AMERICAN SOCIETY FOR MICROBIOLOGY
Microbiology Spectrum®
Check for updates
Identification of intratumor bacteria-associated prognostic risk score in adrenocortical carcinoma
Linyi Tan,1 Dengwei Zhang,2 Yong-xin Li,2 Yuqing Li,1 Ting Guo,3 Yang Sun,4 Ning Li,5 Chenchen Feng1
AUTHOR AFFILIATIONS See affiliation list on p. 10.
ABSTRACT A landmark study by Poore et al. showed intratumor bacteria (ITBs) playing a critical role in most cancers by reproduction of The Cancer Genome Atlas (TCGA) transcriptome data. A recent study by Salzberg et al. argued that ITBs, being overstated as a methodology by Poore et al., were problematic. We previously reported that ITBs were prognostic in adrenocortical carcinoma (ACC), a highly aggressive rare disease using data by Poore et al., and here, we aimed to answer whether ITBs truly existed and were prognostic in ACC. ACC samples from our institutes underwent 16S rRNA sequencing [adrenocortical carcinoma blocks from Huashan Hospital and China Medical University (HS) cohort]. The ITB profile was compared to TCGA data processed by Poore et al. (TCGA-P) and TCGA data processed by Salzberg et al. (TCGA-S), respectively. The primary outcome was overall survival (OS). A total of 26 ACC cases (HS cohort) and 10 paraffin controls were sequenced. The TCGA cohort encompassed 77 cases. Two and four amid the top 10 abundant genera in HS cohort were not detected in TCGA-P and TCGA-S, respectively. Neither was alpha or beta diversity associated with survival nor could ACC be subtyped by ITB signature in the HS cohort. Notably, a five-genera ITB risk score (Corynebacterium, Mycoplasma, Achromobacter, Anaerococcus, and Streptococcus) for OS trained in the HS cohort was validated in both TCGA-P and TCGA-S cohorts and was independently prognostic. Whereas ITB signature on the whole may not be associated with ACC subtypes, certain ITB features are associated with prognosis, and a risk score could be generated and validated externally.
IMPORTANCE In this report, we looked at the role of ITBs in ACC in patients with different race and sequencing platforms. We found a five-genera ITB risk score consis- tently predicted overall survival in all cohorts. We conclude that certain ITB features are universally pathogenic to ACC.
KEYWORDS adrenocortical carcinoma, intratumor bacteria, prognosis
A drenocortical carcinoma (ACC) is a rare disease with occult heralding symptoms. Though curable at the localized stage, most patients are diagnosed at an advanced stage, which confers dismal survival. Identification of novel prognostic markers is urgently needed.
In recent years, next-generation sequencing has brought surprise in cancer studies when scholars identified non-human reads that could be mapped to microbial signature (1). Intratumor microorganisms, especially intratumor bacteria (ITBs), have gained the most attention, given their abundance in comparison to fungi, archaea, or virus. ITB signature has shown a prognostic effect in a variety of cancer types, including gastric cancer (2), colorectal cancer (3), and hepatocellular cancer (4), establishing the micro- biome as a novel omics or “second genome” of cancer (5).
Unlike gene-association studies, microbial signature in cancer is largely impacted by contamination, especially in organs conventionally regarded as “sterile” (6). In breast
Editor Wei-Hua Chen, Huazhong University of Science and Technology, Wuhan, China
Address correspondence to Chenchen Feng, fengchenchen@fudan.edu.cn, or Ning Li, drlining_uro@163.com.
Linyi Tan and Dengwei Zhang contributed equally to this article. Author order was determined based on ther contribution to the article.
The authors declare no conflict of interest.
See the funding table on p. 10.
Received 20 October 2023 Accepted 15 February 2024 Published 29 February 2024
Copyright @ 2024 Tan et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license.
cancer, for instance, only one to two ITBs can be spotted in every ~10,000 cells scanned (7). This makes identification of true ITBs in cancers with low biomass substantially challenging as a vast majority of microbial reads detected are in fact contaminants (8). Nonetheless, since the study of the cancer microbiome is a newly emerging field, a standardized decontamination process is not yet available. Recently, even the landmark study by Poore et al. (1), reproducing RNA-seq data in The Cancer Genome Atlas (TCGA) cohorts to establish an intratumor microbial biomarker, has been challenged recently for potential fallacy in the decontamination process (9).
Our group has been the first to report ITB signature in ACC (10). Using 16S fluores- cence in situ hybridization (FISH) and lipopolysaccharide (LPS) staining, we observed that the existence of ITBs in ACC and reproduction of TCGA-ACC data set using microbial reads converted by Poore et al. (1) yielded an independent prognostic signature that supersedes the current staging system of ACC (10). Interestingly, a recent report by Cantini et al. showed an association between ITBs and response to mitotane, a adreno- lytic medication used in advanced-stage ACC (11).
Together, ITBs could be playing important roles in ACC yet to be elucidated, whereas ITB composition could largely be impacted by contamination, sequencing technique, and even race (12). We thus aim to evaluate whether ITB signature could be extrapolated independently. In the current study, we have sequenced our own ACC samples with a rigorous decontamination process. We owe great thanks to Salzberg et al. (9) for providing their reproduction of TCGA-ACC data. Here, we have developed an ITB risk score in our own cohort and validated our findings in a TCGA data set processed by two top groups (1, 9), where all cohorts differ in sequencing technique, patient race, and processing algorithm.
MATERIALS AND METHODS
Sample collection
A total of 26 formalin-fixed paraffin-embedded (FFPE) ACC blocks from Huashan Hospital and China Medical University [termed adrenocortical carcinoma blocks from Huashan Hospital and China Medical University (HS) cohort] were included. Patients underwent surgical removal of tumors between 2014 and 2022. All samples were re-accessed by an independent pathologist for validation of pathological diagnosis.
