Five Manuka honey samples, graded UMF 5+ through 20+, were dissolved in warm sterile water, pushed through a filter fine enough to strip out any stray microbes, and pipetted into 96-well plates in a descending series of concentrations. Alongside them went the imitations. One was plain sugar syrup mixed to the same carbohydrate profile as honey. The others were that same syrup, and separately plain water, spiked with methylglyoxal at exactly the concentration a certified laboratory had measured in each sample. Then each of four bacterial species was added to its own set of plates, which sat at 37 degrees Celsius for 20 hours.

The honey beat the imitations almost everywhere. The higher the UMF grade, the less honey it took to shut down bacterial growth, according to a study published in Microbiology. And neither the sugar nor the methylglyoxal, the compound most often named as the source of Manuka’s antibacterial punch, could reproduce what the whole honey did.

The ladder from 5+ to 20+

The measure here is the minimum inhibitory concentration, the lowest amount of a substance that stops at least 95 percent of bacterial growth compared with an untreated well. It is expressed as grams of honey per 100 millilitres of solution, so a lower number means a more potent preparation. It is a threshold for suppressing growth rather than a measure of killing, and no kill assay was run. Everything that follows happened in a well plate, on four single reference strains rather than patient isolates, with no biofilm and no airway model, three limits the authors themselves name in their limitations.

The four organisms were methicillin-susceptible Staphylococcus aureus, methicillin-resistant S. aureus, Klebsiella pneumoniae and Pseudomonas aeruginosa. Across five UMF grades the pattern held in the same direction for all four:

  • Methicillin-susceptible S. aureus: 20 percent at UMF 5+, falling to 3 percent at UMF 20+
  • MRSA: 15 percent, falling to 3 percent
  • K. pneumoniae: 30 percent, falling to 10 percent
  • P. aeruginosa: 25 percent, falling to 10 percent

The descent is not a smooth slope on every organism. P. aeruginosa sat at 25 percent for both UMF 5+ and 10+, and at 20 percent for both 12+ and 15+, before dropping.

The two staphylococci, meanwhile, behaved almost identically, which the authors take to suggest that methicillin resistance does not materially alter how these bacteria respond to honey. That is chemically sensible. The resistance mechanism is a modified penicillin-binding protein, and what it defeats is one class of antibiotic, not the general chemical stress a honey applies. The Gram-negative species needed substantially higher concentrations than the staphylococci at every grade, 25 percent against 20 at UMF 5+ and 10 percent against 3 at UMF 20+, which fits the outer membrane and efflux systems that make them harder to treat with most things. In P. aeruginosa both the honey and the sugar control produced a slight increase in growth at the very lowest concentrations before inhibition took over higher up.

Sugar alone did not do it

Honey is roughly four-fifths sugar, and concentrated sugar suppresses bacteria by pulling water out of them. That mechanism has been the standing deflationary explanation for honey’s reputation, and it needed testing rather than assumption. The control was built to match: glucose at 32 percent weight per volume, fructose at 40, sucrose at 10, dissolved and filter-sterilised the same way and diluted down the same series.

It worked, up to a point. Bacterial growth fell as the sugar concentration rose, confirming that osmotic stress does real work. But across all four species the whole honey suppressed growth significantly more than the matched sugar, and in the two staphylococci it drove absorbance to near zero at concentrations where the sugar solution was still supporting measurable growth. Statistical comparisons were made with multiple t-tests under Holm-Sidak correction across four biological replicates.

The authors flag a limitation on their own control. They matched the carbohydrate composition, not the osmolality, and they did not measure osmolality directly, so small differences in osmotic pressure between honey and syrup cannot be ruled out. They recommend that a future version of this experiment measure it.

