Reaction Setup

1. Assemble reactions in 250 µL PCR tubes on ice or cold block.
2. Close tubes, flick tubes a few times to mix, and centrifuge a few seconds to recollect the liquid.
3. Thermocycle according to reaction complexity.

Transformation

Transformation

Chemical/heat-shock transformation: transform ≤10% competent cell volume of Golden Gate assembly reaction to reduce dilution of transformation buffer. For 50 µL transformation, this equates to 1–5 µL of reaction.

Electrotransformation: transform 0.25–0.5 75 µL equivalent of reaction to minimize salt addition and risk of arcing. 1 µL may even be possible if higher efficiency is desired, as for small libraries. For large libraries, column-purify the assembly and elute plasmid with hot water and . Then transform as much DNA as necessary to maximize efficiency without losing too much of the library in double-transformations of library members (multiple plasmids per cell).

Thermocycling

Use the long protocol whenever you have time or trouble, as it is the most time-efficient.
The basis for the slight difference in BsaI and Esp3I protocol is not empirical. Unlike BsaI, Esp3I is not thermostable and so might not digest at 50°, thus the inclusion of a more moderate 42° final digest step.
The "Basic" protocol is said to be sufficient for basic part assemblies, which can also be run isothermal, 37° 1 hr. Reactions containing only 2–3 fragments are especially said to be fine with the basic protocol.

Isothermal Golden Gates are measured by NEB to have the highest fidelity, due to the higher overhang annealing stringency during higher temperature ligation. However, the lower ligation rate at higher temperature requires several-fold more isothermal reaction time to obtain similar product yield and resulting transformation yield.

Notes: Use a 37° incubator shelf or bead bath on the shelf to reduce wear to thermocyclers. Incubator floors and walls are hotter. NEB prescribes 16 hr incubation for 24-part assemblies. Shorter ought to work for as well for less complex assemblies. The 42° step is possibly only advantageous for Esp3I. Inactivation isn't necessary if immediately transforming. A mere 1 h incubation is found to work well for 2–3 part assemblies.

If one or more targeted restriction sites needs to be preserved in the product, it is best to omit final restriction steps and end the thermocycling in ligation-permissive conditions to maximize ligation at the retained restriction site. Requirement for religation greatly reduces efficiency of this kind of Golden Gate assembly, but is still reliable for assembly of fluorescent protein genes, as in fluorescent dropout vectors. Usage of a different type IIS enzyme is an alternative.

^{Endonuclease still has some activity at low hold temp, reducing efficiency ~tenfold. Since end-ligation is generally used to make vectors with fluorescent dropouts, the correct fluorescent colonies can easily be picked. Shyam found high-temp 85° inactivation (following the extended final 16° ligation) further reduced efficiency, but doesn't require a low-temp hold if not immediately transforming or freezing. To eliminate a majority of reactant plasmids with the Type IIs enzyme while also religating the desired dropout: after a standard protocol that includes a final high-temp digest and heat inactivation, a small quantity of additional ligase and ligase buffer can be added at the end and incubated at room-temp for ≥15 min prior to transformation.}

DNA Ligase

Concentration

• High concentration T4 ligase is needed for efficient isothermal Golden Gates of ten parts, but low concentration works just as well as high when cycling restriction/ligation 25–50 times. ^{Golden Gate Shuffling}
  

• High concentration T4 ligase is quite more expensive, so it is best to use low concentration and do thermocycling.

• Using more low-concentration ligase or using high-concentration ligase was found to increase misassembly rate, congruent with T4 ligase's known promiscuity.  ^EMMA
• Contrastingly, CIDAR MoClo and JBEI state using high concentration ligase as "essential". ^{J5 Golden Gate protocol, CIDAR MoClo}