16S rRNA sequencing
All samples were sequenced in one batch. Ten FFPE blocks randomly chosen from the 26 cases were sequenced as control. Total genome DNA from samples was extracted using the cetyltrimethylammonium bromide (CTAB) method. DNA concentration and purity were monitored on 1% agarose gels. According to the concentration, DNA was diluted to 1 ug/uL using sterile water. 16S rRNA genes of a distinct V4 region were amplified using specific primer 16S V4:515 F-806R with the barcode. PCR reactions were carried out with 15 µL of Phusion High-Fidelity PCR Master Mix (New England Biolabs), 0.2 uM of forward and reverse primers, and about 10-ng template DNA. Thermal cycling consisted of initial denaturation at 98℃ for 1 min, followed by 30 cycles of denaturation at 98℃ for 10 s, annealing at 50℃ for 30 s, and elongation at 72℃ for 30 s, and finally, 72℃ for 5 min. The same volume of IX loading buffer (containing SYB green) was mixed with PCR products, and electrophoresis was performed on 2% agarose gel for detec- tion. PCR products were mixed in equidensity ratios. Then, mixture PCR products were purified with Qiagen Gel Extraction Kit (Qiagen, Germany). Sequencing libraries were generated using TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, USA) following the manufacturer’s recommendations, and index codes were added. The library quality was assessed on the Qubit (v.2.0) fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on an Illumina NovaSeq platform, and 250-bp paired-end reads were generated. Paired-end reads were assigned to samples
based on their unique barcode and were truncated by cutting off the barcode and primer sequence. Paired-end reads were merged using FLASH (v.1.2.7, http://ccb.jhu.edu/ software/FLASH/) (13), a very fast and accurate analysis tool which was designed to merge paired-end reads when at least some of the reads overlap the read generated from the opposite end of the same DNA fragment, and the splicing sequences were called raw tags. Quality filtering on the raw tags were performed under specific filtering conditions to obtain the high-quality clean tag according to FASTP (14). The tags were compared with the reference database (Silva database, https://www.arb-silva.de/) using UCHIME algorithm (15) to detect chimera sequences, and then the chimera sequences were removed (16) before effective tags were finally obtained.
Decontamination
Given the retrospective nature of the current study, contamination was inevitable during and after sample embedment. Processing of FFPE samples before sequencing thus maximally conformed to the recommended protocol to minimize additional contami- nation (17). Besides regular decontamination process for sterile organs (11, 18), we designated an algorithm workflow to remove blacklist contaminations (19, 20): ASVs (amplicon sequence variant) of contamination were identified based on the “decontam” algorithm, using the “isNotContaminant” function based on the “prevalence model”; ASVs with a relative abundance of more than 0.5% in the control FFPE samples were removed; ASVs that appeared in less than 5% of the tissue sample were also removed to avoid contingency. Normalized read counts of ITBs were subject to further analyses for clinicopathological association.
Processing of TCGA data
We previously retrieved an NR cluster of TCGA data processed by Poore et al. (TCGA-P cohort), which was ready for analyses (10). Data for the TCGA data processed by Salzberg et al. (TCGA-S) cohort were kindly provided by Professor Steven Salzberg, originally retrieved from the TCGA-ACC data set and processed by a pipeline reported by the team recently (9). In brief, the TCGA-S encompassed microbial counts substantially lower than TCGA-P. Caveat should be taken as our group did not have access to raw transcriptome data of TCGA, so that neither data from TCGA-P nor data from TCGA-S were externally validated by us.
Genomic analysis and visualization
Somatic variants were acquired by the GDCquery_Maf function in the R package “TCGAbiolinks” (21) and were visualized by the R package “maftools” (22). The chi- squared test was used for comparing the frequency of variations between subgroups, and the Wilcoxon test was used for measuring the tumor mutation burden (TMB) and copy number variant (CNV) differences.
Transcriptomic analysis and visualization
The R package “limma” (23) was used for identifying the differentially expressed genes with a cutoff of 1 for log-transformed fold change. The enrichment analysis of differen- tially expressed genes was performed using enrichKEGG and gseKEGG functions in the R package “clusterProfiler” (24). The Quantiseq algorithm was used for evaluating immune cell infiltration by the deconvo_tme function in the R package “IOBR.”
Statistical analyses
Read counts of ITBs were matched to survival data. All statistics were run using R script. Diversity of ITB and subtype probing was investigated as previously reported (10). Log-rank test was used as the univariate test to identify candidate features, which were subjected to least absolute shrinkage and selection operator (LASSO). Cox regression
analysis was conducted using the glmnet package. The multivariate Cox proportional hazards model was employed to investigate independence of ITB risk scores. The P value of <0.05 was accepted as statistical significance.
RESULTS
All cases in he HS cohort were of Chinese Han ethnicity. Loss of follow-up occurred in two cases for overall survival (OS) and in four cases for progression-free survival (PFS), respectively. The TCGA cohort encompassed 77 and 72 cases with OS and PFS, respectively. There were 14 female and 12 male ACC patients in the HS cohort at an average age of 51.7 + 14.3 years. Demographic data for cases in the TCGA-ACC cohort can be found in our previous report (10). Among the five clusters of TCGA-P, we decided to use NR clustering in the current study for validation as it yielded best area under the curve (AUC) in prognosis prediction in our previous model (10). Contamination is the key factor biasing interpretation of ITBs in cancers with low biomass. Using an in-house developed rigorous decontamination process that controlled both for FFPE and blacklist contaminants, we showed a drastic drop in read counts (Fig. 1A). Blacklist contaminants that were highly abundant in both FFPE control and in tumor samples substantially dropped after our decontamination process (Fig. 1B).
