The Anderson Promoter Collection: A Verified Reference Table for BBa_J23100-Series Constitutive Promoters
11 min read · Updated July 14, 2026
This is a cross-checked reference table for the Anderson constitutive promoter collection: the 20 parts named BBa_J23100 through BBa_J23119 in the iGEM Registry of Standard Biological Parts, which function as a de facto standard "expression-strength ladder" for tuning constitutive transcription in E. coli. Every sequence below was pulled from the Registry's own catalog page and independently cross-checked against a second, independently maintained parts mirror; every relative-strength value is labeled with how much weight it can actually bear.
That second part matters more than it sounds. The sequences are simple 35-nucleotide strings and are easy to get right once you've looked at the source. The "relative strength" numbers that get copy-pasted next to them in papers, plasmid maps, and RBS calculators are a different kind of object entirely: a single historical fluorescence measurement, taken in one E. coli strain, in one medium, at one growth stage, in 2006. Treating that number like a physical constant is one of the more common quiet mistakes in promoter selection, and this guide is explicit about where the solid ground ends.
What the Anderson promoter collection actually is
The collection was recovered from a small combinatorial promoter library screened by J. Christopher ("Chris") Anderson and submitted to the iGEM Registry by the 2006 UC Berkeley iGEM team. It consists of 20 parts, BBa_J23100 through BBa_J23119, all built on the same 35 bp scaffold. BBa_J23119 is the literal consensus sequence of the family; the other 19 members are single- to multi-nucleotide variants of it, isolated as hits from the same screen rather than designed one at a time.
The Registry's own description frames the collection's purpose plainly: general protein expression in E. coli, and, in the Registry's own words, "likely" other prokaryotes as well. That said, the actual characterization data behind the strength numbers in this guide is E. coli-only, so treat cross-species use as a design hypothesis the Registry's description points toward, not something this specific dataset has tested. Because all 20 parts share the same backbone with only a handful of positions varying, they function less like 20 independent promoters and more like a titration series: pick a rung on the ladder that's roughly as strong as you need, then fine-tune with RBS strength or copy number if you need more precision than a promoter swap gives you.
Two restriction sites are deliberately built into the shared scaffold: an NheI site (GCTAGC) and an AvrII site (CCTAGG). The Registry describes these as making the promoters "a scaffold for further modification," and because the promoters are only 35 bp, the simplest way to obtain physical DNA is to order two complementary oligonucleotides and anneal them rather than PCR them out of anything.
Sequence architecture: what "consensus-derived" means here
The consensus sequence, BBa_J23119, is 5'-TTGACAGCTAGCTCAGTCCTAGGTATAATGCTAGC-3'. Breaking it down by position (1-based, 35 nt total):
Positions 1-6, TTGACA, match the canonical E. coli sigma70 -35 hexamer consensus. Positions 7-23 are a 17-nucleotide spacer, which is the standard optimal spacer length for sigma70 promoters. Positions 24-29, TATAAT, match the canonical sigma70 -10 hexamer consensus. Positions 30-35 are a trailing GCTAGC.
The two restriction sites mentioned above sit inside this structure rather than outside it: GCTAGC (NheI) appears twice, once at positions 7-12 (the start of the spacer) and once at positions 30-35 (the tail), and CCTAGG (AvrII) appears at positions 18-23, immediately upstream of the -10 hexamer.
This matters for reading the family correctly: every one of the other 19 members differs from this consensus only inside the -35 hexamer, the -10 hexamer, or (in two cases) a single spacer position, never in the flanking NheI/AvrII scaffold itself, and never by changing the spacer length. The library screen sampled sequence identity within the two recognition boxes; it did not sample spacer length or the restriction-site scaffold.
The verified reference table
Sequences below are given 5' to 3' and were pulled directly from the Registry's own consensus-diff table (parts.igem.org/Promoters/Catalog/Anderson), then verified against an independent parts mirror for every one of the 20 members, not just a sample. All 20 sequences are internally consistent with the verified consensus and are HIGH confidence.
The relative-strength values are a different matter and are uniformly LOW-to-MODERATE confidence: they come from one assay run by the 2006 Berkeley team, reported as the fluorescence of an RFP reporter driven by each promoter in E. coli strain TG1 grown in LB medium to saturation, normalized so BBa_J23100 = 1.00. All parts except BBa_J23119 were distributed in the same reporter plasmid (BBa_J61002); BBa_J23119 was distributed in a different backbone (pSB1A2), so it has no strength value that's directly comparable to the other 19 in this same table, despite the Registry's own text elsewhere calling it "the strongest member of the family" by design. Ordered from strongest to weakest reported value:
BBa_J23119 — 5'-TTGACAGCTAGCTCAGTCCTAGGTATAATGCTAGC-3' — the consensus sequence; no strength value directly comparable to the rest of this table (tested in a different plasmid backbone).
