The two scripts largely use the same MicrobiotaProcess functions; the main difference is *which `ps_` object you feed in for each task** (alpha, beta, composition plots).

What the “small” script currently does

• Creates one MPSE (often called mpse_abund) from a taxa-filtered plotting object (e.g. ps.ng.tax_abund / your ps_abund) and then runs alpha diversity + beta diversity + composition plotting all on that same MPSE.
• It also uses rarefied abundance (RareAbundance) as the input for some taxonomy abundance plots/heatmaps.

Implication: alpha/beta diversity are being computed on a taxa-filtered dataset, which can:

• artificially reduce richness (Observed/Chao1),
• shift Shannon/Simpson,
• and change distance structure (especially for presence/absence metrics, but also sometimes Bray).

What the updated MicrobiotaProcess workflow (recommended) does

It splits the workflow into two MPSE objects, each built from the correct upstream ps_*:

1) Diversity MPSE (mpse_div)

Input: ps_filt (QC-filtered samples, full taxa set; not filtered “for plotting”)

• Alpha diversity: do rarefaction inside MPSE (mp_rrarefy()) and compute alpha on RareAbundance.
• Beta diversity (your current Bray+Hellinger): compute Hellinger from non-rarefied Abundance and then Bray/PCoA/PERMANOVA.

2) Plotting MPSE (mpse_plot)

Input: ps_abund_rel (taxa filtered for readability; relative abundance)

• Use this MPSE only for composition plots (stacked bars, heatmaps).

Implication: you keep diversity analyses biologically faithful (no “plotting filter”) while still producing clean, readable composition plots.

Other (non-critical) differences

• The updated script uses prune_samples() + prune_taxa() instead of overwriting otu_table() manually (safer / less error-prone).
• Outputs are consistently written into a figures/ directory.

Bottom line

Same MicrobiotaProcess functions; the updated script mainly fixes the *recommended `ps_` input choice** for each analysis type (diversity vs. plotting).

!!!!! Good candidate for the workshop for the clinicians !!!!!

“First, I generate ps_rarefied from ps_filt. Then, for cleaner composition plots, I create ps_abund / ps_abund_rel by filtering taxa for plotting (e.g., keeping taxa with mean relative abundance > 0.1%; resulting in ~95 taxa across 239 samples). Is it correct to compute alpha diversity and beta diversity using ps_abund (the taxa-filtered object)?”

Not really. ps_abund / ps_abund_rel (taxa filtered “for plotting”) is generally not the right object for alpha- or beta-diversity, unless you explicitly want diversity calculated only within that filtered subset. Luckily, I directly using alpha- and beta diversity from qiime2-result and MicrobiotaProcess –> which means we should feed MicrobiotaProcess with ps_filt (not rarefied) and for differential analysis also using ps_filt (not rarefied)

For that sentence, I mean these specific objects:

• For Jaccard / unweighted UniFrac (presence–absence / detection-sensitive):
  ✅ compute beta diversity on ps_rarefied
  = the rarefied-count phyloseq object created from ps_filt (with the full taxon set, i.e., no plotting-only taxon filter).

• For Bray–Curtis (abundance-based):
  ✅ compute beta diversity on a non-rarefied transformed object, typically:

  • ps_rel = relative abundances computed from ps_filt (not rarefied), or
  • a Hellinger-transformed version derived from ps_filt (if that is your chosen workflow).

• For CLR / Aitchison distance:
  ✅ compute beta diversity on a CLR-transformed version of the non-rarefied counts from ps_filt
  (often named ps_clr / otu_clr, created from ps_filt + pseudocount), not on ps_abund.

The “plotting-filtered object” I’m warning against is:

• ps_abund / ps_abund_rel = taxa filtered for visualization (e.g., mean relative abundance > 0.1%).

Why it’s not “correct” for diversity

Filtering taxa by mean relative abundance (e.g., >0.1%) removes many low-abundance / low-prevalence taxa and changes the community profile.

Alpha diversity

• Observed richness will be artificially lower (because taxa were removed).
• Shannon/Simpson can also change because you altered the abundance distribution. ✅ Therefore, compute alpha diversity on ps_rarefied (unfiltered taxa), not on ps_abund.