Bacterial reads that survived decontamination were presented in relative abundance and annotated phylum and genus levels. Top five abundant phyla are presented in Fig. 1C, four of which, namely, Actinobacteriota, Firmicutes, Bacteroidota, and Proteobacteria were present in all cases (Fig. 1C). Taxa varied at genus level between cases (Fig. 1C). Such diversity was also present across cohorts as solely Bacillus and Pseudomonas were among the top abundant genera in the TCGA-ACC cohort (10). Interestingly, Pseudomonas was also present in the study by Cantini et al. (11).
As no paired healthy adrenal tissue was available, we compared diversity between demographic parameters. Female patients showed a significantly worsened PFS and a numerically worsened OS in the HS cohort (Fig. 1D). We compared alpha diversity by three indices, but none was altered between sexes (Fig. 1E), neither was Shannon index correlated with age (Fig. 1F). Alpha diversity was correlated with neither OS nor PFS (Fig. 2A). We next explored whether the ITB subtype existed in ACC sequenced with 16S rRNA,
100
100
A
B
C
Read counts
Common contaminants
Relative abundance (%)
Relative abundance (%)
Before decontamination
After decontamination
80000-
:
75
75
Number of reads
0.75-
60000-
50-
50
0.50
40000-
0.25-
25
25
20000-
:
a.aa-
4
”
0
Before After Decontamination
FFPE control
ACC tumor
ACC tumor
Phylum
Genus
Pseudomonas
Atopobium
Contaminants in blacklist
Yes
Others
Firmicutes
Others
Alteromonas
Muribaculaceae
No
Acidobacteriota
Bacteroidota
Bacillus
Idiomarina
Marivita
D
Actinobacteriota
Proteobacteria
Lactobacillus
Bacteroides
Phaeodactylibacter
Overall survival
Progression-free survival
E
F
1.00
1.00-
Chaot
Shamnon
Simpson
6
Survival probability
Wilcowon, p = 0.82
5-
Wilcoxon, p = 0.74
1.0
Alpha deversity measure
Wilcoxon, p = 0.82
R2 < 0.01
0.75
Male
Male
300
0.75-
Shannon index
Female
4-
0.50
0.50-
200
+
Female
a-
$
0.25
p = 0.17
0.25
p = 0.043
100-
:
.
2-
0.00
2
0.00
0
40
80
0
40
80
F
M
F
M
F
M
20
40
80
Months
Months
Age (year)
A
Overall survival
Progression-free survival
Survival probability
B
1.00
1.00
R2=0.0343
p-value=0.647
Overall survival
R2=0.0468
p-value=0.372
Progression-free survival
Shannon
0.6
high
0.75
Long survival
0.75-
Shannon high
Long survival
PCoA2: 10.95 %
0.3
PCoA2: 10.95 %
0.4
0.50
0.50
Shannon low
Shannon low
0.0
0.0
0.25
p = 0.51
0.25
p = 0.55
-0.3
Short survival
-0.4
Short survival
0.00
0.00
0
40
80
Months
0
Months
40
80
-0.6
-1.0
-0.5
0.0
PCoA1: 32.35 %
0.5
1.0
-1.0
-0.5
0.0
PCoA1: 32.35 %
0.5
1.0
C
Overall survival
Progression-free survival
D
1.0-
1.00
1.00
Survival probability
PAM cluster 2
Relative abundance (x100%)
0.8
Genus
0.75
0.75
PAM cluster 1
Atopobium
0.6
Bacteroides
Muribaculaceae*#
0.50
PAM cluster 1
0.50
Phaeodactylibacter#
Bacillus
PAM cluster 2
0.
Lactobacillus
Marivita*#
0.25
p = 0.35
0.25
p = 0.47
Alteromonas
Idiomarina#
0.2
Pseudomonas
0.00
0.00
*Not present in TCGA-P “Not presnet in TCGA-S
0
40
80
0
40
80
Months
Months
0.0
HS
TCGA-P TCGA-S
though we successfully established a dichotomized signature by reproduction of the transcriptome data of TCGA based on overall survival (10). By dichotomizing according to median length of survival, we did not find differential beta diversity (Fig. 2B). Using ITB signature as a whole also failed to subtype ACC patients according to OS or PFS (Fig. 2C). Surprisingly, we showed here that ITBs, on the whole, were not associated with key clinicopathological parameters of ACC, despite the limited sample size and the rigorous decontamination process.
The discrepancies with our previous reports that ITB signature was not adequately informative to subtype ACC prompted us to answer another question, namely, whether certain ITB features were associated with prognosis in ACC ubiquitously. We first queried the top 10 abundant ITBs in the HS cohort in TCGA-P and TCGA-S, respectively. Of note, two genera (Muribaculaceae and Marivita) were not present in the TCGA-P cohort, and two other genera (Phaeodactylibacter and Idiomarina) were additionally absent in the TCGA-S cohort (Fig. 2D). In general, composition was much similar between the TCGA-P and TCGA-S cohorts than between the HS and TCGA cohorts (Fig. 2D). Stand- ardized Schoenfeld residual validated a LASSO model for feature selection (Fig. S1), which showed five genera (Corynebacterium, Mycoplasma, Achromobacter, Anaerococcus, and Streptococcus) as risk factors in all three cohorts consistently (Fig. S2). A risk score generated from the five features showed significant prognostication in all three cohorts (Fig. 3A). The score generated an AUC of 0.8 in the HS cohort and 0.7 in both the TCGA-P and TCGA-S cohorts (Fig. 3B). Whereas multivariate analyses showed marginal independence of ITB risk score in HS cohort (P = 0.069), it showed independence in both TCGA-P and TCGA-S cohorts (Fig. 3C). We next explored ITB features associated with PFS and identified two genera showing Paenibacillus being protective and Luteibacter being a risk factor in all three cohorts (Fig. S3). However, we did not obtain a reproducible risk score for PFS with the two features. The best model was trained in the TCGA-P cohort,
A
Cumulative survival (%)
HS cohort
TCGA-P
TCGA-S
1.00
1.00
1.00
Low risk score
0.75
Low risk score
0.75
Low risk score
0.75
0.50
0.50
0.50
High risk score
0.25
p = 0.034
High risk score
0.25
p = 0.037
High risk score
0.25
p = 0.0032
0.00
0.00
0.00
0
20
40
60
80
0
20
40
60
80
100
120
140
0
20
40
60
80
100
120
140
Time (months)
B
HS cohort
1.00
1.00
TCGA-P
1.00
TCGA-S
0.75
Sensitivity
0.75
0.75
0.50
0.50
0.50
0.25
AUC=0.8
0.25
0.25
AUC=0.7
AUC=0.7
0.00
0.00
0.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
C
Specificity
TCGA-P
TCGA-S
Parameter
P value
Hazard ratio (95% CI)
Parameter
P value
Hazard ratio (95% CI)
Riskscore
-
1.4 (1.2-1.7)
Riskscore
**
1.5 (1.1-1.9)
Age
NS
1 (0.98-1.1)
Age
*
1 (1-1.1)
SexMale
NS
1.9 (0.55-6.2)
Primary_therapy_outcome
5.9 (2.7-13)
SexMale
NS
0.78 (0.19-3.1)
Primary_therapy_outcome
4.4 (2.2-8.7)
0
5
0
5
Low risk
High Risk
Low risk
High Risk
showing a significant prognostication with a marginal significance in the HS cohort and a numeric difference in the TCGA-S cohort (Fig. S4).
