A discordant “Ring Test” in Cannabis microbiological safety testing resolved with Next Generation Sequencing

Abstract

In November of 2015, a new cannabis testing lab (Lab XXX) contacted Medicinal Genomics Corporation in regards to microbial testing results that they wanted to present to state regulators. They collected cannabis samples from the local dispensaries and retested them with 3M Rapid Yeast and Mold tests and found colonies. Several other major testing labs in the state use PathoSEEK and 3M PetriFilms thus it was surprising to find passed samples for sale at dispensaries growing colonies. Lab XXX’s voluntary audit revealed colonies and they believed this data suggested a failed testing program in the state.  Lab XXX believed these were P.Citrinum and were questioning if P.Citrinum could evade detection with Medicinal Genomics qPCR. One colony in particular (Colony #4) was picked off of 3M Rapid Total Yeast & Mold (TYM) Petri Film and grown on more traditional agar for additional time in attempt to visually identify the suspected Mold.

To address the P.Citrinum hypothesis, we sent them our preprint publication (Below) demonstrating that PathoSEEK not only detects P.Citrinum but also P.Paxilli and that these often don’t grow on 3M PetriFilm Total Yeast and Mold plates in our experience (Table 1). These microbes also evade SimPlate and Biolumix detection but we quantifiably can detect them withPathoSEEK 18S qPCR (Total Yeast and Mold qPCR).

Cannabis microbiome sequencing reveals several mycotoxic fungi native to dispensary grade cannabis flowers

Lab XXX ordered a PathoSEEK kit and tried to replicate the work of the other dominant testing labs. The initial results presented with occasional TYM signal and with the internal control (SCCG) signal present. This looked normal on its face but drew concerns from Lab XXX as their 3M Yeast and Mold Plating presented a different picture.  Of concern, the TYM qPCR signal on triplicate preps were all producing SCCG signals but inconsistent TYM signals. These samples were  DNA prepped in triplicate (3 times) but with a slightly modified protocol (see below). This invoked further inquiry from our laboratory.

Screen Shot 2015-11-29 at 9.58.59 AM

We flew a Scientist to their lab to assess if the second hand thermocycler (BioRad) they obtained was calibrated and why their 3M Petri Films were presenting a different story. This lab did not purchase their cycler from Medicinal Genomics so we had never performed a validated install of this site. Customers are welcomed to purchase a Biorad directly from Biorad or through Medicinal Genomics. For labs that purchase their equipment directly from Biorad, we sell an install and validation service to for $10,000. This covers flights, hotel and our time. Despite this validation not having been purchased, we flew a scientist there anyway in attempt to resolve the discordance. Our Scientist oversaw a repeat of the experiment while on site.

The result came back with some concerns. The colonies picked off of an agar plate had no plant DNA present so a SCCG control was spiked into the sample. This is similar to how we test for edibles. Despite this spiked in control DNA, SCCG (Single Copy Cannabis Gene) control signal was not amplifying in all samples. This is an internal control designed to amplify Cannabis DNA to verify the DNA purification was capable of capturing DNA. If a reagent lyses plant DNA we know this lysis buffer and resulting DNA capture should also be capturing epiphyte and endophyte DNA. This SCCG amplicon is multiplexed into the qPCR of the microbial target in a different wavelength of light such that it acts as an internal control for every test. If this SCCG amplicon doesn’t amplify there is either a PCR setup error or the DNA prep failed. This had us dig further into the protocol to understand why spiked in DNA was inconsistently amplifying or failing to bind to the DNA purification beads.

After our scientist experienced different failure modes in their lab, it was revealed that the protocol they were using swapped out TSB (tryptic soy broth) for Butterfields Phosphate Buffer at the homogenization step. While labs have reported using water in place in TSB, this does not extend to any other broth or buffer. It is recommend to stick with TSB for all valid tests.

We believed this was one of the causes for confusion but couldn’t address the 3M vs qPCR question without first seeing functional internal controls. After explaining to Lab XXX that safety tests that have failing internal controls cannot be logically used to challenge certified tests where the internal controls behaved, we received this communication.

 “Nov 18th 2015- I will proceed to discuss this with the state at weeks end, I am more than comfortable with the successive sets of failure of the kit on this bug.  I do understand the other lab(s) may have used other traditional methods, however I can only speak to what we have performed here at [Lab XXX].  As to the broth, in the case of Mold and Yeast, its function is simply to strip off the bugs from the plant and serves no need to provide for overnight culture.”

When internal controls fail, it is probably prudent to hold off on challenging State (public) certified labs where the controls are consistent and functional and step back and ask why one cannot replicate what the other labs are performing routinely.  As it turns out, the simple “buffer swap” was not such an irrelevant change and it has us questioning all results generated with this buffer. Phosphate buffers can inhibit solid phase DNA capture protocols as the phosphates soak up the binding capacity of the solid phase designed to capture the Phosphodiester backbone in DNA. Despite sharing our concerns with the lack of functioning internal controls and the deviant homogenization protocol, the lab was still interested in reporting this to Public officials.

