; LIR, liver insulin resistance is assessed at screening using a 7-point oral glucose tolerance test and eligible participants with MIR or LIR are randomized to follow either their hypothesized optimal dark purple or suboptimal light purple diet for 12 weeks.

BMI, body mass index; MUFA, monounsaturated fatty acid. Secondary outcome parameters include whole-body and tissue-specific insulin sensitivity and glucose homeostasis, fasting and postprandial metabolic profile, vascular health, fecal microbiota composition and functionality, body fat distribution, ectopic fat accumulation, adipose tissue morphology and gene expression, skeletal muscle protein and gene expression, fasting immune metabolism, cognitive performance, and perceived well-being.

From May onwards, subjects have been recruited via a volunteer database, flyers, and advertisements in local and online media.

Eligibility is assessed during a screening visit. Body weight and height are measured in duplicate without shoes and heavy clothing to the nearest 0. Waist and hip circumference are measured in duplicate to the nearest 0.

Blood pressure is measured in triplicate on the non-dominant arm with an automated sphygmomanometer after a 5-min rest with the subject in a supine position. The first measurement is used to acclimatize the subject to the measurements, and therefore omitted from the data. Tissue-specific insulin resistance is assessed based on the glucose and insulin responses during a 7-point OGTT.

Hepatic IR and muscle insulin sensitivity are estimated using the calculations of Abdul-Ghani and colleagues We have recently optimized the MISI calculator using the cubic spline method The hepatic IR index HIRI and muscle insulin sensitivity index MISI are calculated according to the following formulas:.

Both indices were developed and validated against gold standard measurements of tissue-specific IR by a hyperinsulinemic-euglycemic clamp 10 , The lowest tertile of MISI represents individuals with MIR, while the highest tertile of HIRI represents individuals with LIR.

The cutoffs for these tertiles are based on values of a selected study population of The Maastricht Study DMS 22 , which resembles the target population of the PERSON study. From the OGTT, incremental area under the curve iAUC is calculated for both glucose and insulin using GraphPad Prism software version 5.

Only values above the fasting value are included in the iAUC. Habitual dietary intake is estimated by a validated item semiquantitative food frequency questionnaire FFQ Dietary misreporting is evaluated by Goldberg's method, using the ratio of daily energy intake EI to estimated basal metabolic rate BMR 27 , Education level is categorized into low no education, primary education, lower or preparatory vocational education, lower general secondary education , medium intermediate vocational education, higher general senior secondary education or pre-university secondary education and high higher vocational education, university.

Perceived chronic stress is assessed with the Long-term Difficulties Inventory 29 and mental well-being with the RAND Item Short Form Health Survey RAND 30 and the Social Production Function Instrument for the Level of Well-being Eligible subjects are randomly allocated to either their hypothesized optimal or suboptimal diet by an independent analyst using center-specific minimization 32 , 33 with randomization factors of 1.

Both researchers and participants are blinded to the participants' metabolic phenotype, and thus blinded to whether participants are allocated to their hypothesized optimal or suboptimal diet. Participants start the study within 3 months of the screening visit. The hypothesized optimal diet for LIR is low in fat, and high in protein LFHP and fiber.

Energy from CHO is similar between diets. The dietary intervention strategy is based on intensive dietary counseling and provision of key products. Before the start of the intervention, a short dietary history is performed to assess the participants' dietary habits and preferences.

This information is used to individualize the dietary plan and counseling accordingly. At the start of the intervention period, participants receive verbal and written instructions on their dietary plan, which lists both types and quantities of foods that they are required to consume daily or weekly in order to meet the targeted nutrient composition of the assigned diet.

The instructions include guidance on what types of foods to choose and avoid within all food groups e. Intake of so-called free-food items e. Key products that largely distinguish the two diets with regards to macronutrient composition are provided in pre-measured amounts. For the HMUFA diet, key products include olive oil, olives, olive tapenade, and low-fat margarine with olive oil.

Key products for the LFHP diet include low-fat yogurt and quark, reduced-fat cheese, very low-fat spread, pumpkin seeds, baking margarine with olive oil, and a dietary fiber supplement 2 g β-glucan per 6 g, PromOat ® , DSM Nutritional Products, Basel, Switzerland providing 6—12 g of additional fiber per day.

Participants are instructed to finish a certain amount of every provided product each day. Apart from the fiber supplement, all products are commercially available. Throughout the intervention period, participants visit the research facilities every week for a to min individual dietary counseling session with a dietitian or research nutritionist to monitor diet adherence, body weight, and adverse events using a semi-structured interview.

These sessions are supported by advice via e-mail or telephone if needed. To be able to assess the effects of the dietary intervention on metabolic health parameters, independent of changes in body weight, we aim to keep participants on a stable body weight throughout the study.

In case of weight loss or gain, participants are reassigned to a higher or lower energy group to prevent further weight change. To promote overall diet adherence, participants are allowed to deviate from their dietary plan on three individual days throughout week 2—10 of the intervention period.

Participants are asked to keep a food record FR on these days. During the COVID restrictions, the weekly on-site visits are replaced by telephone or video-call consultations, key products are home-delivered by courier, and participants weigh themselves at home.

Participants are provided with written and face-to-face instructions on how to record dietary intake. Participants that do not have a smartphone complete the FRs on paper, which are later entered into the app by the researcher.

This week includes three or four depending on study center and participation in additional subgroup measurements clinical test days and three at-home days. Participants wear a continuous glucose monitor CGM and activity monitor throughout the characterization week.

During the clinical test days, participants undergo extensive laboratory testing, which includes challenge tests, body composition analysis, vascular measurements, tissue biopsies, a cognitive test, and questionnaires.

During the at-home days, participants record dietary intake and feelings of well-being, consume various standardized meals, and collect feces and urine. An overview of all measurements can be found in Figures 2 , 3 and are described in more detail below.

Figure 2. Graphical overview of the pre- and post-intervention characterization week. The characterization week contains multiple clinical test days, during which participants are extensively phenotyped. In addition, blood glucose and physical activity are continuously monitored.

At home, participants record their dietary intake and feelings of well-being, collect a fecal sample and h urine, and consume a standardized breakfast on day 4, and on day 5, participants have a full day of standardized meals and snacks, including the standardized breakfast.

In a subgroup of the study population, additional measurements are performed. DXA, dual-energy X-ray absorptiometry; MRI, magnetic resonance imaging; 1 H-MRS, proton magnetic resonance spectroscopy; AGEs, advanced glycation endproducts; CAR, carotid artery reactivity; OGTT, oral glucose tolerance test; GI, gastrointestinal; scAT, subcutaneous adipose tissue; SM, skeletal muscle.

Figure 3. Overview of all measurements performed within the PERSON study. SCR, screening visit; CW, characterization week; DIW, dietary intervention week.