Beta diversity

This depends on the distance metric, but heavy taxon filtering can distort results:

• Presence/absence distances (e.g., Jaccard, unweighted UniFrac) are very sensitive to removing taxa → filtering can strongly change distances and clustering.
• Abundance-weighted distances (e.g., Bray–Curtis, weighted UniFrac) can also shift after filtering; mild filters (e.g., remove singletons) may be acceptable, but mean rel. abundance >0.1% can be quite aggressive (95 taxa suggests strong pruning). ✅ Best practice: compute beta diversity on ps_rarefied (for Jaccard/unweighted UniFrac) or on non-rarefied transformed data (for Bray/CLR workflows), not on a plotting-filtered object.

Recommended workflow (clean separation of purposes)

For analysis (diversity)

• Alpha diversity: ps_rarefied (derived from ps_filt)
• Beta diversity:
  • Jaccard / unweighted UniFrac: ps_rarefied
  • Bray–Curtis: usually from ps_rel (or Hellinger), often without rarefaction
  • Aitchison/CLR: CLR-transformed (non-rarefied) data (no rarefaction)

For visualization (composition plots only)

• Optional cleaner composition plots: ps_abund / ps_abund_rel derived from ps_rel (taxa filtered to make stacked bars/heatmaps readable)

When would ps_abund be acceptable for beta diversity?

Only if your question is explicitly:

“How do samples differ considering only dominant taxa (≥0.1% mean abundance)?”

Then it’s a valid different analysis, but you should label it clearly as “dominant-taxa-only beta diversity.”

If you tell me which beta distance(s) you use (Bray, Jaccard, weighted/unweighted UniFrac, CLR/Aitchison), I can recommend the best object + transformation for your exact setup.

The correct answer is: beta diversity does not universally require rarefaction.

Why this confusion happens

Historically, many microbiome pipelines used rarefied counts for beta diversity to “standardize depth.” That’s why you often see “beta diversity → rarefy” in older workflows. But in modern practice, rarefaction for beta diversity is optional and sometimes not recommended, because it discards data and adds randomness.

When rarefaction is a good idea for beta diversity

Use ps_rarefied when your beta metric is very sensitive to library size, especially:

• Presence/absence distances: Jaccard, unweighted UniFrac These are strongly influenced by detection of rare taxa, which depends on depth. Rarefaction is a common way to make them comparable.
• If your field/reviewer expectations specifically require rarefied beta diversity for certain metrics.

When rarefaction is not necessary (often better avoided)

Prefer non-rarefied approaches when you use:

• Bray–Curtis on relative abundance (or Hellinger-transformed data) Rarefaction can reduce power and add noise; relative abundance already removes the “sum” effect (though compositionality remains).
• Aitchison distance / PCA via CLR transform (Euclidean on CLR) This is a standard compositional approach and typically does not use rarefaction.
• DESeq2 VST / rlog for ordination-like PCA (count-model-based normalization) Again: no rarefaction.

Important note about your original code

Doing rarefaction → compositional (“relative abundance”) is usually redundant for beta diversity/ordination, because after rarefaction every sample has the same total, and compositional transformation just divides by that constant.

Corrected “rule of thumb”

• Alpha diversity: ps_rarefied ✅ (common)
• Beta diversity:
  • Unweighted UniFrac / Jaccard: ps_rarefied ✅ (often recommended)
  • Bray–Curtis / ordination on abundances: ps_rel or Hellinger ✅ (rarefaction optional)
  • Aitchison (CLR): CLR-transformed (non-rarefied) ✅ (no rarefaction)

If you tell me exactly which beta-diversity distance(s) you compute (Bray–Curtis? weighted/unweighted UniFrac? Jaccard? Aitchison/CLR?), I can recommend the cleanest “use this object + this transformation” setup.

Doing rarefaction and then compositional normalization (relative abundance) is often redundant.

Short answer: Don’t need both — rarefaction + compositional normalization is redundant in most cases. Doing rarefaction and then compositional normalization (relative abundance) is often redundant.

# RAREFACTION
set.seed(9242)  # This will help in reproducing the filtering and nomalisation.
ps.ng.tax <- rarefy_even_depth(ps.ng.tax, sample.size = 6389)
total <- 6389

# NORMALIZE number of reads in each sample using median sequencing depth.
total = median(sample_sums(ps.ng.tax))
#> total
#[1] 42369
standf = function(x, t=total) round(t * (x / sum(x)))
ps.ng.tax = transform_sample_counts(ps.ng.tax, standf)
ps.ng.tax_rel <- microbiome::transform(ps.ng.tax, "compositional")

saveRDS(ps.ng.tax, "./ps.ng.tax.rds")

What happens mathematically

1. Rarefaction forces every sample to the same library size (e.g., 6,389 reads).
2. Compositional normalization divides each sample by its total so it sums to 1.