Substrate Preference

• T4 DNA ligase can ligate blunt ends.
• T7 DNA ligase cannot ligate blunt ends, only efficiently ligating ≥2 bp annealed ends under normal conditions.
• NEB T7 ligase lots in 2015–2016 stopped working well for Golden Gate assembly, documented by Dueber Lab and Novome Biotech. T4 ligase should be used instead, despite Yeast Toolkit method.
• T4 ligase is very promiscuous and active, highly efficiently ligating nicks with 1+ mismatches near the ligation junction. ^NEB  They generally have higher fidelity for the 3′ nt than the 5′ nt. Helix-distorting R:R pairs are rejected more than Y:Y and R:Y mismatches, and helix-stabilizing multi-H-bond pairs such as G:T and A:G preferred over ones forming 0–1 H-bond. Both these effects lead to G:T mismatches being nearly as efficiently ligated as correct pairs.
• T7 ligase's higher fidelity is debated, perhaps only slightly higher.  ^Lohman
• Ligases prefer at least 10 bp on either side of the ligation junction, with activity dropping as this shortens, and longer flanking DNA being neutral.  ^Lohman
• Ligases search DNA randomly, so increasing non-substrate DNA inhibits a reaction by sequestering the enzyme.  ^Lohman

Reaction Temperature

• T4 ligase has optimal activity 16–20°C, permissive to maximal cohesive end annealing at 14–16°C.
• Hi-T4 ligase activity is measured at 25°C instead of 16°C, and it retains full activity after 72 h at 45°C or after 30× 5 min 50°–16°C thermocycling.
• T7 ligase has optimal activity at 25°C, much less permissive to equilibrium cohesive end annealing. A few degrees makes a big difference for 4 bp cohesion. When the Dueber Lab used to use T7 ligase, a 20° annealing temperature was used as a compromise between annealing and activity.
• There might be a benefit during long reactions of complex Golden Gates.
• The extremely high activity of T4 ligase still allows high ligation yields at room temp (20–25°C) or even 30–37°C, at which annealing is dynamic. As such, short The extremely high activity of T4 ligase still allows high ligation yields at room temp (20–25°C) or even 30–37°C, at which annealing is dynamic. As such, short 37°C isothermal Golden Gates are efficient for simple reactions or for complex reactions over a much longer 8–16 h incubation.
• Cohesion events at elevated temperatures occur with higher discrimination, increasing cohesive end ligation accuracy.
• T4 ligase is rapidly inactivated above 45°C ^Hi-T4. The highest recommended reaction temperature is thus 37°C ^Lohman .
• T4 and T7 ligases are heat-inactivated at 65°C in 10 min.

Buffer-protocol compatibility

• , as NEB confirms. 20 min not needed. High temperatures induce mutations in DNA, as found in an NEB-coauthored paper that found thermocycling itself causes mutation in DNA. So minimizing high temperature exposure is perhaps better.
• T7 ligase has optimal activity at 25°C, much less permissive to equilibrium cohesive end annealing. A few degrees makes a big difference for 4 bp cohesion. When the Dueber Lab used to use T7 ligase (as published in Yeast Toolkit), a 20° annealing temperature was used as a compromise between annealing and activity.

Buffer-protocol compatibility

• T4 ligase buffer is compatible with heat inactivation. Despite the salts, electroporation has been successful without reaction cleanup when transforming very small volumes (0.5 µL). 
• T7 ligase buffer contains PEG-3000; thus 65°C heat inactivation dramatically reduces transformation efficiency, and the reaction can neither be electroporated. ^NEB The PEG in Golden Gate enhancer, however, does not seem to be enough to have these effects.
• T7 ligase only has 10% activity in T4 ligase buffer. ^NEB  However:
• Higher specificity for correctly base-paired nicks if T7 ligase used with T4 ligase buffer instead of supplied T7 ligase buffer. ^NEB  The Dueber Lab and Novome used T7 ligase in T4 ligase buffer until 2015, when T7 ligase stopped working well, prompting a shift back to using T4 ligase.
• T4 ligase buffer was classically found to be superior to ATP-supplemented NEBuffer 3 for T4 ligase Golden Gate assembly.  ^Engler  Barrick Lab found NEBuffer 3.1 with 42° digests to be best for T7 ligase BsmBI reactions, better than CutSmart buffer, ligase buffer, or any with 37° digests; to be
• T4 ligase buffer is compatible with heat inactivation. Despite the salts, electroporation has been successful without reaction cleanup when transforming very small volumes (0.5 µL). 
• T7 ligase buffer contains PEG-3000; thus 65°C heat inactivation dramatically reduces transformation efficiency, and the reaction can neither be electroporated. ^NEB
• T7 ligase only has 10% activity in T4 ligase buffer. ^NEB  However:
• Higher specificity for correctly base-paired nicks if T7 ligase used with T4 ligase buffer instead of supplied T7 ligase buffer. ^NEB  The Dueber Lab and Novome used T7 ligase in T4 ligase buffer until 2015, when T7 ligase stopped working well, prompting a shift back to using T4 ligase.
• T4 ligase buffer was classically found to be superior to ATP-supplemented NEBuffer 3 for T4 ligase Golden Gate assembly.  ^Engler  Barrick Lab found NEBuffer 3.1 with 42° digests to be best for T7 ligase BsmBI reactions, better than CutSmart buffer, ligase buffer, or any with 37° digests; to be expected as BsmBI prefers high salt/temp. Buffer 3.1 is unlikely to do well for salt-sensitive T4 ligase, and higher incubation temp shouldn't help the mesophilic Esp3I. BsaI-HFv2 is optimized for full activity in T4 ligase buffer, whereas previous forms have 2.5-fold less activity than in CutSmart buffer, 37°.