Furthermore, due to the better performance of the five-genera risk score in pre- dicting the overall survival, we took advantage of the genomics and transcriptomics data of TCGA to characterize the potential correlation between signaling pathways and microbial community. However, the landscape of the genomic alteration showed rarely discrepant genomic events between the high-risk-score group and the low one for both the TCGA-P (Fig. S5A) and TCGA-S (Fig. 4A) cohorts. Moreover, there was no difference in the TMBs (Fig. 4B; Fig. S5B) and CNVs (Fig. 4C; Fig. S5C) between the two subgroups.
We then explored the differentially expressed genes between the subgroups and found that these genes tend to function in pathways related to metabolism, such as glycerophospholipid metabolism, choline metabolism in cancer, arginine and proline metabolism, and retinol metabolism (Fig. 4D). Also, the gene set enrichment analy- sis showed that the AMPK signaling pathway, propanoate metabolism, and carbon metabolism were significantly discriminated between subgroups determined by the five-genera risk score, suggesting the metabolic orchestration performed by some intratumor microbiomes (Fig. 5A; Fig. S6A). Of note, we did not observe the difference in the infiltration of the immune cells between the subgroups, implying the tendency of metabolic participation rather than the immune response of the five-genera community in the tumor microenvironment (Fig. 5B; Fig. S6B).
A
CTNNB1
D
TP53
Neuroactive ligand-receptor interaction
MENT
PAKARTA
APL22
Cytokine-cytokine receptor interaction
1tp15.5
11914.1
12q14.1
Proteoglycans in cancer
13914.2
14qf1.2
16p13.3
Lipid and atherosclerosis
16422.1
17q11.2
CAMP signaling pathway
17g21.31
17424.2
17q25.3
Phospholipase D signaling pathway
-log10(pvalue)
18p13.12
19q12
3.5
1036 23
Sphingolipid signaling pathway
20p12.1
3.0
22q12.1
Glycerophospholipid metabolism
2q37,3
2.5
3q13.31
4p16.3
Axon guidance
4935.1
2.0
5015.33
5935.3
Choline metabolism in cancer
6026
9p21.3
Bp23
PPAR signaling pathway
Count
Xạ2a
☒
2.5
Chemical carcinogenesis - DNA adducts
☒ 5.0
Splice_Site
In_Frame_Del
☒
Missense_Mutation
TCGA_S
Arginine and proline metabolism
7.5
Amp
Frame_Shift_Del
Del
High
☒
Frame_Shift_Ins
Multi_Hit
Low
Steroid hormone biosynthesis
10.0
Nonsense_Mutation
Retinol metabolism
☒
12.5
3
Tumor mutation burden (total per MB)
p = 0.15
C
1.00
p = 0.66
Pertussis
Fraction genome altered (x 100%)
Drug metabolism - cytochrome P450
6
Aldosterone-regulated sodium reabsorption
0.75
African trypanosomiasis
TCGA_S
TCGA_S
Glycosaminoglycan biosynthesis - keratan sulfate
4
0.50
TCGA_P TCGA_S
2
0.25
0
0.00
High
Low
High
Low
DISCUSSION
We reported here a five-genera ITB risk score in our own cohort and validated our findings in the TCGA data set processed by the two top groups (1, 9), where all cohorts differ in sequencing technique, patient race, and processing algorithm. Recent debate over the existence and effect of intratumor microbes has drawn much attention. Besides the TCGA microbiome project, another landmark study showing intratumor fungi playing a pathogenic role in pancreatic cancer (25) was challenged that such mycobiome was solely a bystander (26). To sum up, key questions regarding ITBs in cancer include following: (i) is there such a diversity of ITBs in low-biomass cancers? If so, (ii) can the diversity by used as biomarkers (diagnostic, prognostic, and/or response)? If so, (III) do certain ITBs play a causal role in cancer? Our previous study on ACC showed ITB composition could not only differentiate it from other cancers but also subtype the disease according to survival (10). Though such notion was counter-intuitive for the adrenal gland, which is commonly accepted as a sterile organ, though virus was detected therein long ago (27). In order to present such clinical association, microbial data should be quantitatively abundant to generate statistical significance. Whereas microbial decontamination was accepted widely, Salzberg et al. (9) argued contamina- tion of host genetic misclassified as microbial reads could be a problem resulting from high abundance of ITBs in the study by Poore et al. (1) and has provided stringently
A
NES: - 1.54603790519503
NES: - 1.8728414995568
NES: - 1.87699470544893
Running Enrichment Score
pvalue
p.adju
Running Enrichment Score
0.0
pvalue
p.adjust
0.0
AMPK signaling pathway
0.003846
0.0383
Carbon metabolism
Running Enrichment Score
0.0
p.adjus
-0.1
0.001965
0.0304
pvalue
Propanoate metabolism
0.004228
0.0383
-0.1
-0.2.