On the clinical test days, participants are instructed to travel to the facility by car or public transport. The day prior to and during the characterization weeks, participants are requested to refrain from alcohol and vigorous physical activity. The macaroni meal is prepared in the university kitchen.

Figure 4. Graphical overview of the oral glucose tolerance test and the high-fat mixed-meal HFMM test that are performed during the pre- and post-intervention characterization week. Participants are instructed to drink the glucose drink or HFMM within 5 min, and fasting and postprandial blood samples are drawn at the indicated timepoints for determination of the indicated metabolites.

CHO, carbohydrates; HbA1c, hemoglobin A1c; FFA, free fatty acids; TAG, triglycerides; GLP-1, glucagon-like peptide 1; PYY, peptide YY; NMR, nuclear magnetic resonance; HDL, high-density lipoprotein; SCFA, short-chain fatty acids; PBMCs, peripheral blood mononuclear cells.

On a separate clinical test day, at least 4 days after the OGTT, a high-fat mixed-meal HFMM challenge test is performed after a h overnight fast Figures 3 , 4.

Participants again consume the standardized low-fat macaroni meal the evening before the test. The liquid HFMM g containing 2. An intravenous cannula is inserted in the antecubital vein for blood sampling.

Total cholesterol and HDL cholesterol are determined in fasting serum. Buffy coat is collected from fasting blood for later DNA isolation and genotyping. Blood pressure is assessed according to the same procedures used at screening.

In a subgroup of participants, vascular function is assessed by measuring carotid artery reactivity CAR to a cold pressor test CPT The diameter of the left common carotid artery is monitored during a 1-min baseline assessment and continuously during the 3-min CPT using ultrasound Terason uSmart , Burlington, MA, USA.

Wall-tracking and edge-detecting software is used to calculate the diameter after completion of the test. To confirm sympathetic stimulation, blood pressure is measured after the supine rest, 1-min and 2- min after the start of the CPT, and directly after completion of the CPT Omron M6 Comfort, Omron healthcare Co.

In a subgroup, skin accumulation of advanced glycation end-products AGE is measured by skin autofluorescence AF using the automated AGE reader DiagnOptics Technologies B.

Skin AF is measured at three slightly different places on the volar side of the dominant arm, avoiding impurities of the skin such as scars and birthmarks.

Participants are instructed to not apply any creams, lotions, or sunscreen on their arms on the day of the measurement.

Analyses are performed using a computational modeling method [AMRA Medical AB, Linköping, Sweden 40 ] for quantification of abdominal subcutaneous adipose tissue ASAT , visceral adipose tissue VAT , thigh muscle volume, intrahepatic lipid content IHL , and muscle fat infiltration MFI in the anterior thighs Figure 3.

At WUR, IHL and abdominal fat distribution are assessed with proton magnetic resonance spectroscopy 1 H-MRS and MRI, respectively, on a 3T whole-body scanner Siemens, Munich, Germany; Philips Healthcare, Best, the Netherlands from November onwards.

Spectra for determination of IHL are obtained from a 30 × 30 × 20 mm voxel placed in the right lobe of the liver, avoiding blood vessels and bile ducts.

Participants are instructed to hold their breath when spectra are acquired to reduce respiratory motion artifacts. Spectra are post-processed and analyzed using the AMARES algorithm in jMRUI software.

Abdominal fat distribution is evaluated as subcutaneous ASAT and visceral adipose tissue VAT areas in the abdomen, which are quantified in singles-slice axial T1-weighted spin echo transverse images at the inter-vertebral space L3-L4 using the semi-automatic software program HippoFatTM During one of the at-home days in the characterization week, participants collect fecal samples Figures 2 , 3.

The samples are stored in the participants' home freezer for maximal 72 h before the visit to the research facilities. Participants rate stool consistency of the sample using the Bristol stool scale Fecal microbiota composition is determined by 16S rRNA sequencing as described elsewhere During the HFMM challenge test, fasting and postprandial blood samples are collected for determination of plasma concentrations of GLP-1 and PYY Figure 3.

Fecal concentrations and fasting plasma levels of gut microbiota-derived short-chain fatty acids SCFA acetate, propionate and butyrate are determined using optimized LC-MS protocols Data on self-reported gastrointestinal health are collected by a questionnaire based on the Rome III criteria The questionnaire includes questions on presence of gastrointestinal complaints i.

In addition, oral samples are collected for microbiological and metabolite analyses. Participants are asked to rinse the oral cavity thoroughly for 30 s with 10 ml of sterile 0.

Participants are instructed to refrain from oral hygiene in the morning of the sampling day. The composition of the oral microbiome is determined by 16S rRNA sequencing The samples are washed with saline to remove blood clots.

In a subgroup of participants, at baseline only, ~0. In short, the stromal vascular fraction is isolated from the AT and stained with a cocktail of antibodies for flow cytometry for identification of immune cells The skeletal muscle biopsy is taken from the m.

vastus lateralis under local anesthesia using the Bergström biopsy needle method After the SM biopsy, whole-body and tissue-specific insulin sensitivity are assessed by the gold standard two-step hyperinsulinemic-euglycemic clamp Arterialized blood is frequently sampled from the superficial dorsal hand vein during the insulin infusion to measure glucose concentrations, which are maintained at ~5.

Substrate utilization is measured for 30 min during the basal, low insulin, and high insulin infusion using indirect calorimetry by ventilated hood Omnical, Maastricht Instruments, Maastricht. Resting metabolic rate RMR , fat and carbohydrate oxidation are calculated according to the equations of Weir and Frayn 50 , At WUR only, circulating peripheral blood mononuclear cells PBMCs are isolated from fasted blood samples collected at the HFMM test Figure 3.

In addition, in a random subgroup n ~ , PBMCs are also isolated from fasted blood samples collected at screening. PBMCs are isolated by density gradient isolation using CPT tubes BD vacutainer, cat. Monocytes are subsequently obtained by MACS magnetic activated cell sorting positive selection using CD14 MicroBeads Miltenyi Biotec, cat no.

DY; DY; DY The metabolic potential of monocytes is measured in real-time experiments inflammatory cell activation test and glycolytic stress test using the Seahorse apparatus Agilent Technologies in screening samples only.

Urine collection starts after the first voiding on the morning of the home-day with only standardized meals and finishes 24 h later on the morning of the HFMM. Participants are asked to store the containers in a cool place, preferably a refrigerator, and bring the containers to the facilities on the day of the HFMM.

At the start of the characterization week, a CGM Medtronic iPro2 with Enlite sensor is placed lateral to the umbilicus for 6 days of continuous interstitial fluid glucose measurements Figure 2. The CGM data are calibrated according to the manufacturer's instructions with four daily capillary glucose self-measurements using a blood glucose meter Contour XT, Ascensia Diabetes Care.

To assess glycemic variability and glycemic responses to standardized meals, on one of the home-days, participants consume a standardized breakfast, and on another home-day, participants have a full day of standardized meals and snacks, including the standardized breakfast Figure 2 ; Supplementary Tables 3, 4.

