Like electronics, Synthetic Biology deals with circuits. Thus, over the last twenty years, efforts have been made to adapt electrical engineering concepts and methods to the modular design and modeling of circuits made of DNA. In his famous 2005 paper (Endy, 2005), Drew Endy depicted a possible correspondence between the electrical current—easily measurable with an ammeter—and biological currents that could be responsible for the working of synthetic gene circuits. Following the dogma of molecular biology, RNA polymerases and ribosomes, which lead the synthesis of mRNA and proteins, respectively, appeared to be the biological counterparts of the electrons. Moreover, basic circuit components (in eukaryotic cells) were identified with promoters, coding regions (CDSs), and terminators, i.e., DNA pieces with a well-defined function either in transcription or translation. As the electric current permits to wire together resistors, batteries, solenoids and all the other basic electric components, the fluxes of RNA polymerase (PoPS: polymerases per second) and ribosomes (RiPS: ribosomes per second) should be the biological currents (a shared input/output) that permit to assemble together, first, biological parts into transcription units (TUs) and then TUs into circuits. Hence, biological fluxes permitted to define concepts such as part composability and abstraction hierarchy in Synthetic Biology.

Computational biologists liked these innovatory ideas because they indicated how to deal with biological circuits in a systematic way, whereas wet-lab biologists were skeptical since they were well-aware of the fact that a disordered motion like that of RNA polymerases and ribosomes could not be measured with any instruments. Therefore, PoPS/RiPS-based modeling was never completely accepted by the Synthetic Biology community and, in the end, it showed—in our opinion—the limits of using electronics concepts to study biological systems.

We were among the first to develop a piece of software (Marchisio and Stelling, 2008)—later named “Parts and Pools” (Marchisio, 2014b)—based on an extension of the ideas of Drew Endy. We pointed out that PoPS and RiPS allowed, indeed, parts’ composability but were not enough to describe the wiring among transcription units, which is what really makes a circuit work. Therefore, we had to introduce other signal carriers (transcription factors, small RNAs, and chemicals) and their corresponding fluxes to mediate the interaction among TUs. Our piece of software was an add-on of ProMoT (process modeling tool) (Mirschel et al., 2009), a program for the visual, modular design of complex systems. ProMoT demands to program in MDL (model definition language), a Lisp-based language that permits to define modules that communicate via fluxes (of molecules, in biological systems) calculated and exchanged via the so-called module terminals. ProMoT provides a Graphical User Interface as well where modules are displayed in a drag-and-drop way and terminals are wired together.

Our ProMoT add-on is a collection of Perl and Python scripts each generating one or more MDL modules. The latest version (Marchisio et al., 2013) was successfully applied to the design and modeling of eukaryotic circuits. The usage of fluxes made it straightforward to build models based on mass-action kinetics. Moreover, eukaryotic promoters and coding regions can be characterized by a high number of species and reactions that are calculated by means of BioNeTGen (Blinov et al., 2004). The CDS script generates not only the MDL file for the CDS part, but also those corresponding to the pools of its mRNA and protein. Nucleus and cytoplasm are designed separately (due to their complexity) and then connected by running another script. Finally, a circuit is encoded by a collection of MDL files associated with parts and pools plus a main one that contains the connections among all the modules. ProMoT is able to export the circuit into the SBML2 format that permits circuit analysis and simulations—with, for instance, COPASI (Hoops et al., 2006) as we generally did—but loses the circuit modularity.

SBML Level 3 (Keating et al., 2020) with the hierarchical model composition package (comp) (Smith et al., 2015) allows for a direct modular modeling of genetic circuit without the need for the help of another language, such as MDL, nor the usage of fluxes of molecules to establish the wiring among the modules present in the circuit. Furthermore, SBML Level 3-comp (concisely SBML3-comp) files can be easily generated by means of the libSBML (Bornstein et al., 2008).

In this paper we describe the wet-lab implementation, in the yeast Saccharomyces cerevisiae, of two NOT gates based on protein degradation—a design strategy advanced 15 years ago (Grilly et al., 2007) but never extensively exploited nor optimized—and then we show how they can be modeled, in a modular way, by means of SBML3-comp. We will not explain all the libSBML commands required to generate the full circuit model (for this we refer the reader to our book chapter (Marchisio, 2021) and the files available at GitHub—see Data Availability below) but we will focus on the way to establish connections among circuit modules, which no longer demands to define fluxes of molecules that cannot be measured in the lab. SBML3-comp is already supported by several computational tools such as COPASI, iBioSim (Watanabe et al., 2019), Virtual Part Repository 2 (Misirli et al., 2021), Tellerium (Choi et al., 2018), and BMSS (Yeoh et al., 2019). For this reason, we think it will become the standard language to model genetic circuit in the very next future.

In E. coli, the ClpX and ClpP proteins work in tandem (ClpXP) in order to carry out protein degradation. Upon recognition of a degron sequence, ClpX proceeds to unfold and translocate a protein into the ClpP proteolytic compartment with the release of small peptides (Baker and Sauer, 2012). This bacterial system was first shown to work efficiently in S. cerevisiae by Grilly and co-authors (Grilly et al., 2007)—and later re-proposed by Macia et al. (Macia et al., 2016)—who constructed a four-gene circuit where the yeast enhanced green fluorescent protein—yEGFP (Sheff and Thorn, 2004)—was fused to the ssrA degradation tag and the expression of both ClpX and ClpP was controlled by the LacI/IPTG system. Moreover, yEGFP was produced only in the presence of galactose in the cell culture (see Supplementary Figure S1). Taking a cue from this work, we made use of the ClpXP system to construct two different NOT gates, one responding to galactose, the other to beta-estradiol.

SBML3-comp represents a huge improvement of SBML2 since it permits to construct models of biological systems and, therefore, of synthetic gene circuits, in a modular way without being assisted by other programs. libSBML permits to divide the model of a circuit in different scripts, each generating one or more modules. We think that the best way is to treat transcription units, rather than single DNA parts, as basic modules. A TU can be encoded in the same file together with its corresponding pool of mRNA and, if required, that of the protein it encodes for. The direct use of SBML3 allows also to choose easily the best kinetics for every reaction, without the constraint of using the same in the whole circuit.

In this work we have shown how Boolean gates characterized by high ON/OFF ratio can be built, in S. cerevisiae, by using an almost forgotten, though powerful, yeast-orthogonal system, i.e., the bacterial ClpXP dimer to induce protein degradation. Our Boolean gates served as a case study to explain how to use SBML3-comp to model genetic circuits in a new modular way. Importantly, it was apparent how Python scripts, which generate the SBML3-comp files, can be recycled for the same TUs inside different circuits or just slightly modified to construct models for similar TUs.

We think that the joint use of the System Biology Open Language—SBOL (Galdzicki et al., 2014) and SBML3-comp seems to be the best way to combine, in the future, a detailed description of the actual DNA circuit components (sequences) with modular modeling (Watanabe et al., 2019; Misirli et al., 2021).