-0.2
0.2
0.3
-0.4
-0.3
-0.4
-0.4
-0.6
Ranked List Metric
2.
Ranked List Metric
2
Ranked List Metric
2
1
1
1
0
0
0
-1
-1
-1
-2
-2
-2
5000
10000
Rank in Ordered Dataset
15000
5000
10000
Rank in Ordered Dataset
15000
5000
10000
Rank in Ordered Dataset
15000
B
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
0.75
ssGSEA score
0.50
0.25
..
…
:
0.00
B cells
Dendritic cells
Macrophage M1
Macrophage M2
Monocytes
Neutrophils
NK cells
Other infiltrates
CD4+ T cells
CD8+ T cells
Tregs
decontaminated data. Our findings, validated in both cohorts, support the notion that certain ITB features, namely, Corynebacterium, Mycoplasma, Achromobacter, Anaerococ- cus, and Streptococcus, rather than microbial diversity on the whole, may play patho- genic roles in ACC. However, features associated with PFS should be intepreted with caveat. Limited sample size in our cohort, lack of consistent statistical significance, and differences in definition of progression all render our investigation of PFS solely exploratory and our results heterogenous.
Corynebacterium has been recognized as an important pathogen in immunosuppres- sive patients, and many non-diphtheria corynebacterial can also be virulent in hospital- ized patients. Of note, Corynebacterium has become a common infected pathogen in cancer patients in Brazil due to indwelling instruments like catheters and intravenous tubes (28). This corresponds to the notion that most ITBs migrate to tumors via the bloodstream. Mycoplasma on the other hand, has been reported to play a critical role in various malignant tumors (29-31). In fact, the whole ITB research community was mostly, if not entirely inspired by the interesting report on mycoplasma modulating chemoresistance in solid tumor (32). Similar to Corynebacterium infection, patients with cancer are also at risk of Achromobacter infection due to immunosuppression and the use of prophylaxis with fluoroquinolones (33). Anaerococcus and Streptococcus were both often reported in cancer microbiome studies, particularly in bladder cancer (34).
However, presence and association do not mean causation (35). Though a robust association between those ITBs and prognosis is shown here in ACC, the causal relation between ITBs and adrenal tumorigenesis remain unknown. Whether ITBs play
a commensal or driving role alongside tumor progression depends on human micro- biota-associated murine models and microbe-phenotype triangulation (35). Fortunately, a transgenic murine model for ACC has just been reported this year (36), and culturomics is therefore applicable in the future. Alternatively, one may apply patient-derived ITBs for further functional analyses. However, the rarity of the disease may substantially hinder the process, and multicenter collaboration is needed. Once again limited by the rare nature of the disease, we strived to collect ACC cases with relatively complete clinicopathological parameters. The long time span for sample collection rendered many of the patients lost to follow-ups, and thus, only 26 cases were eligible to the current study. Though we managed to follow up more ACC patients that dated further back, their FFPE blocks decayed substantially and were not applicable for 16S sequencing. Another inherent limitation is that ACC samples are extremely large and invasive that no “adjacent” normal adrenal tissue was available. This is a universal problem as the TCGA data also harbor no paired normal tissue. In our previous analysis, we compared data from ACC with those of paraganglioma/pheochromocytoma, given the consideration of similar organ origin and surgical procedure. Though Cantini et al. showed deferential ITBs between unpaired ACC and healthy adrenal tissue (11), we did not adopt such approach as intratissue microbiome vary drastically and interpersonal diversity could even be magnified. Also, we did not measure ITB loads in ACC, as our primary goal was to identify common ITB features associated with prognosis, and loading information could not be profiled in TGGA cohorts. However, as recent studies point out that absolute, rather than relative, abundance plays a more important role in microbiome study (37) and load is a prognostic in nasopharyngeal cancer (38), we are now setting up a new line to evaluate the association between ITB loads and prognosis.
Using the five genera associated with OS, we were able to perform a functional analysis of those features. We found that, unlike in renal cancer in our another project (39), survival-associated ITBs in ACC were not related to immunity but to metabolism. This could result in a different bacterial community between the diseases. Amid the three metabolic pathways, propanoate metabolism is of interest. Propionate is observed to be among the most common short-chain fatty acids produced in the large intestine of humans by gut microbiota in response to indigestible carbohydrates (dietary fiber) in the diet (40, 41). Though it is known to suppress bacterial immune response, its metabolism activation in the presence of those OS-related genera highly suggested that propanoate is catalyzed or metabolized by the bacteria. Certain species of Corynebacterium (42), Mycoplasma (43), Achromobactin (44), Anaerococcus (45), and Streptococcus (46) were reported to either metabolize or generate propanoate. How propanoate metabolism of the genera impacts on ACC warrants further functional study.
Lastly, we did not put much effort in imaging ITBs in the current study. 16S FISH staining and bacterial LPS staining of ACC samples were shown in our previous report (10) using samples that overlap with the HS cohort. For low-biomass cancer, we tend to consider both LPS and FISH staining could harbor magnified signals from extra-tumor bacterial contamination, a notion supported by a recent report (47). Our findings, together with validation in the TCGA-P/TCGA-S cohorts, has undoubtfully proven the existence and prognostication of ITBs in ACC.
Conclusion
Whereas ITB signature on the whole may not be associated with ACC subtypes, certain ITB features are associated with prognosis, and a risk score could be generated and validated externally.
ACKNOWLEDGMENTS
We owe great thanks to the Salzberg team for sharing their data and granting us permission to use these for publication.
This study was sponsored in part by the National Natural Science Foundation of China (grant number 81874123).