Participants are instructed to consume the meals according to standardized instructions including time frames, to fast for 2 h after the breakfast, and to only drink water alongside the standardized meals.

Participants keep a diary to record the times they go to sleep and wake up while wearing the monitor. Sedentary and physical activity parameters are quantified with a modified version of the script of Winkler et al.

In addition, participants are asked to report on hunger, mood, and sleepiness every 2 h from to h Figure 2. Executive function is evaluated with the multitasking test and spatial span test; memory with the delayed matching to sample test and paired associates learning test; and attention and psychomotor speed is assessed with the motor screening task and reaction time task.

Each test is preceded by standardized instructions and a practice round for familiarization. After the CANTAB, participants complete the computer-based Macronutrient and Taste Preference Ranking Task MTPRT for assessment of food preferences The task assesses liking and ranking for 32 food products that are categorized as high in carbohydrates, high in fat, high in protein, or low-calorie, as well as either sweet or savory.

In addition, during one of the clinical test days, participants provide information on general well-being, sleep characteristics, and physical in activity by questionnaire Figure 3.

Mental well-being is assessed using the RAND 30 and perceived stress is measured with the item Perceived Stress Scale PSS Physical and mental fatigue are assessed using the item Chalder fatigue scale Sleep quality is assessed with the item Pittsburgh Sleep Quality Index 60 and sleep duration and chronotype are derived from the Munich ChronoType Questionnaire Daytime sleepiness is assessed with the 8-item Epworth Sleepiness scale 62 Figure 3.

A wide range of biological samples are collected in the present study, including blood plasma and serum, SAT, SM tissue, feces, urine, saliva, and PBMCs. EDTA Becton Dickinson, Eysins, Switzerland tubes are centrifuged at 1, g, 4°C for 10 min and plasma is aliquoted subsequently. Serum tubes are left at room temperature for at least 30 min to allow clotting after sampling and centrifuged at 1, g, 20°C for 10 min before aliquoting of serum.

Samples from both centers are analyzed at central laboratories. Plasma glucose, insulin, and FFA are measured on a Cobas Pentra C using ABX Pentra Glucose HK CP reagens Horiba ABX Diagnostics, Montpellier, France , ELISA Meso Scale Discovery, Gaithersburg, USA , and NEFA HR 2 reagens 2 Wako chemicals, Neuss, Germany , respectively.

Serum TAG, total cholesterol, and HDL cholesterol are measured on a Cobas Pentra C using ABX Pentra Triglycerides HK CP reagens, ABX Pentra Cholesterol CP reagens, and ABX Pentra HDL Direct, respectively. During the HFMM challenge test, fasting and postprandial blood samples are collected in EDTA tubes and aprotinin tubes containing dipeptidyl peptidase-IV inhibitor Milipore Merck, Billerica, MA, USA for determination of plasma GLP-1 and PYY, respectively.

Total GLP-1 immunoreactivity is assessed using an antiserum that reacts equally with intact GLP-1 and the primary N-terminally truncated metabolite as previously described PYY concentrations are determined with a commercially available radioimmunoassay for Human PYY Millipore Corporation, MA, USA.

Data are collected on paper case report forms CRF and are entered in an electronic CRF designed for the study, using the web-based data capturing platform Caster EDC 68 that is compliant with good clinical practice GCP requirements.

All relevant raw and processed data e. We show in Fig. The predicted dynamics closely follow the measured lac levels obtained from minutes IPTG induction cyan line. The highlighted regions in Fig. The modeling results support a picture in which response memory provides a large adaptive advantage when external fluctuations occur faster than the cell division time, while phenotypic memory is beneficial for slower fluctuations, spanning several generations.

We have presented two distinct memory mechanisms in the lac operon of E. coli , phenotypic and response memory, each of which is beneficial over different timescales. Phenotypic memory allows cells to maintain an adapted state for multiple generations after a specific carbon source is removed from the environment.

Since phenotypic memory operates through the transmission of stable cytoplasmic proteins, it may be employed as a general strategy in other organisms to transmit metabolic information between generations, as observed e. in the yeast galactose system [9] , [36]. More generally, the intrinsic mechanism behind phenotypic memory being passive — based on intracellular proteins whose lifetime is longer than a typical generation — similar memory effects are expected to be present for other fluctuations and in other organisms.

Adaptation mechanisms that rely on the expression of long-lived permease molecules — e. We used fast fluctuating environments to dissect the determinants of lag phases following a transition from glucose to lactose. Our results suggest a simple biological model of the lag phase in which lac protein activity and the stringent response are mutually inhibitory processes: Lac protein activity in lactose has an inhibitory effect on the stringent response due to glucose production and amino acid synthesis, while the stringent response initially inhibits lac protein production through its global inhibitory effects on translation.

To see this, we consider two examples. First, we compare for uninduced and pre-induced cells Fig. In the pre-induced case, after the first lactose exposure cells rapidly recover full growth in glucose, whereas if no lac proteins are initially available, cells experience a slow recovery in glucose.

The stringent response due to the lactose exposure is therefore much less severe when a small amount of LacZ induced is available to hydrolyze lactose and initiate positive autoregulation. Second, we note that for Fig. However, for Fig. We conclude that protein production during the stringent response is too slow to allow cells to cross the threshold during the short lactose exposures for.

In particular, if lactose encounters unexpectedly cease, this cost will no longer be temporary, but sustained by the population indefinitely. Cells employing response memory avoid such long-term cost by transiently expressing the required genes for a short amount of time following an initial exposure to the stimulus, with a maximal metabolic cost that is limited by the duration of this transient expression.

Should environmental fluctuations cease, cells will suffer only a small, short-term fitness cost. On the other hand, should fast fluctuations persist, as we have shown the cells reap a significant fitness benefit. In particular, we showed that cells reach higher induction levels more rapidly by maintaining their response profile following the removal of an external inducer.

Memory in different genetic network architectures could affect not only the cost of gene expression, but also the evolution of gene expression levels. The timescale over which phenotypic memory persists is determined to a large extent by the gene's expression level provided the protein is sufficiently stable.

Expression levels may be evolutionarily tuned not only to support growth in a single environment, but also to provide cells' progeny with memory of past environments. The interplay of memory and metabolic constraints could thus dramatically change the nature of evolutionary trajectories and optima.

We expect theoretical analyses may be fruitfully applied to explore these possibilities. The power of the memory mechanisms we have described lies in their universality.

Protein lifetimes and regulatory networks can be tuned in simple ways to give rise to physiological memory under rapidly changing conditions. Microorganisms have to handle both internal and external sources of noise, and while many genetic networks have evolved to exploit stochastic fluctuations of intracellular molecular components to regulate key cellular processes [41] , we have shown that molecular rates of signal transduction reactions can be modulated to optimize response profiles for growth in fluctuating environments.