After rarefaction, every sample has the same total, so compositional normalization is essentially: relative_abundance = rarefied_counts / constant_depth. That step is valid, but typically unnecessary.

Practical implication

• If your goal is relative abundance–based visualization or many beta-diversity analyses, you can usually skip rarefaction and compute relative abundances (or other transformations) on the full data.
• If your goal is alpha diversity, rarefaction is commonly used, but then you typically keep that rarefied object for alpha diversity only (not as a general-purpose normalized dataset).

So: rarefaction alone can be fine for specific tasks (especially alpha diversity), but rarefaction + compositional normalization together is usually not needed.

Mapping new ps_* objects to the old ps.ng.tax* objects (and overwrite notes)

Overwrite summary (old script):

• ps.ng.tax was overwritten 1 time:
  • Before overwrite: ps.ng.tax = absolute abundance (raw integer counts).
  • After overwrite: ps.ng.tax = rarefied counts, produced by rarefy_even_depth(ps.ng.tax, sample.size = 41764).

这里说“证据链没有闭环”，不是否定结果，而是从审稿人最严格的因果标准来看，ΔadeIJ 这条线目前更像“相关性很强 + 合理推断”，但还缺少几步能把“推断”锁死成“因果”的关键证据。

1) 目前已有的证据（强相关）

• 表型：ΔadeIJ 的 ROS 更高，对 Cu²⁺ / SNP 更敏感
• 转录组：ΔadeIJ 上调金属外排/解毒/应激相关基因

因此很自然会推断：

AdeIJ 参与金属/氧化应激稳态（redox & metal homeostasis）

这属于 “一致性证据（consistent with）”，说服力已经很不错。

2) 为什么说“还没闭环”：少了把“推断 → 因果”钉死的步骤

审稿人可能会问：
“你怎么证明是 缺失 AdeIJ 导致金属/ROS 失衡，而不是其他原因？”

通常闭环至少需要补上一类证据（不一定要做，但逻辑上缺）：

A. 互补实验（complementation / rescue）

理想闭环：

• ΔadeIJ 表型变差
• 把 adeIJ（或 adeIJK）补回去 → ROS/Cu²⁺/SNP 表型恢复到 WT（rescue）

没有这一步，审稿人可能会担心：

• 极性效应（polar effect）
• 二次突变（secondary mutation）
• 背景差异（background effects）

B. 直接测量“因果中间量”（mechanistic intermediate）

你们推断的是“金属/毒性底物积累 → ROS 上升 → Cu²⁺/SNP 敏感”。
但目前缺少对“中间量”的直接测量，例如：

• 细胞内 铜含量 是否在 ΔadeIJ 增加（ICP-MS/比色法等）？
• 是否有更明确的蛋白氧化损伤/Fe–S 破坏指标？
• 是否存在呼吸链/膜电位异常导致 ROS 增加？

目前是“结果（ROS 高）+ 反应（相关基因上调）”，但还没直接证明“金属积累/底物积累”这一环节。

C. 排除替代解释（alternative explanations）

ΔadeIJ 的高 ROS 也可能来自：

• 代谢状态改变导致呼吸链泄漏增加
• 生长状态差异影响 ROS 测定
• Cu²⁺ 敏感来自包膜改变而非金属外排不足

如果未排除，结论更适合写成 consistent with，而非 crucial determinant。

3) “闭环”长什么样（最理想的因果链）

AdeIJ 缺失
→ 细胞内金属/毒性底物积累（直接测到）
→ ROS 上升（你们已测到）
→ Cu²⁺/SNP 敏感（你们已测到）
→ 补回 adeIJ 后表型恢复（rescue）

这就是“证据链闭环”。

4) 对写作的建议（不一定要加实验）

投稿时完全可以不补实验，但建议在文字上更稳健：

• 避免：“AdeIJ is crucial/essential for maintaining …”
• 推荐：
  • “These data support / are consistent with a role for AdeIJ in …”
  • “We suggest AdeIJ contributes to …”
  • “Further work (e.g., complementation or intracellular metal quantification) would be needed to establish causality.”