Conceptualization: C.F. and N.L .; methodology: C.F., L.T., Y.L., D.Z., and Y.x.L .; validation: L.T., Y.L., and D.Z .; investigation: C.F., L.T., Y.L., and D.Z .; original draft: N.L .; revision: C.F.
AUTHOR AFFILIATIONS
1Department of Urology, Huashan Hospital, Fudan University, Shanghai, China
2Department of Chemistry and The Swire Institute of Marine Science, The University of Hong Kong, Hong Kong, China
3Department of Gynecology, Obstetrics and Gynecology Hospital of Fudan University, Shanghai, China
4Cancer Institute, Xuzhou Medical University, Xuzhou, Jiangsu, China
5Department of Urology, Fourth Affiliated Hospital of China Medical University, Shenyang, Liaoning, China
AUTHOR ORCIDS
Ning Li @ http://orcid.org/0000-0001-6814-7947 Chenchen Feng @D http://orcid.org/0000-0002-1854-356X
FUNDING
| Funder | Grant(s) | Author(s) |
|---|---|---|
| MOST | National Natural Science Foundation of China (NSFC) | 81874123 | Chenchen Feng |
AUTHOR CONTRIBUTIONS
Linyi Tan, Investigation, Methodology, Writing - original draft | Dengwei Zhang, Investigation, Methodology | Yong-xin Li, Investigation | Yuqing Li, Conceptualization, Methodology | Ting Guo, Conceptualization, Investigation | Yang Sun, Conceptualization, Investigation | Ning Li, Conceptualization | Chenchen Feng, Conceptualization, Method- ology
DATA AVAILABILITY
Read counts of un-decontaminated intratumor bacteria were deposited at China National Center for Bioinformation (CRA015044). Codes for reproduction of The Cancer Genome Atlas (TCGA) data processed by Poore et al. (NR cluster) are linked to the GitHub release (https://github.com/ZhangDengwei/ACC_Project). Request for TCGA-adrenocort- ical carcinoma data processed by Salzberg et al. should be addressed to Professor Steven Salzberg (steven.salzberg@gmail.com). Clinicopathological data may be provided upon request to the corresponding author (C.F.).
ETHICS APPROVAL
Informed consent was obtained for all patients. The study was approved by Huashan Institutional Review Board and China Medical University.
ADDITIONAL FILES
The following material is available online.
Supplemental Material
Supplemental figures (Spectrum03727-23-s0001.docx). Figures S1 to S6.
REFERENCES
1. Poore GD, Kopylova E, Zhu Q, Carpenter C, Fraraccio S, Wandro S, Kosciolek T, Janssen S, Metcalf J, Song SJ, Kanbar J, Miller-Montgomery S, Heaton R, Mckay R, Patel SP, Swafford AD, Knight R. 2020. Microbiome analyses of blood and tissues suggest cancer diagnostic approach. Nature 579:567-574. https://doi.org/10.1038/s41586-020-2095-1
2. Oosterlinck B, Ceuleers H, Arras W, De Man JG, Geboes K, De Schepper H, Peeters M, Lebeer S, Skieceviciene J, Hold GL, Kupcinskas J, Link A, De Winter BY, Smet A. 2023. Mucin-microbiome signatures shape the tumor microenvironment in gastric cancer. Microbiome 11:86. https://doi.org/ 10.1186/s40168-023-01534-w
3. Mouradov D, Greenfield P, Li S, In E-J, Storey C, Sakthianandeswaren A, Georgeson P, Buchanan DD, Ward RL, Hawkins NJ, Skinner I, Jones IT, Gibbs P, Ma C, Liew YJ, Fung KYC, Sieber OM. 2023. Oncomicrobial community profiling identifies clinicomolecular and prognostic subtypes of colorectal cancer. Gastroenterology 165:104-120. https:// doi.org/10.1053/j.gastro.2023.03.205
4. Sun L, Ke X, Guan A, Jin B, Qu J, Wang Y, Xu X, Li C, Sun H, Xu H, Xu G, Sang X, Feng Y, Sun Y, Yang H, Mao Y. 2023. Intratumoural microbiome can predict the prognosis of hepatocellular carcinoma after surgery. Clin Transl Med 13:e1331. https://doi.org/10.1002/ctm2.1331
5. Sepich - Poore GD, Guccione C, Laplane L, Pradeu T, Curtius K, Knight R. 2022. Cancer’s second genome: microbial cancer diagnostics and redefining clonal evolution as a multispecies process. BioEssays 44. https://doi.org/10.1002/bies.202100252
6. Nejman D, Livyatan I, Fuks G, Gavert N, Zwang Y, Geller LT, Rotter- Maskowitz A, Weiser R, Mallel G, Gigi E, et al. 2020. The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science 368:973-980. https://doi.org/10.1126/science.aay9189
7. Fu A, Yao B, Dong T, Chen Y, Yao J, Liu Y, Li H, Bai H, Liu X, Zhang Y, Wang C, Guo Y, Li N, Cai S. 2022. Tumor-resident intracellular microbiota promotes metastatic colonization in breast cancer. Cell 185:1356-1372. https://doi.org/10.1016/j.cell.2022.02.027
8. Dohlman AB, Arguijo Mendoza D, Ding S, Gao M, Dressman H, Iliev ID, Lipkin SM, Shen X. 2021. The cancer microbiome atlas: a pan-cancer comparative analysis to distinguish tissue-resident microbiota from contaminants. Cell Host & Microbe 29:281-298. https://doi.org/10.1016/ j.chom.2020.12.001