• NEB and Promega T4 ligase buffers are 10×, allowing more DNA volume in a small reaction.
• NEB T7 ligase buffer is 2×, allowing 4 µL DNA volume in a typical 10 µL rxn with 1 µL total enzyme.
• T3 ligase has has a much higher salt tolerance, better resisting salt contamination in DNA samples (max 250 mM).

Golden Gate Enhancer

• Dueber Lab / Novome Biotech found 0.5–2.5% PEG-3350 in the reaction can improve Esp3I and BsaI Golden Gate reaction efficiencies 2-fold for a ten-part assembly, without reducing fidelity.
• Shyam likes to add 1× BSA to reactions to adsorb plasmid miniprep contaminants, increase enzyme stability in the reaction, and reduce enzyme loss to adsorption to the tube (reasons for which BSA is added restriction enzyme buffers).
• Shyam's formula for 10× Golden Gate enhancer: 1 mg/mL BSA, 10% PEG-3350.
  1° storage of the enhancer seems to grow contaminants sometimes. Might be best to freeze periodically., so look out for solids.
• PEG-containing reactions are typically instructed to not be heat-inactivated or electroporated, as the PEG quantity dramatically reduces transformation efficiency. The quantity used in the enhancer, along with the dilution for transformation, seems not to hamper transformation efficiency to a degree that is apparent. (≤10× dilution for chemical transformation, 5 µL rxn in 50 µL cells; or ≤66× dilution for electrotransformation, 0.5–0.75 µL rxn in 50 µL cells)
• NEB sells 3× Golden Gate master mix (NEBridge Ligase Master Mix) in which NEB sells 3× Golden Gate master mix (NEBridge Ligase Master Mix) in which you add your Type IIS enzyme of choice and DNAs. 

Type IIS Restriction Endonucleases

Here are all commercial ≥3 bp overhang-creating type IIS restriction IIS restriction endonucleases, and analysis of their properties for Golden Gate assembly.
Shyam's rank-order of preference are BsaI-HFv2, PaqCI, BbsI-HF, SapI (3bp), Esp3I/BsmBIv2.