With the same methylglyoxal, the plates behaved differently

Methylglyoxal is the compound that made Manuka’s reputation. It is a highly reactive dicarbonyl found in honey derived from the nectar of Leptospermum scoparium, it modifies bacterial proteins and nucleic acids by glycation, and it has been proposed for years as the principal driver of Manuka’s non-peroxide activity. It is also one of three marker compounds the UMF grading system measures, alongside leptosperin and dihydroxyacetone, and its concentration does climb steeply with grade: 138 parts per million at UMF 5+ against 1,114 at UMF 20+ in the samples used here.

So the test was direct. Take the measured methylglyoxal concentration for each sample, add that exact amount to water and to the sugar control, and see whether the copies match the original.

They did not, mostly. At UMF 5+ the honey stopped the two staphylococci at 20 and 15 percent while both methylglyoxal controls needed 40. From UMF 12+ upward the honey was inhibiting the staphylococci at 5, 4 and finally 3 percent while the controls stalled at 5 to 10. Against the Gram-negatives the methylglyoxal controls were not merely weaker but consistently so: for P. aeruginosa they never reached the inhibition threshold below 40 percent at any grade, and at the two lowest grades in water they had not reached it even at 40 percent, while the honey came down to 10.

The exception is UMF 10+, where honey, methylglyoxal in water and methylglyoxal in sugar all inhibited both staphylococci at exactly 10 percent. At that one grade against those two organisms, methylglyoxal alone reached the same threshold as the whole honey. The paper’s claim is not that methylglyoxal does nothing; the controls plainly inhibited growth on their own. The claim is narrower, that methylglyoxal alone, and methylglyoxal combined with a matched sugar matrix, cannot fully reproduce what intact honey does. What accounts for the remainder is not identified here. Acidity, phenolic compounds, peptides and other small molecules are the candidates the authors name, and one earlier study found that 3-phenyllactic acid enhances methylglyoxal’s bacteriostatic effect in model honeys tested against Bacillus subtilis.

The methylglyoxal arm is also the thinner half of the experiment. Its controls were run at three technical replicates per concentration, with no biological replication reported and no statistical comparison against the honey. The honey-versus-sugar comparison carries the statistics; the honey-versus-methylglyoxal comparison is a table of thresholds.

One disclosure belongs next to those numbers rather than in a footnote. Two of the four authors, Jackie Evans and Troy Merry, are employees of Comvita Ltd, the New Zealand company that supplied the honey. Comvita jointly funded lead author Gemma Allcott’s PhD studentship with Aston University, where she worked with Jonathan Cox, and Comvita determined the methylglyoxal concentrations the control solutions were matched to by high-performance liquid chromatography. The paper states that the funders had no role in study design, analysis, interpretation or writing, and adds that the methylglyoxal analyses were run by an IANZ-accredited laboratory under ISO/IEC 17025 as part of routine workflows without knowledge of their eventual use. A company selling UMF-graded honey has a commercial interest in both halves of this result: that the certified grade tracks potency, and that the single compound most often credited for that potency does not account for the honey on its own.

From a plate to a patient

Every experiment here was done in vitro, and the authors are direct about what that rules out. They write that oral consumption would not be expected to achieve therapeutically relevant concentrations at sites of respiratory infection, particularly in the lower airways. Any respiratory use would need purpose-built formulation, throat sprays, lozenges, gargles or inhaled preparations, each of which the authors say would have to address viscosity, dose delivery, airway tolerability, sterility, stability and possible effects on the respiratory microbiome. Their own framing is that these results support further mechanistic and formulation work, not clinical use.

The broader case for looking at alternatives at all is real enough. A modelling study estimated that bacterial antimicrobial resistance was associated with about 4.95 million deaths worldwide in 2019, with a 95 percent uncertainty interval of 3.6 to 6.6 million, and the antibiotic pipeline has not kept pace.

But the finding that came out of those plates is smaller and more specific than a cure. Across every organism tested, the number printed on the label tracked how much honey it took to shut their growth down. The compound that number is partly built on tracked it less well. In these five samples the grade was a reliable guide to what the honey did in the well plate, and the chemistry meant to explain the grade has some catching up to do.