1. 针对第 9、10、11 组以及混合的 pre-FMT 组的 PCA 和多样性分析
   筛选条件：Group 9（J1–4, J10, J11）、Group 10（K1–K6）、Group 11（L2–L6）以及 pre-FMT（G1–G6, H1–H6, I1–I6）
   1.1 这些组之间在 α 多样性或 β 多样性方面是否存在显著差异？你能否为此生成一个 PCoA 图？
   1.2 在仅比较第 9、10 和 11 组时，α 多样性或 β 多样性之间是否存在显著差异？你能否也为此生成一个 PCoA 图？
   1.3 第 9 组与第 10 组之间有哪些差异表达的基因（DEGs）？
   1.4 第 9 组与第 10 组之间是否存在差异富集/差异表达的通路？
   1.5 你是否可以使用 R 绘制一棵系统聚类树（树状图），包含数据集中检测到的所有细菌科（bacterial families）？我附了一篇论文作为示例（Figure 1C）。

2. Group 3 与 Group 4 之间的通路分析
   筛选条件：Group 3：C1–C7；Group 4：E1–E10。
   是否可以请你对 Group 3 vs Group 4 做一次通路分析？

3. 2022 年数据集（第一组筛选条件）
   筛选条件：Group 1：1, 2, 7, 6；Group 5：29, 30, 31。
   这些组之间在 α 多样性或 β 多样性方面是否存在显著差异？你能否为此生成一个 PCoA 图？
   有哪些 DEGs？
   是否存在差异富集/差异表达的通路？

4. 2022 年数据集（第二组筛选条件）
   筛选条件：Group 1：1, 2, 7, 6；Group 5：29, 30, 31；Group 2（不加筛选）；Group 6（不加筛选）。
   这些组之间在 α 多样性或 β 多样性方面是否存在显著差异？你能否为此生成一个 PCoA 图？

5. 微生物组分析的方法部分
   我们已经写好了材料和方法部分，其中包含微生物组分析的内容，我已将这一部分附在邮件中。你有没有什么意见？特别是你认为是否还需要补充描述 QC（质控）分析的内容？

   Microbiome analysis（微生物组分析）
   在所示时间点，于上午 9–10 点之间采集小鼠粪便样本，并立即在 -80°C 冻存直至使用。DNA 提取使用 QIAamp Fast DNA Stool Mini Kit（Qiagen, #51604），并根据已发表的小鼠粪便专用方案进行操作⁸。简而言之，粪便颗粒在 Precellys® 24 匀浆器（Bertin Technologies）中、使用 Soil grinding tubes（Bertin Technologies, #SK38）进行均质处理，随后按照试剂盒说明书提取 DNA。通过 Nanodrop 测定 DNA 产量，并将浓度调至 10 ng/µl。16S rRNA 扩增子（V3–V4 区域）采用带有 Illumina 接头通用序列的简并引物进行扩增：前向引物 F（5′- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′），反向引物 R（5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGG-TATCTAATCC-3′），具体方法参考已发表的方案⁹,¹⁰。随后使用 Illumina Nextera XT Index Kit（Illumina, #FC-131-1001）对样本进行多重上样，构建带条形码的文库。文库在 MiSeq 平台（Illumina, #MS-102-2003）上进行 500PE 测序。

   QC
   不同组之间微生物群落组成的差异使用 PERMANOVA（置换多元方差分析）进行评估，基于 R 包 vegan 中的 adonis 函数，设置 9999 次置换。该分析基于 Bray–Curtis 距离矩阵，该矩阵综合考虑了物种的有无以及各物种在每个样本中的相对丰度。该方法用于评估不同组之间整体群落组成是否存在显著差异。
   群体间的差异丰度分析使用 R 包 DESeq2 完成，该方法通过负二项式广义线性模型对计数数据进行建模¹¹。在气泡图中，x 轴表示相对丰度的 log₂ 倍数变化，每个气泡代表一个 OTU，气泡颜色表示其所属的细菌目（bacterial order），气泡大小表示经过 Benjamini–Hochberg 校正后的调整 p 值。

6. 关于 PCoA 和 PCA 的问题
   我们目前还不是完全确定你所做的分析是主坐标分析（PCoA）还是主成分分析（PCA）。你能否就此再给一些说明或评论？