Screen Shot 2015-11-28 at 2.34.58 PM

Notice the 34 grams of phosphate used in Lab XXXs  Butterfield Phosphate buffer.
This is 13X more phosphate than what is present in TSB. Phosphate buffers are published as a mechanism to elute DNA off of solid phase capture beads thus preventing any DNA binding if this is used as a buffer to rinse microbes off of the plant. This would explain why the DNA prep failed, why no SCCG signal was observed even when we spiked an SCCG amplicon into the DNA prep. This also highlights why the internal controls are important as they signal poor protocol adherence and an inconclusive test. These types of internal controls are a benefit of qPCR that cannot be replicated with Petri films or cultures. This will become a very important detail as we move into Whole Genome Shotgun of these bugs.

For reference -From Wikipedia-

Tryptic soy broth

From Wikipedia, the free encyclopedia
Tryptic soy broth or Trypticase soy broth (frequently abbreviated as TSB) is used in microbiology laboratories as a culture broth to grow aerobicbacteria. It is a complex, general purpose medium that is routinely used to grow certain pathogenic bacteria, which tend to have high nutritional requirements (i.e., they are fastidious). Its agar counterpart is tryptic soy agar (TSA).[citation needed]
TSB is frequently used in commercial diagnostics in conjunction with the additive Thio which promotes growth of anaerobes.[citation needed]
Preparation[edit]
To prepare TSB, the following ingredients are dissolved under gentle heat and then autoclaved for 15 minutes at 121 °C (250 °F).

Nov 12, 2015- Medicinal Genomics Scientist brought DNA back from this lab and the lab shipped colonies of this 3M Rapid Yeast and Mold test to our facility for qPCR and whole genome sequencing to settle the issue. This was a colony that will not produce PathoSEEK results for Total Yeast and Mold but grows on a 3M Rapid Total Yeast and Mold Petri Film. This was suspected to be P.Citrinum. Samples were using Butterfields Phosphate Buffer as a homogenization buffer. The only sample using TSB were the colonies picked from the 3M Petri Film and the agar stab.

Multiple PathoSEEK tests are run using

  • Total Yeast and Mold & SCCG
  • Total Aerobic Counts & SCCG
  • Total Enterobacteria & SCCG
  • Total Coliform & SCCG

Screen Shot 2015-11-30 at 11.51.19 AMScreen Shot 2015-11-28 at 4.41.48 PMScreen Shot 2015-11-28 at 4.43.30 PM

Of note, the Total Aerobic Count (TAC) qPCR was positive for 4 samples of concern (Red) and implies 16S sequence from Bacteria was likely responsible for the colonies on the 3M Petrifilm Rapid Total Yeast and Mold test. Also notice that these 4 samples produced SCCG signal but not TYM qPCR signal. Below are the plating results from 3M performed at Lab XXX. This would explain the initial experiments where the SCCG controls worked, PathoSEEK implied no Yeast and Mold (counter to the 3M Petrifilm) yet later tests demonstrate positive TAC qPCR (16S). To resolve the conflicting data between qPCR of 16S DNA sequence (not 18S) and 3M PetriFilm Total Yeast and Mold, we resorted to Whole Genome Sequencing of the colonies in question.

Screen Shot 2015-11-28 at 4.52.25 PM

The YM=26.4 is the qPCR Cq. A Cq of 26 is usually 10,000 CFU/gram. Higher Cq = lower CFU/gram. TAC is the Total Aerobic Count Cq. All 4 Petri films have samples with 16S DNA from the Total Aerobic test. Sample TOG New (Right) were picked and Whole Genome Shotgun sequenced to ID the micro-organism.

Screen Shot 2015-11-30 at 11.19.49 PM

 

Heisenberg microbial uncertainty 

When the LabXXX was presented the sequencing data below, they claimed we sequenced the wrong samples and should have sequenced other samples from the experiment despite the fact that LabXXX shipped these films to MGC for sequencing. LabXXX supplied the MSDS and postage for these samples but upon seeing the data believes different samples should be sequenced. One artifact of the experiment is the shipment time on ice could be altering the colony and our 24 hr incubation in TSB could have altered the colony. While the later is certainly a possibility with a complex microbiome its seems difficult to believe with a picked colony. If culturing something for 24 hours (less than the recommended time for a 3M Total Yeast and Mold test), radically alters the microbiome then this theory alone undermines the entire use of Petri Films as a safety test to begin with as its longer growth time would induce a more significant representative change. This Heisenberg uncertainty (the act of measuring alters what you measure) seems more credible when a complex microbiome is plated on a new carbon source but less likely to occur form a picked colony. One benefit of qPCR is that the cells are lysed and DNA measured in absence of a carbon source selective pressure that could alter the readout of the original microbiome being measured.