Together, phenotypic and response memory allow bacteria to adapt to a wide range of fluctuation timescales in sophisticated, history-dependent ways. These memory mechanisms constitute general strategies that bacteria can employ to adapt to diverse environmental fluctuations — including nutrients, antibiotics, and other physiological stresses.

The microfluidic device used in this study was made using standard soft lithography and microfabrication techniques and consists of growth chambers and a main flow channel patterned from two SU-8 layers 1.

The devices were fabricated by making polydimethylsiloxane PDMS replicates of the SU-8 master. The PDMS devices were peeled from the silicon master and 1.

When transitioning between two media, both valves were closed for 15 seconds before the new one was opened to let the pressure equilibrate inside the device and to avoid backflow problems.

A T-junction upstream of the growth chambers ensured that transitions between the different media occurred very rapidly. By flowing a fluorescent dye inside the device, the transition between each type of media was measured to occur in less than milliseconds Supp.

Cells inside the growth chambers push their immediate neighbors toward the main flow channel as they increase in size, and the lateral speed at which the cells move is proportional to the population's mean elongation rate.

An optical flow algorithm implemented using openCV [42] was used to measure the displacement between successive frames. This displacement was used to find the average cell speed over the region between 10 and 15 away from the closed end of the growth chamber.

The lateral speed reports on the cumulative growth rate of cells in the first 10 microns of the growth chamber providing a measure of the relative growth rate of the population. Error bars on relative growth plots report the standard error of the mean as averaged over the 5 chambers present in a single field of view.

These error bars measure intrinsic cell-to-cell variability in growth, due to stochasticity in cell division rates, elongation rates, and gene expression processes. The lac induction dynamics of a population subjected to sudden environmental changes are modeled as described in [35] , with an additional equation to account for mRNA transcription.

The model assumes that LacY protein levels are proportional to LacZ levels. Unless otherwise noted in Table 1 , refer to [35] for a complete rationalization behind each parameter's value. The set of equations are given by 1 2 3 4 where , , , and are the intracellular concentrations of lactose, allolactose, mRNA and LacZ proteins, respectively parameters are specified in Table 1.

The expression for the lactose hydrolysis rate is given by 5 where , , and are obtained by solving Eqns. To qualitatively compare behaviors with and without response memory, we artificially reduced the rate of allolactose turnover in glucose environments taking to attain response memory in our simple model.

Similar results were obtained by reducing instead the mRNA degradation rate in the transition from lactose to glucose. Full methods as well as further details of microfluidic fabrication, strain description, image acquisition and analysis, and any associated references are provided in Text S1.

Timescale of the environmental change inside a chemoflux. The transition are accurately described by exponential functions red lines, seconds and seconds. Growth rate measurement. Since each cell division event yields two cells at age zero, the fraction of cells at age 0 is twice the population's growth rate.

Duration of the lag phase. B - F The duration of the lag and recovery phases is computed from a linear regression of the lateral cell speed and the results are presented in Fig. S1 for quantification. A video of cell growth in the chemoflux growth chambers during a glucose-to-lactose transition.

We thank Calin Guet, Yuichi Wakamoto, Michael Rust, David Gresham, Joao Xavier, Matthew Eames, and Wei-Hsiang Lin for comments on the manuscript. We also thank Jose Vilar, Gene Huber, and Bruce Levin for discussions. We especially thank Tobias Bergmiller and Calin Guet for the over-expression plasmids, Wei-Hsiang Lin for his assistance in carrying out the constitutive Lac expression experiments, and Michael Rust for providing a microscopy setup for additional experiments.

We thank the Wakamoto and Xie labs for providing E. coli strains. Conceived and designed the experiments: GL EK. Performed the experiments: GL.

Analyzed the data: GL. Interpreted the data: GL EK. Wrote the paper: GL EK. Article Authors Metrics Comments Media Coverage Reader Comments Figures. Correction 16 Oct The PLOS Genetics Staff Correction: Memory and Fitness Optimization of Bacteria under Fluctuating Environments.

Abstract Bacteria prudently regulate their metabolic phenotypes by sensing the availability of specific nutrients, expressing the required genes for their metabolism, and repressing them after specific metabolites are depleted.

Author Summary Bacterial adaptation to new environments typically involves reorganization of gene expression that temporarily decreases growth rates. Introduction Escherichia coli cells grown in the presence of both glucose and lactose first consume glucose, which is more easily metabolized, before expressing the genes necessary for lactose catabolism [1] — [3].

Download: PPT. Figure 1. Chemoflux device for growth rate measurement in changing environments. Results Phenotypic memory in response to sudden environmental changes The growth rate of E.

Figure 3. Molecular components of phenotypic memory in the lac operon. Figure 4. Lag phase and recovery in rapidly fluctuating environments. Response memory and dynamics of lac protein expression To reproduce the conditions of Fig.

Figure 5. In vivo measurement of LacY expression in fluctuating environments. Modeling lac operon dynamics with memory in a fluctuating environment While the major determinants of the lag phase were found to be the initial induction steps and the recovery from stringent response, the potential fitness gains that cells might reap from response memory remained unclear.

Figure 6. Mathematical modeling quantifies fitness advantage of memory. Discussion We have presented two distinct memory mechanisms in the lac operon of E.

Materials and Methods Device description and fabrication The microfluidic device used in this study was made using standard soft lithography and microfabrication techniques and consists of growth chambers and a main flow channel patterned from two SU-8 layers 1.

Growth rate measurements Cells inside the growth chambers push their immediate neighbors toward the main flow channel as they increase in size, and the lateral speed at which the cells move is proportional to the population's mean elongation rate. Mathematical model of lactose metabolism The lac induction dynamics of a population subjected to sudden environmental changes are modeled as described in [35] , with an additional equation to account for mRNA transcription.

Supporting Information. Figure S1. s PDF. Figure S2. Figure S3. Text S1. Supplementary methods. Video S1. s MOV. Video S2. Acknowledgments We thank Calin Guet, Yuichi Wakamoto, Michael Rust, David Gresham, Joao Xavier, Matthew Eames, and Wei-Hsiang Lin for comments on the manuscript.

Author Contributions Conceived and designed the experiments: GL EK. References 1. Monod J Recherches sur la croissance des cultures bactériennes. Monod J The growth of bacterial cultures.

Annu Rev Microbiol 3: — View Article Google Scholar 3. Dekel E, Alon U Optimality and evolutionary tuning of the expression level of a protein.

Nature —

Overall, both Simulink TM simulations and experimental results demonstrated that the GMPC approach provided more robust and precise control than traditional PID controllers.

While the model could anticipate the future behavior of the fermentation and take appropriate control action, the PID controller did not have this capability resulting in oscillations and overshoot behavior in both simulations and experiments.

Thus, our study demonstrates how GMPC systems can serve as a bridge between genome-scale metabolic modeling and control algorithms. Since the cultivation conditions can change and affect algal cellular metabolism, our system connected feedback measurements with genome-scale metabolic models and achieved more efficient nutrient utilization and higher product yields for dynamic algal cultivation conditions.

