Living cells maintain a steady state of biochemical reaction rates by exchanging energy and matter with the environment. These exchanges usually do not occur in in vitro systems, which consequently go to chemical equilibrium. This in turn has severely constrained the complexity of biological networks that can be implemented in vitro. We developed nanoliter-scale microfluidic reactors that exchange reagents at dilution rates matching those of dividing bacteria. In these reactors we achieved transcription and translation at steady state for 30 h and implemented diverse regulatory mechanisms on the transcriptional, translational, and posttranslational levels, including RNA polymerases, transcriptional repression, translational activation, and proteolysis. We constructed and implemented an in vitro genetic oscillator and mapped its phase diagram showing that steady-state conditions were necessary to produce oscillations. This reactor-based approach will allow testing of whether fundamental limits exist to in vitro network complexity.synthetic biology | cell-free protein synthesis | computational biology | minimal artificial cell I nstead of complex and ill-characterized cellular hosts, in vitro systems have recently become popular alternatives for implementing synthetic networks. In vitro systems can be completely defined, easily manipulated, interrogated, and have been used to study a number of biological phenomena. For example, periodic temporal patterns were observed in systems based on nucleic acid synthesis and degradation (1, 2), and ordered spatial patterns were created from purified cell division regulators (3). In vitro transcription and translation (ITT)-based systems should, in principle, be able to use all regulatory functionalities found in living cells. Reconstituted, defined ITT systems like the PURE mix (4), are particularly appealing for bottom-up synthetic biology. A number of recent examples show that various genetic (5-10) and metabolic (11) networks can be implemented in ITT systems. Genetic network complexity has, however, been limited to genetic cascades, where the output of one stage acts on the next stage, whereas examples of positive and negative feedback have been basic (8,9,12). The main limitation to network complexity in vitro derives from its batch reaction format. In batch, synthesis rates decrease as precursors are consumed, enzymatic activities degrade, and reaction products accumulate. This rapid approach to chemical equilibrium severely limits network size. In addition, negative feedback is particularly difficult to implement, because regulators from earlier stages are not removed. The implementation of active degradation mechanisms for RNA and proteins (13) could solve the problem of product removal, and synthesis times can be increased by using reactors that allow an exchange of small molecules between the ITT mix and a feeding solution. Large-volume continuous flow and exchange systems were developed to increase the amounts of protein produced by ITT systems and are based on di...