BsaI,

BsaI-

HFv2

• • Instructed to be used with BSA (bovine serum albumin) in old buffer system, indicating that it is ≈essential at the recommended reaction temperature, 37°C. The new buffer system adds BSA to all standard restriction enzyme buffers for simplification, as it almost never reduces activity of enzyme that don't benefit from it.
  • NEB writes that BsaI digest does not require BSA for 100% activity at 50°C, but does for activity at the lower 37°C.
  • Neither T4 nor T7 ligase buffers of Promega or NEB contain BSA. Though the PEG in T7 ligase buffer may function similarly.
  • J5 Golden Gate protocol says that BsaI only has 10% activity at 37°C without BSA (bovine serum albumin), and thus requires adding BSA to Golden Gate reactions, as their protocol is performs digestion at 37°C. The CIDAR protocol seems to have overlooked BSA or discounted its utility in their optimization. ^{CIDAR MoClo}
  • NEB found that T4 ligase activity drops rapidly at temperatures starting 45°C, so 37° ought to be used for cycling digestion steps to protect the ligase, with BSA added to get full BsaI activity at 37° (along with absorbing miniprep contaminants). 
  • 0.5–2.5% PEG-3350 in the reaction can improve Golden Gate efficiency 2-fold for a ten part assembly. NEB warns that heat-inactivating reactions containing PEG inhibits their transformation, but the aforementioned PEG enhancement was determined with an 80°C, 10 min heat-inactivation; thus the transformation-inhibition effect may not manifest at lower PEG concentrations. ^{Dueber Lab / Novome / Shyam} NEB also warns not to electroporate PEG solutions, but in practice, direct electroporation of a few µL of unpurified GG reaction yields plenty of colonies for cloning purposes.
    

  • BsaI requires a full unit to digest a microgram of substrate in one hour or sixteen hours, meaning that it's active for only one hour. ^NEB BsaI-HF is active 4-8 hours, as is expected for -HFv2. 

  • Impaired by Dcm methylation at CCWGGTCTC sites.  ^NEB
  • NEB's BsaI-HF was found to have lower efficiency cutting at sites with an A/T spacer and generally performs worse in ten-part test Golden Gate assemblies. ^{Dueber Lab / Novome}
  • NEB's BsaI-HFv2 eliminates the spacer sequence context bias of BsaI and BsaI-HF, suggesting the bias of BsaI-HF was removed. is active 4-8 hours 
  • BsaI-HF and BsaI-HFv2 have has reduced star activity over WT BsaI and is active ≥16 hours, as is T4 ligase.
  • BsaI-HFv2 is optimized for full activity in T4 ligase buffer, whereas previous forms have 2.5-fold less activity than in CutSmart buffer, 37°.^{NEB help} This suggests BsaI-HFv2 does not require 1× BSA for full activity at 37°C in T4 ligase buffer like previous forms. However, 1×BSA may still be worthwhile to add with HFv2 as it might replicate in ligase buffer the 2-fold enhancement in activity at 55° that it sees in CutSmart buffer, as BSA's function is to stabilize protein.
  • BsaI-HFv2 appears to greatly reduce vector carry-through in a reaction and improves assembly efficiency with non-isomolar assemblies, but is neutral relative to BsaI in efficiency with isomolar assemblies and in misassembly rate.^Shyam 
  • All BsaI versions have 50–100% activity on sites that are 1 to 5 bp from DNA termini. ^NEB The 4–6 bp that are typical to add before the site on 5′ ends of primers are apparently unnecessary.

Esp3I

• • I only use 1 bp for inward-facing sites, where overhang is away from the end. Too little experience with the outward orientation.

PaqCI / AarI_Thermo

• • 7 bp recognition site, 4 nt overhang, 4 bp spacer
  • Requires supplied oligonucleotide addition for optimal digestion, as it requires two sites for cleavage. 
  • 7-fold as expensive as BsaI by activity. Lower concentration.
  • NEB PaqCI Golden Gate guide, protocol with T4 ligase
  • Both are reported to function in Golden Gate assemblies as well and robustly as BsaI-HFv2.

Esp3I

• • Mesophilic isoschizomer of BsmBI
  • Temperature optimum is 37°C, so it has full activity at a temperature conducive to T4 ligase survival.
  • Esp3I performs far better for Golden Gate assembly than 2015-era problematic NEB BsmBI lots, though BsmBI
  • Mesophilic isoschizomer of BsmBI
  • Temperature optimum is 37°C, so it has full activity at a temperature conducive to T4 ligase survival.
  • Esp3I performs far better for Golden Gate assembly than 2015-era problematic NEB BsmBI lots, though BsmBI still appears to digest DNA normally..   ^{Dueber Lab / Novome}
  • Thermo Esp3I performs far, far better for Golden Gate assembly than the newer NEB Esp3I, though NEB Esp3I still appears to digest DNA normally. ^{Dueber Lab / Novome.}
    