In this way, genome-scale metabolic models can be effectively utilized to improve biomanufacturing of microalgae and other industrially important microbial cell factories. Fed-batch cultivation and PID controllers have been widely used in bioprocess development. Unfortunately, fed-batch cultivation often results in poor nutrient control and wasted nutrients and conventional PID control can lead to oscillating cell behaviors and poor performance under dynamic conditions.

In this study, we have utilized the power of genome-scale metabolic models to predict and control glucose and nitrate supply for C. vulgaris cultures under light and dark cycles and compared this approach to conventional autotrophic and heterotrophic processes.

Our results first showed that utilizing genome scale models to track and limit glucose and nitrate feeding led to higher titers of biomass, FAs, and lutein than autotrophic conditions and more efficient glucose utilization and higher product yields than heterotrophic conditions.

Next, implementing these models into an open loop system modestly improved performance. Finally, both computational simulations and experimental results demonstrated that this genome-based MPC system exhibits superior controller performance compared to conventional PID methods.

Green microalgae C. vulgaris UTEX was obtained from the Culture Collection of Algae at the University of Texas at Austin and maintained on sterile agar plates 1. Liquid cultures were inoculated with a single colony in For alternating light and dark cycles, autotrophic conditions were used for light sections and heterotrophic conditions were used for dark sections.

The lyophilized algal dry biomass was weighted gravimetrically using an analytical balance. The glucose concentration was measured using YSI biochemistry analyzer Yellow Springs, OH. FAME production followed the procedure provided by Dong et al.

Helium was used as carrier gas. Lutein extraction followed the procedure provided by Yuan et al. The solution was filtered before HPLC analysis. The mobile phases are eluent A dichloromethane: methanol: acetonitrile: water, 5. The i CZ model, including six different biomass compositions for autotrophic conditions PAT1-PAT6 and five different biomass compositions for heterotrophic conditions HT1-HT5 , was obtained from Zuniga et al.

GSM simulations were performed using the Gurobi Optimizer Version 5. The experimental setup is shown in Supplementary Fig. The manipulated variables were glucose demand F G and nitrate demand on a per L basis F N for 8-h period.

Two pumps were used to control both variables automatically by Matlab TM through Arduino chip. All the control algorithms were run on Matlab TM and the codes are provided in Supplementary information.

The Simulink TM simulation is shown in Fig. The blue box in Fig. Four equations were built inside the blue box as shown in Supplementary Fig.

The inputs were F G and F N. The outputs were biomass, nitrate level, glucose level and volume. Only nitrate levels and glucose levels were fed into the PID and GMPC controller.

For the proportional-integral-derivative PID controller, the proportional gain K p , integral gain K i and derivative gain K d equal to 1.

The PID controller and GMPC controller were used to control glucose supply and nitrate supply every hour in both simulation and experiment. Changes in the setpoint for glucose were introduced to see how both PID and GMPC responded to those changes. Initial biomass levels x 0 , glucose levels G 0 and nitrate levels N 0 were measured as described above and used as inputs into the open-loop system.

Three equations shown below were used to predict biomass growth, nitrate consumption rate, and glucose consumption rate in the open-loop system. The growth rates under light and dark cycles were determined based on previous experimental data. After that, the growth rates were constrained in the autotrophic and heterotrophic GSMs, respectively to determine nutrient exchange rates r N and r G under light and dark cycles.

The methods for using growth rate to estimate nutrient exchange rates have been described previously in Chen et al. We assumed a rapid switch to a new operational steady state following the transition between light and dark cycles. Initial biomass levels x 0 , glucose levels G 0 and nitrate levels N 0 were measured and used as inputs into the closed-loop system.

During the experiment, biomass levels x m , glucose levels G m and nitrate levels N m were msured and used as inputs into the closed-loop system. For the light cycle, two equations were built to describe and predict biomass accumulation rate and nitrate consumption rate.

Unlike the open loop system, the light shielding effect was considered and the growth rate would decrease as the biomass concentration increased as described in the equation below and shown in Fig.

The GSM was used to predict nutrient exchange rate r N based on the measured growth rate. For the dark cycles, three model equations were built to predict biomass accumulation rate, nitrate consumption rate and glucose consumption rate as listed below and shown in Fig.

In the biomass equation, we assumed a fraction of heterotrophic biomass, a , was derived from autotrophic metabolism and the simulated growth rate was μ A. Meanwhile, some biomass was derived through heterotrophic metabolism with the simulated growth rate, μ H.

The nutrient exchange rates r NA , r NH , r GH were determined by inputting simulated growth rates into the autotrophic and heterotrophic GSMs respectively. where μ A is simulation growth rate from autotrophic metabolism, μ H is the growth rate from heterotrophic metabolism, r NA is nitrate exchange rate from autotrophic metabolism, r NH is the nitrate exchange rate from heterotrophic metabolism, r GH is the glucose exchange rate from heterotrophic metabolism.

Next, we applied a fitting objective function J to minimize the difference between calculated values and simulated model values in order to estimate the optimal parameter values a , μ A , μ H , r NA , r NH , r GH for dictating the actual nitrate and glucose feeds to the bioreactor.

The actual bolus nitrate demand F N and the glucose demand F G were thus determined by using values obtained from this fitting objective function.

The data that support the findings of this study are available from the corresponding author upon reasonable request. Rosenberg, J. A green light for engineered algae: redirecting metabolism to fuel a biotechnology revolution.

Article CAS PubMed Google Scholar. Shene, C. Metabolic modelling and simulation of the light and dark metabolism of Chlamydomonas reinhardtii. Plant J. Kato, Y. et al. Biofuels 12 , 39 Article PubMed PubMed Central Google Scholar. Cheirsilp, B. Enhanced growth and lipid production of microalgae under mixotrophic culture condition: effect of light intensity, glucose concentration and fed-batch cultivation.

Zheng, Y. High-density fed-batch culture of a thermotolerant microalga Chlorella sorokiniana for biofuel production.

Energy , — Article CAS Google Scholar. Shi, X. High-yield production of lutein by the green microalga Chlorella protothecoides in heterotrophic fed-batch culture.

Bordbar, A. Constraint-based models predict metabolic and associated cellular functions. Chang, R. Article Google Scholar. Zuñiga, C. Genome-scale metabolic model for the green alga Chlorella vulgaris utex accurately predicts phenotypes under autotrophic, heterotrophic, and mixotrophic growth conditions.

Plant Physiol. Loira, N. Reconstruction of the microalga Nannochloropsis salina genome-scale metabolic model with applications to lipid production. BMC Syst. Chang, L. Nonlinear model predictive control of fed-batch fermentations using dynamic flux balance models.