9. Gihawi A, Ge Y, Lu J, Puiu D, Xu A, Cooper CS, Brewer DS, Pertea M, Salzberg SL. 2023. Major data analysis errors invalidate cancer microbiome findings. mBio 14:e0160723. https://doi.org/10.1128/mbio. 01607-23
10. Li Y, Zhang D, Wang M, Jiang H, Feng C, Li YX. 2023. Intratumoral microbiota is associated with prognosis in patients with adrenocortical carcinoma. iMeta 2. https://doi.org/10.1002/imt2.102
11. Cantini G, Niccolai E, Canu L, Di Gloria L, Baldi S, Propato AP, Fei L, Nannini G, Puglisi S, Nesi G, Ramazzotti M, Amedei A, Luconi M. 2023. Intratumour microbiota modulates adrenocortical cancer responsive- ness to mitotane. Endocr Relat Cancer 30:e230094. https://doi.org/10. 1530/ERC-23-0094
12. Luo M, Liu Y, Hermida LC, Gertz EM, Zhang Z, Li Q, Diao L, Ruppin E, Han L. 2022. Race is a key determinant of the human intratumor microbiome. Cancer Cell 40:901-902. https://doi.org/10.1016/j.ccell.2022.08.007
13. Magoč T, Salzberg SL. 2011. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27:2957-2963. https://doi.org/10.1093/bioinformatics/btr507
14. Bokulich NA, Subramanian S, Faith JJ, Gevers D, Gordon JI, Knight R, Mills DA, Caporaso JG. 2013. Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nat Methods 10:57-59. https://doi.org/10.1038/nmeth.2276
15. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. 2011. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics 27:2194-2200. https://doi.org/10.1093/bioinformatics/btr381
16. Haas BJ, Gevers D, Earl AM, Feldgarden M, Ward DV, Giannoukos G, Ciulla D, Tabbaa D, Highlander SK, Sodergren E, Methé B, DeSantis TZ, Human Microbiome Consortium, Petrosino JF, Knight R, Birren BW. 2011. Chimeric 16S rRNA sequence formation and detection in Sanger and 454-pyrosequenced PCR amplicons. Genome Res. 21:494-504. https:// doi.org/10.1101/gr.112730.110
17. Yao B, Dong T, Fu A, Li H, Jiang C, Li N, Cai S. 2022. Quantification and characterization of mouse and human tissue-resident microbiota by qPCR and 16S sequencing. STAR Protoc 3:101765. https://doi.org/10. 1016/j.xpro.2022.101765
18. Wheeler C, Yang Y, Spakowicz D, Hoyd R, Li M. 2021. 942 the tumor microbiome correlates with response to immune checkpoint inhibitors in renal cell carcinoma. J Immunother Cancer 9:A988-A989. https://doi. org/10.1136/jitc-2021-SITC2021.942
19. Eisenhofer R, Minich JJ, Marotz C, Cooper A, Knight R, Weyrich LS. 2019. Contamination in low microbial biomass microbiome studies: issues and recommendations. Trends Microbiol. 27:105-117. https://doi.org/10. 1016/j.tim.2018.11.003
20. Salter SJ, Cox MJ, Turek EM, Calus ST, Cookson WO, Moffatt MF, Turner P, Parkhill J, Loman NJ, Walker AW. 2014. Reagent and laboratory contamination can critically impact sequence-based microbiome analyses. BMC Biol. 12:87. https://doi.org/10.1186/s12915-014-0087-z
21. Colaprico A, Silva TC, Olsen C, Garofano L, Cava C, Garolini D, Sabedot TS, Malta TM, Pagnotta SM, Castiglioni I, Ceccarelli M, Bontempi G, Noushmehr H. 2016. TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res. 44:e71-e71. https:// doi.org/10.1093/nar/gkv1507
22. Mayakonda A, Lin D-C, Assenov Y, Plass C, Koeffler HP. 2018. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 28:1747-1756. https://doi.org/10.1101/gr.239244.118
23. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. 2015. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43:e47-e47. https://doi.org/10. 1093/nar/gkv007
24. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, Fu X, Liu S, Bo X, Yu G. 2021. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2:100141. https:// doi.org/10.1016/j.xinn.2021.100141
25. Aykut B, Pushalkar S, Chen R, Li Q, Abengozar R, Kim JI, Shadaloey SA, Wu D, Preiss P, Verma N, Guo Y, Saxena A, Vardhan M, Diskin B, Wang W, Leinwand J, Kurz E, Kochen Rossi JA, Hundeyin M, Zambrinis C, Li X, Saxena D, Miller G. 2019. The fungal mycobiome promotes pancreatic oncogenesis via activation of MBL. Nature 574:264-267. https://doi.org/ 10.1038/s41586-019-1608-2
26. Fletcher AA, Kelly MS, Eckhoff AM, Allen PJ. 2023. Revisiting the intrinsic mycobiome in pancreatic cancer. Nature 620:E1-E6. https://doi.org/10. 1038/s41586-023-06292-1
27. Flaegstad T, Andresen PA, Johnsen JI, Asomani SK, Jørgensen GE, Vignarajan S, Kjuul A, Kogner P, Traavik T. 1999. A possible contributory role of BK virus infection in neuroblastoma development. Cancer Res 59:1160-1163.
28. Martins C, Faria L, Souza M, Camello T, Velasco E, Hirata R Jr, Thuler L, Mattos-Guaraldi A. 2009. Microbiological and host features associated with corynebacteriosis in cancer patients: a five-year study. Mem Inst Oswaldo Cruz
104:905-913.