  • NEB Esp3I seems to substantially reduce in activity over a year, well before expiration.
  • 0.5–2.0.5–2.5% PEG-3350 in the reaction can improve Golden Gate efficiency 2-fold for a ten part assembly. NEB warns that heat-inactivating reactions containing PEG inhibits their transformation, but the aforementioned PEG enhancement was determined with an 80°C, 10 min heat-inactivation; thus the transformation-inhibition effect may not manifest at lower PEG concentrations. ^{Dueber Lab / Novome / Shyam} NEB NEB also warns not to electroporate/heat-inactivate PEG solutions, but in practice, direct electroporation of a few µL of unpurified GG reaction yields plenty of colonies for cloning purposesthis amount of PEG produces no problems.
  • Active for at least 5 hr at 37°C, as it requires 0.2 U to digest 1 µg substrate in 16 hr. ^manual
  • Has 50–100% activity on sites that are 1 to 5 bp from DNA termini. ^Thermo The 4–6 bp that are typical to add before the site on 5′ ends of primers are apparently unnecessary. Shyam switched to using 2 bponly 1 bp, at least when overhang is away from the end. Too little experience with the outward orientation.

BsmBI, BsmBI-v2

• • NEB lots in 2015–2016 stopped working well for Golden Gate assembly, documented by Dueber Lab and Novome Biotech with ten-part Golden Gate assemblies. Esp3I should be used instead, despite Yeast Toolkit methods section.
  • NEB recommends reaction at 55°C, but this harms the ligase ^Hi-T4. CIDAR MoClo doesn't use BsmBI, so doesn't make a recommendation.
  • Dueber Lab performed the digestion step at 45°C, 10° off optimum to less hurt the ligase, but it worked fine at the time.
  • Barrick Lab found NEBuffer 3.1 with 42° digests to be best for BsmBI reactions using T7 ligase, better than CutSmart buffer, ligase buffer, or any with 37° digests. ^Barrick
  • Not instructed to be used with BSA in the old buffer system; therefore BSA must not be required.
  • Active for 2–4 hr, as it requires only 0.5 U to completely digest 1 µg substrate in 16 hr, but cannot with 0.25 U. ^NEB
  • Has 50–100% activity on sites that are 1 to 5 bp from DNA termini. ^NEB The 4–6 bp that are typical to add before the site on 5′ ends of primers are apparently unnecessary.
  • v2 version has lower 10% reported activity at 37°C, compared to the WT's 20%.
  • Esp3I produces much more product yield than BsmBI-v2, and faster, according to Shyam's tests, especially under cycling conditions (37° temps for former, 42° for latter). Perhaps a lot more BsmBI-v2 enzyme is required.
  • NEB's BsmBI-v2 NEB's BsmBI-v2 Golden Gate kit is packed with concentrated enzyme to, I believe, make up for the much lower activity at the prescribed 42°C reaction temperature. If paired with Hi-T4 ligase, both enzymes ought to survive well at higher temperatures, but NEB recommendations may not be higher because overhang annealing rate further decreases, increasing required reaction time beyond reason.
  • Esp3I produces much more product yield and faster, according to Shyam's tests, especially under cycling conditions identical to that of BsaI-HFv2.

BbsI, BbsI-HF, BpiI

BbsI, BbsI-HF, BpiI

• • BbsI as ≤25% activity at temperatures ≤25°C, according to NEB. BbsI as ≤25% activity at temperatures ≤25°C, according to NEB. Thus, some activity during the ligation step.
  • BbsI must be stored at -80°C, not -20°. NEB tested full activity after ten freeze-thaws. Storing and using aliquots is safest.
  • BbsI-HF is improved to allow storage at -20°C.
  • BbsI-HF is improved to reduce star activity.
  • Not instructed to be used with BSA in the old buffer system; therefore BSA is not required.
  • BpiI is an isoschizomer of BbsI

Eco31I

• • Mesophilic isoschizomer of BsaI, performing at 37°C.
  • Untested for superiority over BsaI or BsaI-HFv2.
  • Active for at least 3⅓ hr, as it requires 0.3 U to digest 1 µg substrate in 16 hr. ^manual

BtgZI

• • Cuts 10 bp away from recognition, used optionally in GoldenBraid to cut over a BsaI site to release the same overhang. This could allow, say, a BtgZI+BsmBI-reactive multigene acceptor vector to be used in a BsaI cassette assembly, by using BtgZI to cut over BsmBI sites and reveal the same BsmBI overhangs, where using BsmBI would unwantedly cut inside the connectors. ^{[GoldenBraid]}
  • 100% activity at 60°C, 75% activity at 37°C.
  • Ineffective longer than 1 hr in rxn, like BsaI.