BBa_J23100 — 5'-TTGACGGCTAGCTCAGTCCTAGGTACAGTGCTAGC-3' — 1.00 (the reference value the rest of the table is normalized to).
BBa_J23102 — 5'-TTGACAGCTAGCTCAGTCCTAGGTACTGTGCTAGC-3' — 0.86.
BBa_J23104 — 5'-TTGACAGCTAGCTCAGTCCTAGGTATTGTGCTAGC-3' — 0.72.
BBa_J23101 — 5'-TTTACAGCTAGCTCAGTCCTAGGTATTATGCTAGC-3' — 0.70. (Note: this is the promoter Kelly et al. 2009 later chose as the reference for the unrelated RPU standard, discussed below — not the same reference as this table.)
BBa_J23111 — 5'-TTGACGGCTAGCTCAGTCCTAGGTATAGTGCTAGC-3' — 0.58.
BBa_J23118 — 5'-TTGACGGCTAGCTCAGTCCTAGGTATTGTGCTAGC-3' — 0.56.
BBa_J23108 — 5'-CTGACAGCTAGCTCAGTCCTAGGTATAATGCTAGC-3' — 0.51.
BBa_J23106 — 5'-TTTACGGCTAGCTCAGTCCTAGGTATAGTGCTAGC-3' — 0.47.
BBa_J23107 — 5'-TTTACGGCTAGCTCAGCCCTAGGTATTATGCTAGC-3' — 0.36.
BBa_J23110 — 5'-TTTACGGCTAGCTCAGTCCTAGGTACAATGCTAGC-3' — 0.33.
BBa_J23105 — 5'-TTTACGGCTAGCTCAGTCCTAGGTACTATGCTAGC-3' — 0.24.
BBa_J23116 — 5'-TTGACAGCTAGCTCAGTCCTAGGGACTATGCTAGC-3' — 0.16.
BBa_J23115 — 5'-TTTATAGCTAGCTCAGCCCTTGGTACAATGCTAGC-3' — 0.15.
BBa_J23114 — 5'-TTTATGGCTAGCTCAGTCCTAGGTACAATGCTAGC-3' — 0.10.
BBa_J23117 — 5'-TTGACAGCTAGCTCAGTCCTAGGGATTGTGCTAGC-3' — 0.06.
BBa_J23109 — 5'-TTTACAGCTAGCTCAGTCCTAGGGACTGTGCTAGC-3' — 0.04.
BBa_J23103 — 5'-CTGATAGCTAGCTCAGTCCTAGGGATTATGCTAGC-3' — 0.01. Effectively at the assay's noise floor; treat as indistinguishable from BBa_J23112 below rather than as a meaningfully weaker promoter.
BBa_J23113 — 5'-CTGATGGCTAGCTCAGTCCTAGGGATTATGCTAGC-3' — 0.01. Also at the noise floor.
BBa_J23112 — 5'-CTGATAGCTAGCTCAGTCCTAGGGATTATGCTAGC-3' — 0.00. Sequence-identical to BBa_J23103 (the Registry notes these two are "twins," independently isolated hits from the same screen that happen to share one sequence) and, like it, at the noise floor.
Getting the physical DNA
The Registry itself recommends two routes, and there isn't a strong reason to prefer a third:
If you're scripting a design pipeline rather than hand-copying from a wiki table, general-purpose sequence tooling can save the transcription step entirely — for example, SeqBench's MCP server exposes part lookups and sequence operations directly to an agent or script, which removes the specific failure mode this article opened with (a slightly-wrong copy of a 35-mer propagating into a plasmid map).
- De novo synthesis: because each promoter is only 35 bp, order it as two complementary single-stranded oligonucleotides and anneal them. This is the cheaper and faster option for a single promoter or a handful of variants.
- Registry distribution: request the physical part from the iGEM Registry's distribution. All members except BBa_J23119 arrive in plasmid BBa_J61002, which places an RFP reporter immediately downstream of the promoter; BBa_J23119 arrives in pSB1A2 instead.
Reading the "relative strength" numbers correctly
A relative strength of 0.47 for BBa_J23106 does not mean "47% of maximal promoter activity" in any absolute sense. It means: in one 2006 assay, an RFP reporter driven by BBa_J23106 in E. coli TG1 grown in LB to saturation fluoresced at 47% of the level produced by the same reporter driven by BBa_J23100 under the same conditions. It is a unitless ratio anchored to one reference promoter, one strain, one medium, and one timepoint. It is not transcripts per cell, not RNA polymerase molecules per second (PoPS), and not directly convertible into either without additional assumptions.