https://doi.org/10.1590/s0074 — 02762009000600015
29. Ainsworth JG, Easterbrook PJ, Clarke J, Gilroy CB, Taylor-Robinson D. 2001. An association of disseminated Mycoplasma fermentans in HIV-1 positive patients with non-Hodgkin’s lymphoma. Int J STD AIDS 12:499- 504. https://doi.org/10.1258/0956462011923589
30. Barykova YA, Logunov DY, Shmarov MM, Vinarov AZ, Fiev DN, Vinarova NA, Rakovskaya IV, Baker PS, Shyshynova I, Stephenson AJ, Klein EA, Naroditsky BS, Gintsburg AL, Gudkov AV. 2011. Association of Mycoplasma hominis infection with prostate cancer. Oncotarget 2:289- 297. https://doi.org/10.18632/oncotarget.256
31. Henrich B, Rumming M, Sczyrba A, Velleuer E, Dietrich R, Gerlach W, Gombert M, Rahn S, Stoye J, Borkhardt A, Fischer U. 2014. Mycoplasma salivarium as a dominant coloniser of Fanconi anaemia associated oral carcinoma. PLoS One 9:e92297. https://doi.org/10.1371/journal.pone. 0092297
32. Vande Voorde J, Sabuncuoğlu S, Noppen S, Hofer A, Ranjbarian F, Fieuws S, Balzarini J, Liekens S. 2014. Nucleoside-catabolizing enzymes in mycoplasma-infected tumor cell cultures compromise the cytostatic activity of the anticancer drug gemcitabine. J Biol Chem 289:13054- 13065. https://doi.org/10.1074/jbc.M114.558924
33. Stutzman T, Sánchez-Vargas FM, Nanjappa S, Velez AP, Greene JN. 2016. Achromobacter bacteremia in patients with cancer. Infect Dis Clin Pract 24:339-342. https://doi.org/10.1097/IPC.0000000000000430
34. D’Antonio DL, Marchetti S, Pignatelli P, Piattelli A, Curia MC. 2022. The oncobiome in gastroenteric and genitourinary cancers. Int J Mol Sci 23:9664. https://doi.org/10.3390/ijms23179664
35. Lv B-M, Quan Y, Zhang H-Y. 2021. Causal inference in microbiome medicine: principles and applications. Trends Microbiol. 29:736-746. https://doi.org/10.1016/j.tim.2021.03.015
36. Warde KM, Smith LJ, Liu L, Stubben CJ, Lohman BK, Willett PW, Ammer JL, Castaneda-Hernandez G, Imodoye SO, Zhang C, Jones KD, Converso- Baran K, Ekiz HA, Barry M, Clay MR, Kiseljak-Vassiliades K, Giordano TJ, Hammer GD, Basham KJ. 2023. Senescence-induced immune remodel- ing facilitates metastatic adrenal cancer in a sex-dimorphic manner. Nat Aging 3:846-865. https://doi.org/10.1038/s43587-023-00420-2
37. Maghini DG, Dvorak M, Dahlen A, Roos M, Kuersten S, Bhatt AS. 2024. Quantifying bias introduced by sample collection in relative and absolute microbiome measurements. Nat Biotechnol 42:328-338. https:/ /doi.org/10.1038/s41587-023-01754-3
38. Qiao H, Tan X-R, Li H, Li J-Y, Chen X-Z, Li Y-Q, Li W-F, Tang L-L, Zhou G-Q, Zhang Y, Liang Y-L, He Q-M, Zhao Y, Huang S-Y, Gong S, Li Q, Ye M-L, Chen K-L, Sun Y, Ma J, Liu N. 2022. Association of intratumoral microbiota with prognosis in patients with nasopharyngeal carcinoma from 2 hospitals in China. JAMA Oncol 8:1301-1309. https://doi.org/10. 1001/jamaoncol.2022.2810
39. Li Y, Zhang D, Tan L, Xu J, Guo T, Sun Y, Zhang R, Cheng Y, Jiang H, Zhai W, Li Y, Feng C. 2023 Intratumor bacteria is associated with prognosis in clear-cell renal cell carcinoma. medRxiv. https://doi.org/10.1101/2023. 12.29.23300629
40. Cani PD, Knauf C. 2016. How gut microbes talk to organs: the role of endocrine and nervous routes. Mol Metab 5:743-752. https://doi.org/10. 1016/j.molmet.2016.05.011
41. den Besten G, van Eunen K, Groen AK, Venema K, Reijngoud D-J, Bakker BM. 2013. The role of short-chain fatty acids in the interplay between diet, gut microbiota, and host energy metabolism. J Lipid Res 54:2325- 2340. https://doi.org/10.1194/jlr.R036012
42. Veit A, Rittmann D, Georgi T, Youn J-W, Eikmanns BJ, Wendisch VF. 2009. Pathway identification combining metabolic flux and functional genomics analyses: acetate and propionate activation by Corynebacte- rium glutamicum. J Biotechnol 140:75-83. https://doi.org/10.1016/j. jbiotec.2008.12.014
43. Masukagami Y, De Souza DP, Dayalan S, Bowen C, O’Callaghan S, Kouremenos K, Nijagal B, Tull D, Tivendale KA, Markham PF, McConville MJ, Browning GF, Sansom FM. 2017. Comparative metabolomics of Mycoplasma bovis and Mycoplasma gallisepticum reveals fundamental differences in active metabolic pathways and suggests novel gene annotations. mSystems 2:00055-17. https://doi.org/10.1128/mSystems. 00055-17
44. Liu Y-J, Pei X-Q, Lin H, Gai P, Liu Y-C, Wu Z-L. 2012. Asymmetric bioreduction of activated alkenes by a novel isolate of Achromobacter species producing enoate reductase. Appl Microbiol Biotechnol 95:635- 645. https://doi.org/10.1007/s00253-012-4064-6
45. Ezaki T, Ohkusu K. 2015. Anaerococcus, p 1-5. In Bergey’s manual of systematics of archaea and bacteria.
46. Wampler JL, Martin SA, Hill GM. 1998. Effects of laidlomycin propionate and monensin on glucose utilization and nutrient transport by Streptococcus bovis and Selenomonas ruminantium. J Anim Sci 76:2730- 2736. https://doi.org/10.2527/1998.76102730x
47. de Miranda NF, Smit VT, van der Ploeg M, Wesseling J, Neefjes J. 2023. Absence of lipopolysccharide (LPS) expression in breast cancer cells. bioRxiv. https://doi.org/10.1101/2023.08.28.555057