The MSDS Lab XXX included in the shipment of the Petri Films are below. Notice the assumption that P.Citrinum is what was shipped from Lab XXX and these actions seem to directly contradict the claim of MGC sequencing the wrong colonies.

MSDS_P.Citrinin MSDS-P.Citrinin_2


Whole Genome Sequencing

Colonies numbered in the image above were picked and DNA was purified with SenSATIVAx. Nextera libraries were generated for 2 x 75bp sequencing on ILMN MiSeq V3 chemistry. 4 Libraries were generated and multiplex sequenced with Nextera indices.  Over 50 Million 75bp reads were sequenced generating over 3.8Gb of sequence with over 93% of the bases generated being over Q30 (99.9% accurate). For perspective, the Human Genome Project  (many of the Medicinal Genomics team members are authors on the paper) historically targeted a 2.8Gb human genome.
Screen Shot 2015-11-28 at 5.14.07 PM
These reads were assembled using CLCbio V4.9 genome workbench software. The attached history file has the assembly parameters utilized.
Fastq reads can be downloaded from Amazon S3 storage system and have been submitted to NCBI.
Forward Reads Culture 1– Reverse Reads Culture 1
Forward Reads Culture 2– Reverse Reads Culture 2
Forward Reads Culture 3– Reverse Reads Culture 3
Forward Reads Culture 4– Reverse Reads Culture 4
Assembly Parameters-
Assembly results-
All 4 cultures produce high quality shotgun assemblies with >250,000 base contigs and paired end read contiguity across large gene families.
Screen Shot 2015-11-28 at 5.42.10 PM
Top Contig Fasta Files
Top BLAST hits of Culture 1
All 4 cultures have irrefutable BLAST hits to Staphylococcus Saprophyticus, S.Xylosus and S.Aureus.
Screen Shot 2015-11-28 at 11.21.22 AM
Culture 2
Screen Shot 2015-11-28 at 1.07.46 PM
Culture 3
Screen Shot 2015-11-28 at 1.09.48 PM
Culture 4
Screen Shot 2015-11-28 at 1.38.52 PM
Having contigs over 200,000 bases long at 300X coverage with 99% identity to a Staphylococcus strain is an ironclad identification of a Staph strain. To identify the exact strain and if it is novel requires a more nuanced view of the genome and this literature on The Bacterial species definition in the genomic era.
We also threw the reads into OneCodex as they have user friendly user interfaces for whole genome shotgun data. These reads were not trimmed before OneCodex analysis and the insert sizes are short enough to read over adapters and lower the yield of reads matching the isolate. Take theses results with some caution until we present them with trimmed reads.
Trimmed of Nextera sequence and quality filtered at 0.01 threshold in CLC Bio V4.9

Schrödinger’s Panda
Staphylococcus Xylosus is a published pathogen from mammalian fecal matter and soils. The >100Kb hit to strain HKUOPL8 is a strain discovered in Panda feces. The Schrödinger’s Panda reference is to express the fact that we doubt this panda actually exists in this state but also doubt this Staphylococcus is very selective over its carbon source. Since it grows on 3M Yeast and Mold Petri Films, it likely behaves facultatively (like many of its diverse relatives do) and thrives in more places than just Pandas. S.Saphrophyticus is the pathogen responsible for most human Urinary Tract Infections (UTIs). While these draft assemblies undeniably identify the colony as a Staphylococcus strain (NOT A Yeast and Mold), its exact strain ID may be novel or require further analysis. The species is known to be diverse and a pathogenic concern in  Canada and the European Herbal Pharmacopoeia. 
Panda
3M’s Specification sheet is here and specifies its Exclusion criteria.

 “EXCLUSIVITY ORGANISMS

Enterococcus faecalis, Bacillus spizinzenii, Aeromonas hydrophila and Escherichia coli were tested as representative Gram positive and Gram negative bacteria to demonstrate the selectivity of three lots of 3M Petrifilm Rapid Yeast and Mold Count Plates against bacteria”.

Of note- there is no mention of Staphylococcus or that this was tested with a terpene and cannabinoid rich background matrix (30% w/v). Some terpene and cannabinoids are known antiseptic compounds while others are known carbon sources.  Thus any validation performed on food products must be re-evaluated with a matrix that could alter the carbon source on the Petri Film.


Discussion

After protocol discrepancies were resolved, samples in question were showing signal on Total Aerobic Count qPCR, and on some samples qPCR for total Yeast and Mold. Samples that were negative for Total Yeast and Mold (TYM) qPCR but we positive for TAC were shipped to Medicinal Genomics and selected for sequencing. These colonies that were cultured on 3M Rapid Total Yeast and Mold were whole genome shotgun sequenced to demonstrate Staphyloccus.