This is not just a theoretical caveat. A separate, peer-reviewed effort to standardize promoter measurement, Kelly et al., 2009, in the Journal of Biological Engineering, defined the Relative Promoter Unit (RPU) system specifically because absolute fluorescence readings varied substantially across labs and instruments. Their fix was to always report activity relative to an in vivo reference promoter rather than in absolute units. The promoter they picked as that reference was BBa_J23101, defined as exactly 1 RPU. That is a different reference promoter than the one this table normalizes to (BBa_J23100 = 1.00). So the same underlying idea, "strength relative to a reference," has already been operationalized twice in mainstream synthetic biology with two different reference promoters and, necessarily, two different numbering systems. A "relative strength of 1" in one paper's table and a "1 RPU" in another's are not interchangeable unless you check which reference each one used.
The same paper is also useful for the general point about context-dependence: Kelly et al. found that host strain alone shifted measured GFP synthesis rates by up to 2-fold, and plasmid copy number by up to 3-fold, on top of whatever the promoter sequence itself contributes. If your expression host, plasmid backbone, or growth conditions differ from E. coli TG1 in LB at saturation, don't assume the rank order in this table, let alone the specific ratios, will reproduce exactly.
A worked example: from one base change to a 100-fold reported drop
To make the sequence-to-strength relationship concrete rather than abstract, compare three members directly against the consensus (BBa_J23119, positions numbered 1-6 = -35 hexamer, 24-29 = -10 hexamer):
BBa_J23100 (strength 1.00) differs from the consensus at three positions, all inside the two hexamers: position 6 (A→G, inside the -35 hexamer) and positions 26 and 28 (inside the -10 hexamer). It's the reference value, not the consensus sequence itself.
BBa_J23108 (strength 0.51) differs from the consensus at exactly one position: position 1, where the consensus T is changed to C, right at the start of the -35 hexamer (TTGACA becomes CTGACA). One substitution, in this one assay, corresponds to roughly a 2-fold drop in reported activity relative to BBa_J23100 (1.00 → 0.51).
BBa_J23103 (strength 0.01) carries that same position-1 T→C change plus three more substitutions: one more in the -35 hexamer (position 5) and two in the -10 hexamer (positions 24 and 27, which converts TATAAT toward a much weaker match to the -10 consensus). Going from BBa_J23108's one hexamer substitution to BBa_J23103's four brings reported activity down roughly another 50-fold (0.51 → 0.01), for a combined ~100-fold drop relative to BBa_J23100.
The generalizable point isn't the specific fold-numbers (those inherit all the caveats above) but the pattern: across all 19 non-consensus members of this collection, essentially every difference from the consensus falls inside the -35 or -10 hexamer itself (two members, BBa_J23107 and BBa_J23115, each carry one additional spacer substitution). None of the strength variation in this panel comes from changing spacer length; it comes entirely from how well each variant's hexamers still resemble the sigma70 consensus boxes, which is exactly what you'd expect mechanistically from a sigma70-dependent promoter.
Why iGEM leans on this collection so heavily
The Anderson collection shows up in an outsized share of iGEM projects for reasons that are mostly about logistics rather than performance:
- It's short and cheap. At 35 bp, every member can be built as annealed oligonucleotides in an afternoon, without PCR, gene synthesis vendors, or sequence-verification turnaround.
- It's a BioBrick part in the conventional sense, meaning it follows the Registry's standard prefix/suffix convention, so it drops into the standard BioBrick assembly workflow (restriction-digest-and-ligate cloning with the shared flanking sites) without needing a bespoke adapter.
- It's pre-ranked. A team that needs "a weak constitutive promoter" or "the strongest constitutive promoter we can get quickly" doesn't have to design and characterize one from scratch in a single summer; they pick a rung on an existing, if imperfectly documented, ladder.
- It's already in the distribution kit most teams start from, which lowers the activation energy for using it to roughly zero compared to sourcing a promoter from a paper that isn't in Registry format.
Common mistakes, limitations, and what this table doesn't tell you
Being explicit about what this data does not establish is at least as useful as the table itself:
- None of these numbers are absolute expression levels. They're one lab's fluorescence ratios from 2006. If you need transcripts per cell, protein copies per cell, or PoPS, this table doesn't give you that, and neither does any secondhand reproduction of it.
- The assay conditions were narrow: one E. coli strain (TG1), one medium (LB), one timepoint (culture saturation, i.e., stationary phase), one reporter (RFP), one plasmid backbone for 19 of the 20 parts. Different strain backgrounds, media, growth phases, or plasmid copy numbers can and do change both absolute output and relative ranking, not just the scale.
- The bottom of the ladder is likely noise, not signal. BBa_J23103, BBa_J23112, and BBa_J23113 (0.01, 0.00, 0.01) and, to a lesser extent, BBa_J23109 and BBa_J23117 (0.04, 0.06) are close enough to zero that treating the differences between them as meaningful is almost certainly over-interpreting assay noise rather than reading real biology.