Process Control 42 , — Jabarivelisdeh, B. Model predictive control of a fed-batch bioreactor based on dynamic metabolic-genetic network models. IFAC-PapersOnLine 51 , 34—37 Juneja, A. Model predictive control coupled with economic and environmental constraints for optimum algal production. Ogbonna, J.

Zuniga, C. Predicting dynamic metabolic demands in the photosynthetic eukaryote Chlorella vulgaris. Li, C. Utilizing genome-scale models to optimize nutrient supply for sustained algal growth and lipid productivity.

NPJ Syst. Tebbani, S. Nonlinear predictive control for maximization of CO2 bio-fixation by microalgae in a photobioreactor. Bioprocess Biosyst. Hu, D. The design and optimization for light-algae bioreactor controller based on Artificial Neural Network-Model Predictive Control.

Acta Astronaut 63 , — White, R. Long-term cultivation of algae in open-raceway ponds: lessons from the field. Gu, C. Current status and applications of genome-scale metabolic models. Genome Biol.

Using genome-scale models to predict biological capabilities. Cell , — Colarusso, A. Computational modeling of metabolism in microbial communities on a genome-scale. Han, F. Enhancement of microalgal biomass and lipid productivities by a model of photoautotrophic culture with heterotrophic cells as seed.

Xiong, W. Double CO 2 fixation in photosynthesis—fermentation model enhances algal lipid synthesis for biodiesel production. Xiao, Y. Photosynthetic accumulation of lutein in Auxenochlorella protothecoides after heterotrophic growth. Drugs 16 , Vidotti, A. Analysis of autotrophic, mixotrophic and heterotrophic phenotypes in the microalgae Chlorella vulgaris using time-resolved proteomics and transcriptomics approaches.

Algal Res. Park, J. The contribution ratio of autotrophic and heterotrophic metabolism during a mixotrophic culture of Chlorella sorokiniana.

Public Health 18 , Huesemann, M. A validated model to predict microalgae growth in outdoor pond cultures subjected to fluctuating light intensities and water temperatures. Mears, L. A review of control strategies for manipulating the feed rate in fed-batch fermentation processes.

Sommeregger, W. Quality by control: Towards model predictive control of mammalian cell culture bioprocesses. Chen, G. Viable cell density on-line auto-control in perfusion cell culture aided by in-situ Raman spectroscopy. Lee, H. In situ bioprocess monitoring of Escherichia coli bioreactions using Raman spectroscopy.

Xu, J. Application of metabolic controls for the maximization of lipid production in semicontinuous fermentation. Natl Acad. USA , E—E Article CAS PubMed PubMed Central Google Scholar. Dong, T. Direct quantification of fatty acids in wet microalgal and yeast biomass via a rapid in situ fatty acid methyl ester derivatization approach.

Yuan, J. Carotenoid composition in the green microalga Chlorococcum. Food Chem. Chen, Y. An unconventional uptake rate objective function approach enhances applicability of genome-scale models for mammalian cells. Download references. This work was supported by the U.

National Science Foundation EFRI program Grant number: and CBET program Grant number: and the Department of Energy Grant number: DE-SC Department of Chemical and Biomolecular Engineering, Johns Hopkins University, North Charles Street, Baltimore, MD, , USA.

Department of Pediatrics, University of California, San Diego, Gilman Drive, La Jolla, CA, , USA. Department of Biology, San Diego State University, San Diego, USA. Department of Bioengineering, University of California, San Diego, Gilman Drive, La Jolla, CA, , USA.

Center for Microbiome Innovation, University of California, San Diego, Gilman Drive, La Jolla, CA, , USA. You can also search for this author in PubMed Google Scholar.

contributed to conception and design of the experiment. conducted the experiments. analyzed the data. drafted the paper. All authors read and approved the paper. Correspondence to Michael J. Open Access This article is licensed under a Creative Commons Attribution 4. Reprints and permissions.

Li, CT. Optimization of nutrient utilization efficiency and productivity for algal cultures under light and dark cycles using genome-scale model process control. npj Syst Biol Appl 9 , 7 Download citation.

Received : 27 September Accepted : 08 November Published : 15 March Anyone you share the following link with will be able to read this content:. Sorry, a shareable link is not currently available for this article. Provided by the Springer Nature SharedIt content-sharing initiative.

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

Skip to main content Thank you for visiting nature. nature npj systems biology and applications articles article. Download PDF. Subjects Computer modelling Metabolic engineering Plant sciences.

Abstract Algal cultivations are strongly influenced by light and dark cycles. Introduction Microalgae represent promising microorganisms for transforming renewable resources and inorganic carbon sources into biomass, biofuel precursors, and high-value products 1.

Results and discussion Advantages of using genome-scale model predictions on C. Full size image. a PID controller. b GMPC controller. Conclusions Fed-batch cultivation and PID controllers have been widely used in bioprocess development. Methods Algal strain and cultivation conditions Green microalgae C.

Summary of equations for 2L bioreactor cultures Open-loop system Initial biomass levels x 0 , glucose levels G 0 and nitrate levels N 0 were measured as described above and used as inputs into the open-loop system. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request.

Code availability The codes that support the findings of this study are provided in supplementary information. References Rosenberg, J.

Article CAS PubMed Google Scholar Shene, C. Article CAS PubMed Google Scholar Kato, Y. Article PubMed PubMed Central Google Scholar Cheirsilp, B. Article CAS PubMed Google Scholar Zheng, Y. Article CAS Google Scholar Shi, X. Article CAS PubMed Google Scholar Bordbar, A.

These results confirm our conclusion, based on growth measurements in Fig. A Induction dynamics in response to a single pulse of lactose lasting 10, 20, 30, 45, or 60 minutes.

In each case, the permease density continues to increase and levels start to decay approximately 40 minutes after lactose is removed from the environment. Measurements from a long experiment in a periodic environment inset are superimposed onto a single period.

The LacY-Venus reporter delay lasts C Induction by either 60 minutes of 0. We term this behavior response memory : the ability of a regulatory network to continue to respond after the stimulus has been removed.

Hysteresis and expression delays are to be expected in multi-level gene regulatory circuits, and in the particular case of lac regulation these delays can involve the kinetics of mRNA degradation [29] , repressor re-binding [30] , [31] , catabolite repression mediated by cAMP [32] , and dynamics of allolactose, the intracellular inducer of the lac operon [33].

We therefore characterized the relative contributions of these effects to the observed response memory. First, the ability to detect in vivo changes in lac protein levels is set by the maturation time of LacY-Venus both folding and chromophore formation, measured to last less than 7 minutes in vivo [34] , which introduces a delay between observed and actual protein levels.

To accurately measure the delay associated with the LacY-Venus protein maturation, we analyzed the LacY-Venus fluorescence levels when glucose and lactose environments alternate with an environmental duration of 90 minutes and measured the phase difference between the environment and the LacY-Venus levels.

The average delay measured under these experimental conditions is The observed peak at 40 minutes in Fig. In contrast to induction using lactose, which requires LacZ activity to produce the inducer allolactose, IPTG induces the lac operon directly.