PaqCI / AarI_Thermo

• • 7 bp recognition site, 4 nt overhang, 4 bp spacer
  • Requires supplied oligonucleotide addition for optimal digestion, as it requires two sites for cleavage. 
  • 7-fold as expensive as BsaI by activity. Lower concentration.
  • NEB PaqCI Golden Gate guide, protocol with T4 ligase

SapI  BspQI  LguI

SapI  BspQI  LguI

• • 7 bp recognition site, 3 bp overhang
  • SapI and BspQI are ineffective longer than 1 hr in rxn, like WT BsaI.

BfuAI  BspMI  BveI

• • Require two sites on the DNA to cleave.
  • BfuAI cleaves plasmid DNAs more efficiently than BspMI.
  • BfuAI requires 50°C, with 50% activity at 37°C.
  • Oligonucleotide provided with BveI to add to reaction and assist cleavage, probably releasing unwanted overhangs into reaction.

Eco31I

• • Mesophilic isoschizomer of BsaI, performing at 37°C.
  • Untested for superiority over BsaI or BsaI-HFv2.
  • Active for at least 3⅓ hr, as it requires 0.3 U to digest 1 µg substrate in 16 hr. ^manual

BtgZI

• • Cuts 10 bp away from recognition, used optionally in GoldenBraid to cut over a BsaI site to release the same overhang. This could allow, say, a BtgZI+BsmBI-reactive multigene acceptor vector to be used in a BsaI cassette assembly, by using BtgZI to cut over BsmBI sites and reveal the same BsmBI overhangs, where using BsmBI would unwantedly cut inside the connectors. ^{[GoldenBraid]}
  • 100% activity at 60°C, 75% activity at 37°C.
  • Ineffective
  • 7 bp recognition site, 3 bp overhang
  • SapI and BspQI are ineffective longer than 1 hr in rxn, like WT BsaI.

BfuAI  BspMI  BveI

• • Require two sites on the DNA to cleave.
  • BfuAI cleaves plasmid DNAs more efficiently than BspMI.
  • BfuAI requires 50°C, with 50% activity at 37°C.
  • Oligonucleotide provided with BveI to add to reaction and assist cleavage, probably releasing unwanted overhangs into reaction.

EarI  EamEarI  Eam1104I

• • 3 bp overhang
  • Active >8 hr.

Literature Recommendations

CIDAR MoClo

The CIDAR MoClo protocol calls for ligase either from Promega or NEB, but each measures units differently. They must be referring to Promega's Weiss units and be using the high concentration Promega ligase, as 20 CEU NEB T4 or T7 ligase is unreasonable, and 20 U low concentration Promega ligase is too much volume.

Conversions based on Sambrook's Molecular Cloning (2001) definition: ^[2]  In 30 minutes at 16°C, 0.015 Weiss units of T4 DNA ligase should ligate 50% of the fragments derived from 5 µg of lambda DNA digested with HindIII.

• • NEB T4 DNA ligase
    1 CEU_T4= 50% ligation of 0.12 µM 5′termini (6 µg/20 µL) λ/HindIII, 16°C 30 min in 1× T4 Ligase Buffer 
    Implying: 55.5 CEU_T4 = 1 Weiss, using NEB's T4 ligase unit definition. (5 µg DNA / 0.015 Weiss)/(CEU_T4/6 µg DNA).
    Someone else says 66.6 CEU_T4 = 1 Weiss
    