The 3M Petri Film was shipped over night on ice and endured slightly more growth time than the 24-48 hour recommendation but the colonies were identified prior to shipment and marked with a pen for sequencing (See picture). It is rare for a mold to outgrow a bacteria on an overnight shipment thus we believe the Staphylococcus strain found on 3M Rapid Yeast and Mold Petri Film is the source of the TAC signal witnessed with freshly extracted DNA (no growth).

Conclusions

  • 3M Rapid Total Yeast and Mold Petri films are not specific to Yeast and Molds and can harbor Staphylococcus growth. Staphylococcus is a human pathogen and is a required microbial food test in Europe and Canada according to the EHP. The TAC qPCR results, while detecting Staphylococcus, would not have failed it according to TAC thresholds. Species specific tests are often suggested for this reason.
  • Internal controls are unique to qPCR methods and offer guidance on failed samples and false negatives.
  • PathoSEEK is a DNA based test that can specifically differentiate different species and avoid false conclusions. Culture based tests can not Speciate microbes and are a gross measure of risk. Evolution however has usually ensured most microbes can utilize multiple carbon sources and these pathways are rarely unique to different species. DNA is unique to different species and is the scientifically agreed upon taxological tool in the case of visual discrepancies.
  • Given the samples were acquired and tested outside of a seed to sale auditing system, it is unclear if the Staphylococcus originated in the dispensary or in the testing lab that choose to re-audit this process. Given the testing lab had limited experience with the relevance of culture broths and internal controls, it is unclear if sterile technique and GLP precautions were taken to avoid human contamination of samples being tested. In any microbial assay, the contaminant must be grown to detect it. This makes microbiological labs uniquely different from pesticide or cannabinoid testing labs as microbiology labs amplify their contaminant 1000s to millions fold in order to measure it making contamination a constant risk. For this reason PathoSEEK has a decontamination system designed into the assay to avoid post amplified products from contaminating pre-amplified laboratories. This method was published in Nature Biotechnology and PLOSOne. There is no such decontamination safeguard with culture based methods and therefore they are particularly prone to False positive and False negative results that can lead into expensive rabbit holes full of Panda feces (to be fair, it is unlikely this Staph.X originated from Panda Feces as it can be human or soil originating).
  • 3M has 19 Yeasts and 21 Molds they have tested with their Rapid Petri Film Total Yeast and Mold kit. None of these tests were performed with Cannabis background matrix. It is well known that Cannabis flowers are rich in terpenes and carbon sources that influence and harbor microbial Epiphytes and Endophytes. The 3M platform should be Cannabis validated given the carbon rich nature of the cannabis flower background and the known anticeptic properties of cannabinoids to resolve the lack of selectivity of its tests in this background matrix.
  • A proper ring test should bulk homogenize samples and test them simultaneously. The temporal nature of sterility would advise that methods that require long periods of time to provide a result may be irrelevant by the time the test results are in. Thus any ring test that is spatially or temporally uncontrolled will always lead to lab to lab discrepancies unrelated to the skill of the lab but more related to the fragmented nature of the experimental design.

Appendix

Through the course of this work it became clear that the 3M platform has been tested on only a limited number of Yeasts and Molds and all of these species must be culturable in order for a lab to purchase them from a vendor like ATCC. This severely limits their capability of validating against all microbes potentially present on cannabis.

To put this in to perspective, the below microbes have 18S evidence of existence on Cannabis samples and a large number of them are not available at ATCC to even verify if they can be grown on 3M. The numeric columns are the number of high quality 250bp reads discovered via sequencing the 18S amplicons. With such rich molecular data some curation is in order. Notice Zea Mays (corn) present as hits in this data. Corn has been sequenced and these hits could either be cannabis ITS sequences appearing as corn or microbial contamination in  the corn genome assembly. For this reason, our paper above confirms pathogenic microbes of concern with toxin specific gene sequencing. Both P.Citrinum and P.Paxilli have been confirmed on Cannabis samples with pathway specific amplification and sequencing.

 

Screen Shot 2015-12-04 at 11.38.29 AM

The below alignment shows the primer sequences match to the 1.4kb Corn region. You can see there are so many(9) mismatches (3 prime ones too) in the primers that this is unlikely to have amplified this region of corn and is more likely a misclassification of an uncultureable sequence yet to be described in NCBI.

Screen Shot 2015-12-05 at 2.24.01 PM

 

P.paxilli and P.citrinum were ordered from ATCC. These were grown in TSB and amplified with PathoSEEK to verify the qPCR methods can in fact detect these organisms in a monoculture.

Screen Shot 2015-12-19 at 12.31.34 PM

 

Why Sequencing is better than Plating

Why the FDA is moving all food borne pathogen testing to genomics