Furthermore, this experiment shows that the stringent response caused by carbon starvation does not significantly affect the induction dynamics. We next tested whether residual intracellular inducer could account for the observed response memory, by using 2-nitrophenyl β -D-fucopyranoside ONPF , an anti-inducer molecule that competitively binds LacI, excludes IPTG, and increases LacI's affinity for its operator site.

However, they exhibited significantly different response memory profiles when the inducer was removed at minutes: lac expression in the presence of ONPF started to decrease 20 minutes after IPTG removal, compared to the 40 minutes measured in the absence of ONPF.

The remaining 6 minutes of response memory, not accounted for by reporter delay, can be explained by the measured time for LacI to fully rebind lac operator sites in the presence of ONPF min, [30] , [31] as well as the lifetime of lac mRNA min, [29].

These in vivo measurements support our predictions above based on the growth rate dynamics. First, we found that response memory enables cells to continue responding to lactose through the glucose exposures.

Second, we showed in Fig. We note that because our experiment is not designed for single-molecule sensitivity, we cannot measure the initiation events themselves. However, we clearly see that cells cross our detection threshold at approximately the same time when induced with IPTG in glucose i.

without any carbon stress or with lactose under carbon stress. Third, we measured the post-initiation rate constant for lac protein production to be. This implies that post-initiation the time to increase lac induction levels to would be approximately minutes, which is consistent with our prediction that barrier B can be crossed in as little as 3 minutes.

While the major determinants of the lag phase were found to be the initial induction steps and the recovery from stringent response, the potential fitness gains that cells might reap from response memory remained unclear. To better quantify the fitness advantage of response memory in the lac operon, we adapted the established metabolic model described in [35] to fluctuating environments see Materials and Methods.

We focused exclusively on the observed memory effects and their impact on metabolic activity, and did not model the stringent response since it did not significantly affect the induction dynamics Fig.

The model explicitly accounts for intracellular concentrations of lactose, allolactose, lac operon mRNA, and lac proteins, and captures several features we observed in experiments. For example, in response to a single pulse of extracellular inducer, protein levels can continue to increase after the stimulus is removed, causing an overshoot, if sufficient mRNA and intracellular inducer levels are maintained Fig.

Likewise, the model exhibits phenotypic memory consistent with our observations. A Schematic of the gene regulation model, including extracellular inducer , mRNA , and protein. A few representative examples of how lac levels evolve under fluctuating conditions are shown in the insets.

C A difference in lac expression levels is observed for models with solid line and without dashed line response memory. The model that includes response memory correctly predicts the experimentally measured IPTG induction dynamics cyan line.

Response memory leads to increased intracellular LacZ levels and higher catabolic activity. Since response memory can be explained by the LacI-mediated repression kinetics Fig. We used the model to test this conclusion by artificially reducing the allolactose degradation rate to zero during glucose environments.

We obtained similar results across a range of slower but non-zero degradation rates. We show in Fig. The predicted dynamics closely follow the measured lac levels obtained from minutes IPTG induction cyan line.

The highlighted regions in Fig. The modeling results support a picture in which response memory provides a large adaptive advantage when external fluctuations occur faster than the cell division time, while phenotypic memory is beneficial for slower fluctuations, spanning several generations.

We have presented two distinct memory mechanisms in the lac operon of E. coli , phenotypic and response memory, each of which is beneficial over different timescales. Phenotypic memory allows cells to maintain an adapted state for multiple generations after a specific carbon source is removed from the environment.

Since phenotypic memory operates through the transmission of stable cytoplasmic proteins, it may be employed as a general strategy in other organisms to transmit metabolic information between generations, as observed e.

in the yeast galactose system [9] , [36]. More generally, the intrinsic mechanism behind phenotypic memory being passive — based on intracellular proteins whose lifetime is longer than a typical generation — similar memory effects are expected to be present for other fluctuations and in other organisms.

Adaptation mechanisms that rely on the expression of long-lived permease molecules — e. We used fast fluctuating environments to dissect the determinants of lag phases following a transition from glucose to lactose.

Our results suggest a simple biological model of the lag phase in which lac protein activity and the stringent response are mutually inhibitory processes: Lac protein activity in lactose has an inhibitory effect on the stringent response due to glucose production and amino acid synthesis, while the stringent response initially inhibits lac protein production through its global inhibitory effects on translation.

To see this, we consider two examples. First, we compare for uninduced and pre-induced cells Fig. In the pre-induced case, after the first lactose exposure cells rapidly recover full growth in glucose, whereas if no lac proteins are initially available, cells experience a slow recovery in glucose.

The stringent response due to the lactose exposure is therefore much less severe when a small amount of LacZ induced is available to hydrolyze lactose and initiate positive autoregulation. Second, we note that for Fig. However, for Fig. We conclude that protein production during the stringent response is too slow to allow cells to cross the threshold during the short lactose exposures for.

In particular, if lactose encounters unexpectedly cease, this cost will no longer be temporary, but sustained by the population indefinitely. Cells employing response memory avoid such long-term cost by transiently expressing the required genes for a short amount of time following an initial exposure to the stimulus, with a maximal metabolic cost that is limited by the duration of this transient expression.

Should environmental fluctuations cease, cells will suffer only a small, short-term fitness cost. On the other hand, should fast fluctuations persist, as we have shown the cells reap a significant fitness benefit.

In particular, we showed that cells reach higher induction levels more rapidly by maintaining their response profile following the removal of an external inducer. Memory in different genetic network architectures could affect not only the cost of gene expression, but also the evolution of gene expression levels.

The timescale over which phenotypic memory persists is determined to a large extent by the gene's expression level provided the protein is sufficiently stable. Expression levels may be evolutionarily tuned not only to support growth in a single environment, but also to provide cells' progeny with memory of past environments.

The interplay of memory and metabolic constraints could thus dramatically change the nature of evolutionary trajectories and optima. We expect theoretical analyses may be fruitfully applied to explore these possibilities.

The power of the memory mechanisms we have described lies in their universality. Protein lifetimes and regulatory networks can be tuned in simple ways to give rise to physiological memory under rapidly changing conditions. Microorganisms have to handle both internal and external sources of noise, and while many genetic networks have evolved to exploit stochastic fluctuations of intracellular molecular components to regulate key cellular processes [41] , we have shown that molecular rates of signal transduction reactions can be modulated to optimize response profiles for growth in fluctuating environments.

Together, phenotypic and response memory allow bacteria to adapt to a wide range of fluctuation timescales in sophisticated, history-dependent ways.

These memory mechanisms constitute general strategies that bacteria can employ to adapt to diverse environmental fluctuations — including nutrients, antibiotics, and other physiological stresses.

The microfluidic device used in this study was made using standard soft lithography and microfabrication techniques and consists of growth chambers and a main flow channel patterned from two SU-8 layers 1.