    • 20 Weiss U of NEB T4 ligase => 2.78 µL of the low concentration 400 U/µL stuff.
    • 20 Weiss U of NEB T4 ligase => 0.55 µL of the high concentration 2000 U/µL stuff.
  • NEB T7 DNA ligase: 
    1 CEU_T7 = 50% ligation of 100 ng/20 µL λ/HindIII, 25° 30 min in 1× T7 Ligase Buffer 
    Implying: 3,333 T7 CEU_T7 = 1 Weiss, using the T7 ligase unit definition. (5 µg DNA / 0.015 Weiss)/(CEU_T7/0.1 µg DNA).
    • 20 Weiss U of NEB T7 ligase => 22 µL of the 3000 U/µL stuff.
  • Promega T4 DNA Ligase 
    0.01 Weiss = >95% ligation of 1 µg λ/HindIII, 16° 20 min 
    
    • 20 Weiss U of Promega ligase => 6.7–20 µL of the low concentration 1–3 U/µL low stuff.
    • 20 Weiss U of Promega ligase => 1–2 µL of the high concentration 10–20 U/µL stuff.

Golden Gate Shuffling

The Golden Gate Shuffling paper finds that high concentration T4 ligase is needed for isothermal Golden Gates of ten parts, but low concentration works just as well as high when cycling restriction/ligation 25–50 times.
They used 2.5 U BsaI (0.25 µL).

They claimed their results from altering annealing temperature from 16° to 20°, 25°, 30°, and 37° did not increase cloning efficiency, but the product they assembled to test this didn't have a visible phenotype like the Dueber Lab GFP cassette assembly test plasmid set.

Many incorrect assemblies they examined were formed from 3 correct and 1 G:T base pair. G:T base pairs are known to be accepted by T4 ligase. G:T must be regarded as a match when choosing maximally different overhangs.
The most common misassembly mechanism involved elimination of at least 1 nt at the 3′ end of an overhang, leaving a 5 bp overhang to which the end of an incorrect overhang anneals and is ligated. This reveals that three consecutive nucleotide matches must optimally be avoided in overhang choices.

EMMA

The EMMA paper presented a mammalian Golden Gate toolkit that makes a 2 transcriptional unit multigene directly from parts. 

They tested adding 0.25, 1, and 1.75 µL normal-concentration T4 ligase to equimolar 14-part 10 µL Esp3I Golden Gate assemblies, finding white colony percentages of 43%, 88%, and 98%, but with 6/6, 3/6, and 2/6 tested colonies correct. More ligase or high-concentration ligase can be concluded to increase misassembly rate, which makes sense given T4 ligase's promiscuity. Oddly, CIDAR and JBEI state using high concentration ligase as "essential".

EMMA also showed 0.5 and 1 µL Esp3I performed similarly, but 0.25 µL made more wrong constructs, though produced slightly more white colonies. Equimolar versus 25 ng of each of 11 parts made 68% and 55% white colonies and 8/8 and 6/8 correct assemblies. So equimolarity perhaps helps yield and fidelity.

EMMA used an initial 5 min 37° digest, longer cycling steps: 21×(37° 5 min,  16° 10 min), a unique final 20 min 16° ligation, followed by a long 30 min final 37° digest, before a 6 min 75° inactivation.

A unique step used in the EMMA paper was the addition of a nuclease (Plasmidsafe) + ATP after the Golden Gate reaction, presumably to digest linear, unligated DNA.

CIDAR MoClo Protocol

Densmore Lab CIDAR MoClo protocol, from supplement:

"Three protocols have been developed to optimize reaction time and cloning efficiency. The Standard Protocol provides >80% efficiency (>80% of all clones are correct with more than 200 colonies per plate in an 82.5 minute reaction time. It is ideal for simple 4-part + vector assemblies. The Troubleshooting protocol is used for more than 6 part + vector reactions and in Multiplexed MoClo to provide a larger number of correctly assembled clones. The Rapid protocol was designed for quickly assembling basic parts from PCR product or annealed oligos where only one part is being ligated to a vector. The longer initial digestion time could also be adopted in the Standard and Troubleshooting protocols to increase efficiency if needed." ^[1]a vector. The longer initial digestion time could also be adopted in the Standard and Troubleshooting protocols to increase efficiency if needed." ^[1]