The spatiotemporal control of gene expression is orchestrated in part by distally located DNA
sequences known as enhancers. The first viral and cellular enhancers were identified by cloning
fragments of DNA into a plasmid with a reporter gene and promoter 1–4 . Enhancement of
transcription in such a reporter assay is a widely used approach for evaluating whether a putative
regulatory element is a bona fide enhancer. However, conventional, one-at-a-time reporter assays
are insufficiently scalable to test the >1 million putative enhancers in the human genome 5–8 .
Massively parallel reporter assays (MPRAs) modify the in vitro reporter assays described above
to facilitate simultaneous testing of thousands of putative regulatory elements or variants
thereof 9–11 in a single experiment. Instead of relying on measurement of a conventional reporter,
MPRAs characterize each element in a multiplex fashion, through sequence-based quantification
of barcodes incorporated into the RNA, each associated with a different element 9–15 . MPRAs (a
term we use broadly to encompass related methods including STARR-seq 16 and lentiMPRAs 17 )
have facilitated the scalable study of putative regulatory elements for goals ranging from
functional annotation 16–18 to variant effect prediction 10–15,19 to evolutionary reconstruction 20,21 .
To date, several groups have implemented enhancer-focused MPRAs, but with diverse designs.
Some of the major differences include whether the enhancer is upstream 10,11 vs. within the 3′ UTR of the reporter gene 16 , and whether the construct remains episomal vs. integrated 17 .
Additionally, most MPRAs test sequences cloned in one of two possible orientations, effectively
assuming that enhancer activity is independent of orientation. Finally, while larger sheared
genomic DNA fragments 16,22 , PCR amplicons 12 or captured sequences 23,24 have been used in
MPRAs, most studies using MPRAs synthesize libraries of candidate enhancers on microarrays,
and are therefore limited to testing shorter sequences (typically less than 200 bp).
Unfortunately, we have, as a field to date, largely failed to evaluate how these design choices
impact or bias the results of MPRAs. First, although assays like STARR-seq wherein the
enhancer serves as the barcode are more straightforward to implement, our understanding of how
position (3′ of the promoter, rather than 5′ as in a more conventional reporter assay vector) or the
fact that the sequence is serving as the barcode, influences results, remains incomplete. Arnold et
al. notably benchmarked STARR-seq against 142 conventional luciferase assays (r = 0.83), but
STARR-seq has yet to be systematically compared to other MPRAs 16 . Second, although we
previously showed differences between episomal vs. integrated MPRAs 17 , it is not clear how
these differences rank relative to those resulting from other design choices. Third, although the
orientation-independence of enhancers has been evaluated in Drosophila 16,25,26 , to our knowledge
the robustness of this assumption has not previously been systematically tested in a mammalian
system. Finally, the typical choice to test <200 bp fragments, each corresponding to a putative
enhancer, is entirely based on technical limitations of massively parallel DNA synthesis, rather than on any principled understanding of the actual size of enhancers. The consequences of this
choice for the results obtained remain largely unquantified.
Particularly as efforts to validate the >1 million putative human enhancers 5–8 , as well as the
growing number of disease-associated noncoding variants, begin to scale, a clear-eyed
understanding of the biases and tradeoffs introduced by various MPRA experimental design
choices is needed. To this end, we performed a systematic comparison, testing the same 2,440
sequences for regulatory activity using nine different MPRA strategies, including conventional
episomal, STARR-seq, and lentiviral designs. Second, we tested the same sequences in both
orientations relative to the promoter. Finally, we further developed our multiplex pairwise
assembly protocol 27, and applied it to test short (192 bp), medium (354 bp), and long (678 bp)
versions of the same enhancers. Our results quantify the impact of MPRA experimental design
choices and also provide further insight into the nature of enhancers.