The devices were fabricated by making polydimethylsiloxane PDMS replicates of the SU-8 master. The PDMS devices were peeled from the silicon master and 1. When transitioning between two media, both valves were closed for 15 seconds before the new one was opened to let the pressure equilibrate inside the device and to avoid backflow problems.

A T-junction upstream of the growth chambers ensured that transitions between the different media occurred very rapidly. By flowing a fluorescent dye inside the device, the transition between each type of media was measured to occur in less than milliseconds Supp.

Cells inside the growth chambers push their immediate neighbors toward the main flow channel as they increase in size, and the lateral speed at which the cells move is proportional to the population's mean elongation rate.

An optical flow algorithm implemented using openCV [42] was used to measure the displacement between successive frames. This displacement was used to find the average cell speed over the region between 10 and 15 away from the closed end of the growth chamber.

The lateral speed reports on the cumulative growth rate of cells in the first 10 microns of the growth chamber providing a measure of the relative growth rate of the population.

Error bars on relative growth plots report the standard error of the mean as averaged over the 5 chambers present in a single field of view. These error bars measure intrinsic cell-to-cell variability in growth, due to stochasticity in cell division rates, elongation rates, and gene expression processes.

The lac induction dynamics of a population subjected to sudden environmental changes are modeled as described in [35] , with an additional equation to account for mRNA transcription.

The model assumes that LacY protein levels are proportional to LacZ levels. Unless otherwise noted in Table 1 , refer to [35] for a complete rationalization behind each parameter's value. The set of equations are given by 1 2 3 4 where , , , and are the intracellular concentrations of lactose, allolactose, mRNA and LacZ proteins, respectively parameters are specified in Table 1.

The expression for the lactose hydrolysis rate is given by 5 where , , and are obtained by solving Eqns.

To qualitatively compare behaviors with and without response memory, we artificially reduced the rate of allolactose turnover in glucose environments taking to attain response memory in our simple model. Similar results were obtained by reducing instead the mRNA degradation rate in the transition from lactose to glucose.

Full methods as well as further details of microfluidic fabrication, strain description, image acquisition and analysis, and any associated references are provided in Text S1.

Timescale of the environmental change inside a chemoflux. The transition are accurately described by exponential functions red lines, seconds and seconds. Growth rate measurement. Since each cell division event yields two cells at age zero, the fraction of cells at age 0 is twice the population's growth rate.

Duration of the lag phase. B - F The duration of the lag and recovery phases is computed from a linear regression of the lateral cell speed and the results are presented in Fig.

S1 for quantification. A video of cell growth in the chemoflux growth chambers during a glucose-to-lactose transition. We thank Calin Guet, Yuichi Wakamoto, Michael Rust, David Gresham, Joao Xavier, Matthew Eames, and Wei-Hsiang Lin for comments on the manuscript.

We also thank Jose Vilar, Gene Huber, and Bruce Levin for discussions. We especially thank Tobias Bergmiller and Calin Guet for the over-expression plasmids, Wei-Hsiang Lin for his assistance in carrying out the constitutive Lac expression experiments, and Michael Rust for providing a microscopy setup for additional experiments.

We thank the Wakamoto and Xie labs for providing E. coli strains. Conceived and designed the experiments: GL EK.

Performed the experiments: GL. Analyzed the data: GL. Interpreted the data: GL EK. Wrote the paper: GL EK. Article Authors Metrics Comments Media Coverage Reader Comments Figures. Correction 16 Oct The PLOS Genetics Staff Correction: Memory and Fitness Optimization of Bacteria under Fluctuating Environments.

Abstract Bacteria prudently regulate their metabolic phenotypes by sensing the availability of specific nutrients, expressing the required genes for their metabolism, and repressing them after specific metabolites are depleted.

Author Summary Bacterial adaptation to new environments typically involves reorganization of gene expression that temporarily decreases growth rates. Introduction Escherichia coli cells grown in the presence of both glucose and lactose first consume glucose, which is more easily metabolized, before expressing the genes necessary for lactose catabolism [1] — [3].

Download: PPT. Figure 1. Chemoflux device for growth rate measurement in changing environments. Results Phenotypic memory in response to sudden environmental changes The growth rate of E. Figure 3. Molecular components of phenotypic memory in the lac operon.

Figure 4. Lag phase and recovery in rapidly fluctuating environments. Response memory and dynamics of lac protein expression To reproduce the conditions of Fig. Figure 5. In vivo measurement of LacY expression in fluctuating environments.

Modeling lac operon dynamics with memory in a fluctuating environment While the major determinants of the lag phase were found to be the initial induction steps and the recovery from stringent response, the potential fitness gains that cells might reap from response memory remained unclear.

Figure 6. Mathematical modeling quantifies fitness advantage of memory. Discussion We have presented two distinct memory mechanisms in the lac operon of E.

Materials and Methods Device description and fabrication The microfluidic device used in this study was made using standard soft lithography and microfabrication techniques and consists of growth chambers and a main flow channel patterned from two SU-8 layers 1. Growth rate measurements Cells inside the growth chambers push their immediate neighbors toward the main flow channel as they increase in size, and the lateral speed at which the cells move is proportional to the population's mean elongation rate.

Mathematical model of lactose metabolism The lac induction dynamics of a population subjected to sudden environmental changes are modeled as described in [35] , with an additional equation to account for mRNA transcription.

Supporting Information. Figure S1. s PDF. Figure S2. Figure S3. Text S1. Supplementary methods. Video S1. s MOV. Video S2. Acknowledgments We thank Calin Guet, Yuichi Wakamoto, Michael Rust, David Gresham, Joao Xavier, Matthew Eames, and Wei-Hsiang Lin for comments on the manuscript.

Author Contributions Conceived and designed the experiments: GL EK. References 1. Monod J Recherches sur la croissance des cultures bactériennes. Monod J The growth of bacterial cultures. Annu Rev Microbiol 3: — View Article Google Scholar 3.

Dekel E, Alon U Optimality and evolutionary tuning of the expression level of a protein. Nature — View Article Google Scholar 4. Magasanik B Catabolite repression. Cold Spring Harb Sym — View Article Google Scholar 5.

Gorke B, Stulke J Carbon catabolite repression in bacteria: many ways to make the most out of nutrients. Nat Rev Microbiol 6: — View Article Google Scholar 6.

New A, Cerulus B, Govers S, Perez-Samper G, Zhu B, et al. PLoS Biol e View Article Google Scholar 7. Scott M, Gunderson CW, Mateescu EM, Zhang Z, Hwa T Interdependence of cell growth and gene expression: Origins and consequences.

Science — View Article Google Scholar 8. Deris JB, Kim M, Zhang Z, Okano H, Hermsen R, et al. Science View Article Google Scholar 9. Razinkov IA, Baumgartner BL, Bennett MR, Tsimring LS, Hasty J Measuring competitive fitness in dynamic environments.

J Phys Chem B —