- BBa_J23103 and BBa_J23112 are the same DNA sequence. They were independently picked up as separate hits in the original screen and given separate part numbers, but they are not two different promoters; anything that reports them as having meaningfully different strengths (0.01 vs 0.00) is reporting noise on top of an identical input.
- BBa_J23119 is not actually ranked against the other 19. It's frequently described, including in the Registry's own prose, as "the strongest member of the family" because it's the literal consensus and consensus sequences are expected to bind sigma70 best. But it was never measured in the same reporter plasmid as the other 19 members, so there is no apples-to-apples number for it in this dataset. Treat "J23119 is strongest" as a design expectation, not a measured fact from this table.
- The built-in NheI and AvrII sites are a double-edged design feature. They make the promoter easy to re-engineer, but they are also just NheI and AvrII recognition sequences; if your finished construct has other copies of GCTAGC or CCTAGG elsewhere (in a multiple cloning site, a tag, or another part), a restriction-digest-based assembly step can cut somewhere you didn't intend. It's worth screening a full construct sequence with a restriction site finder, such as SeqBench's, before committing to a digest-based cloning plan built around one of these promoters.
- Promoter strength alone doesn't predict expression. RBS strength, 5' UTR secondary structure, codon usage of the downstream gene, mRNA and protein stability, plasmid copy number, and terminator efficiency all interact with promoter output, often outweighing small differences between adjacent rungs on this ladder (say, 0.47 vs 0.51). Use this collection for coarse tuning and expect to need RBS-level or copy-number-level adjustment for anything more precise.
- No burden or fitness cost data is included. A promoter that drives more reporter fluorescence is not automatically "better" for your construct; strong constitutive expression of a burdensome protein can slow growth or select against high producers over time, and that effect isn't visible in a single-timepoint fluorescence ratio.
Frequently asked questions
What is the strongest Anderson promoter?
Among the 19 members with a directly comparable measured value, BBa_J23100 is reported as strongest, with a relative strength of 1.00 (it's also the value the rest of the table is normalized to). BBa_J23119, the literal consensus sequence, is often called the strongest by design, but it was tested in a different plasmid backbone than the other 19 parts, so it has no strength value directly comparable to them in the original dataset.
Where can I get the exact Anderson promoter sequences?
The canonical source is the iGEM Registry of Standard Biological Parts, at parts.igem.org/Promoters/Catalog/Anderson, which lists all 20 sequences (BBa_J23100 through BBa_J23119) alongside the original 2006 relative-strength measurements. Because each promoter is only 35 bp, the cheapest way to obtain the physical DNA is to order it as two annealed complementary oligonucleotides rather than requesting it from a distribution kit.
Are the Anderson promoter relative-strength values accurate for my plasmid and strain?
Not necessarily. The published values come from one 2006 assay in E. coli strain TG1, grown in LB medium to saturation, using an RFP reporter in one specific plasmid backbone. A separate study (Kelly et al. 2009) found that host strain alone can shift measured promoter output up to 2-fold and plasmid copy number up to 3-fold, so the numbers here are a reasonable starting point for a coarse ranking, not a guarantee of the same ratios in a different experimental context.
Why do BBa_J23103 and BBa_J23112 have the same sequence?
They were independently isolated as separate hits from the same combinatorial promoter library screen and happened to be sequence-identical. The iGEM Registry documents them explicitly as "twins." Their reported strengths (0.01 and 0.00) differ only because of assay noise at the low end of the scale, not because the underlying DNA is different.
What's the difference between the Anderson promoter table and Relative Promoter Units (RPU)?
The Anderson table (from the 2006 Berkeley iGEM team) normalizes all values to BBa_J23100 = 1.00. The RPU standard, introduced by Kelly et al. in a 2009 Journal of Biological Engineering paper to reduce cross-lab measurement variation, normalizes to a different promoter, BBa_J23101, defined as exactly 1 RPU. The two systems are not interchangeable without knowing which reference each number is relative to.
Can Anderson promoters be used in bacteria other than E. coli?
They are sigma70-dependent promoters, and the iGEM Registry's own description states they are suitable for E. coli and "likely" other prokaryotes that use a similar sigma70 recognition system. That said, the Registry's actual characterization data (the strength numbers in this table) is E. coli-only, so function and relative strength in a different host still has to be verified experimentally rather than assumed.
Related references
Common restriction enzymes: recognition sites, cut positions, NEB buffer activity, star activity and an interactive double-digest buffer finder.
PCR primer design rules for length, GC content and Tm.
Reference thermal-cycling steps, temperatures, times and cycle numbers.