The engineering of many-component, synthetic biological systems is being made easier by the development of collections of reusable, standard biological parts [1
]. Standardization of components has been instrumental in managing complexity in other engineering fields by helping engineers to reliably design and deploy systems comprised of combinations of parts [7
]. However, it is an open question whether the overwhelming complexity of living systems will prevent biological engineers from fully achieving similar capabilities (below). To help answer this question, a Registry of Standard Biological Parts started at MIT now maintains and distributes thousands of BioBrick standard biological parts [8
]. BioBrick parts provide the first popular example of standard biological parts. However, BioBrick parts are currently only standardized in terms of how individual parts are assembled into multi-component systems (that is, "physical composition") [1
The utility of so-called standard biological parts would increase if the behavior of parts, both in isolation and in combination, were more predictable (that is, "functional composition")[9
]. Prediction of behavior, in turn, depends on the initial designs and refinement of the parts themselves, the characterization of part functions, and the representation of part functions via abstract models (for related examples see [10
]). Today, most BioBrick parts are directly derived from natural DNA sequences with only slight modifications to support at least one physical assembly standard, and many parts remain to be characterized. For example, fewer than 50 out of over 500 transcriptional promoters now available via the Registry have been characterized. Making matters worse, for the 50 characterized promoters, the methods of characterization are disparate and the resulting data incomparable. Shared and standardized approaches are needed in order to begin to address the challenge of characterizing promoters (and other types of standard biological parts) across a distributed community of biological engineers.
Making reliable and comparable in vivomeasurements of biological parts has proven challenging. For example, five different efforts to measure the abundances of proteins in the yeast pheromone mating response system, one of the best characterized eukaryotic signalling systems, produced reports for the numbers per cell (abundances) of key system proteins that vary over a factor of ~12 [Thomson TM, Benjamin KR, Bush A, Love T, Pincus D, Resnekov O, Yu R, Gordon A, Colman-Lerner A, Endy D, Brent R: Scaffold number in yeast signaling system sets tradeoff between system output and dynamic range. Molecular Systems Biology. Unpublished
]. Such examples suggest that measurement of the state or activity of biological systems, whether natural or engineered, may be unlike past engineering experiences, in that the minor differences in experimental conditions (relative to what can be readily controlled for, below) may cause large changes in the properties being measured. Even if conditions could be controlled for, it has proven challenging for researchers to develop and adopt standard approaches for characterizing biological parts. For example, an analysis of 80 published papers in which researchers used beta-galactosidase (β
-gal) activity as a measure of gene expression found that at least six different protocols were used to measure enzyme activity [13
]. In addition, nearly all activities were reported in "Miller units" even though in several cases there were differences in the substrates used to quantify enzymatic activity (CPRG or ONPG), the experimental conditions (pH and temperature for the assay), and even the absolute units of the Miller unit (nmol/min or μ
]. Differences in conditions such as using either CPRG or ONPG as a substrate for enzymatic assays lead to incompatible results [15
], and thus Miller units should generally not be considered comparable unless they have been calibrated against a common reference standard [13
The challenge of making reliable in vivomeasurements of biological parts is further compounded by the need to measure many part properties indirectly via biological "measurement instruments" such as reporter proteins whose production can also be sensitive to experimental conditions. For example, β
-gal activity can be used as an indirect measure of the behavior of a promoter, but the translation and activity of the β
-gal protein is itself sensitive to experimental conditions such as temperature or choice of media. Since both the measurement instrument (β
-gal) and the property being measured (promoter activity) are sensitive to measurement conditions (perhaps in differing ways) correcting for errors in measured promoter activity due to changes in conditions is more difficult. In theory such challenges could be addressed by strict adherence to standard measurement conditions. However, the adoption of standard measurement conditions in biological engineering is prevented by both practical constraints (as noted above) and also engineering constraints, such as culture or performance requirements that are specific to a particular biotechnology application. The overall situation is summed up nicely via the following quote: "There is no such thing as a standard (biological) component, because even a standard component works differently depending on the environment" [16
Although the characterization of standard biological parts is challenging, lessons from the measurement of other types of physical objects are worth considering. For example, one approach to controlling for variation in the measured property of an object in response to changing experimental conditions is to collect data from which to develop a model that describes the relationship between the measured property and experimental conditions. As a specific example, models based on empirically determined coefficients of thermal expansion for common building materials (for example, Oak = 54E-6/K at 20°C; Stainless Steel = 17.3E-6/K at 20°C) are now sufficient to enable the reliable construction of structures across a range of environments [17
]. However, given the complexity of living matter, the relationships between the measured properties of biological parts and experimental conditions may be difficult to determine (at first). Thus, a second lesson worth considering is the measurement of relative (or ratio) properties rather than absolute characteristics. A relative measure is the ratio of the measurement of some aspect of the object being characterized in comparison to a standard reference object that is measured under the same conditions. For example, early methods for the diagnosis of osteoporosis made use of a measure of spinal cord deformity that was based on the ratio of various length measurements of vertebra within an individual patient [18
]. Doctors, by using a relative measurement for length, could account for variation in vertebra sizes between individuals of different body types or heights. As a second example, microarray experiments are frequently performed by co-hybridizing probes synthesized from both a reference and experimental RNA sample that have been labelled with different colors [19
]; gene expression levels are then reported as the ratio of the experimental and reference intensities on each array spot. Thus, measurements made in relation to defined reference standards may provide an important first approach in characterizing the in vivoactivity of biological parts and, over time, could enable the collection of empirical data sufficient to support the development of models that describe the effect of varying conditions on part properties.
Here, we characterized the in vivoactivity of BioBrick promoters in order to evaluate if measuring relative activities might provide a useful initial framework for measuring the activity of standard biological parts across varying conditions. We chose to characterize promoters as a first example since they are ubiquitous in engineered biological systems, relatively well-understood, practically useful to biological engineers, and poorly characterized in the existing BioBrick collection [20
]. We developed a system that allows indirect measurement of the activity of promoters via observation of the synthesis rate of Green Fluorescent Protein (GFP) encoded by mRNA transcribed from each promoter. Our system requires the use of a quantitative model that allows promoter activity to be estimated from observed rates of GFP synthesis (below). Using this approach we demonstrate that normalizing the apparent absolute activity of a promoter to a defined reference standard promoter can help account for variation in conditions that would otherwise lead to significant differences in reported measurements.