1. 第 9、10、11 组及混合 pre-FMT 组的 PCA 和多样性分析
   你之前已经向我们展示过，这些组在 α 多样性和 β 多样性方面存在显著差异。我们能否据此得出结论：每一组在 α 和 β 多样性上都与所有其他组显著不同？如果不能，是否可以做一个事后检验（例如 Bonferroni 校正的 post-hoc 分析），来查看哪些组在 α 或 β 多样性上存在显著差异？在这里，我们主要感兴趣的比较是：9 vs 10，9 vs 11，10 vs 11。

2. 通路分析：第 9 组 vs 第 10 组
   是否可以预先筛选我们感兴趣的通路？在这里，我们特别想关注参与短链脂肪酸（SCFAs）生成的相关通路。这个在分析上可行吗？你对这样的做法有什么看法或建议？

3. 2022 年数据集：第 1、2、5 和 6 组
   在第 1 组和第 5 组之间，是否能检测到任何差异表达基因（DEGs）？
   你之前已经向我们展示过，第 1、2、5 和 6 组之间在 β 多样性上存在显著差异。结合我在第一个问题中的疑问：这些组在 β 多样性上是否彼此之间都显著不同，还是说我们需要进行事后分析（post-hoc）来判断具体是哪些组之间存在显著差异？在这里，我们主要关心的比较是 1 vs 5 和 2 vs 6。你能帮我们一起解读这些结果吗？

The table lists Hamburg state schools’ 2025 Abitur average grades and graduate counts, now with an index column as first column.¹

This guide explains how to complete the SRA metadata table for ten bulk RNA-seq samples derived from human skin organoids (SkOs). It includes one fully filled example row and shows the pattern to follow for the remaining samples.

Columns That Are the Same for All 10 Samples

According to the Methods section, all samples correspond to bulk RNA-seq using Lexogen RiboCop rRNA depletion and the Lexogen RNA-Seq V2 library kit. Sequencing was performed as single-end 1×83 bp reads on the Illumina NextSeq 500. Therefore, the following columns remain identical for all samples (untreated_rep1 … HSV_d8_rep2):

• Library strategy: RNA-Seq
• Library source: TRANSCRIPTOMIC
• Library selection: RANDOM (rRNA-depleted total RNA; “RANDOM” is the standard choice)
• Library layout: SINGLE
• Platform: ILLUMINA
• Instrument model: Illumina NextSeq 500
• Filetype: fastq (use this if uploading *.fastq.gz; otherwise, select the matching file type)

Columns That Differ Per Sample

1. Sample name This information comes from BioSample (e.g., untreated_rep1, HSV_d4_rep2, etc.).
2. Library ID Each row must have a unique internal ID. You can use a simple sequential pattern:

untreated_rep1 → LIB01  
untreated_rep2 → LIB02  
HSV_d2_rep1   → LIB03  
…  
HSV_d8_rep2   → LIB10

Alternatively, a more descriptive format such as SkO_bulk_untreated_rep1 is acceptable. The key requirement is that all IDs are unique.

3. Title Use the following pattern: Bulk RNA-seq of human skin organoid (SkO), CONDITION, replicate X

Examples:

untreated_rep1 → Bulk RNA-seq of human skin organoid (SkO), untreated, replicate 1  
untreated_rep2 → Bulk RNA-seq of human skin organoid (SkO), untreated, replicate 2  
HSV_d2_rep1   → Bulk RNA-seq of human skin organoid (SkO) infected with Human alphaherpesvirus 1 (HSV-1), 2 dpi, replicate 1  
HSV_d4_rep2   → Bulk RNA-seq of human skin organoid (SkO) infected with Human alphaherpesvirus 1 (HSV-1), 4 dpi, replicate 2  

Repeat this pattern for 6 dpi and 8 dpi samples.

4. Design description Use this base text and modify it per condition:

   Bulk RNA-seq of human hair-bearing skin organoids (SkOs) derived from hiPSC line UKEi001-A. Total RNA from two organoids per condition was pooled, depleted of rRNA (Lexogen RiboCop rRNA Depletion Kit for Human/Mouse/Rat V2), and libraries were prepared with the Lexogen RNA-Seq V2 Library Prep Kit with UDIs (short insert variant). Libraries were sequenced as 1×83 bp single-end reads on an Illumina NextSeq 500.

Then append the condition information as follows:

• For untreated rows: Condition: untreated control.
• For HSV_d2 rows: Condition: skin organoids infected with Human alphaherpesvirus 1 (HSV-1) for 2 days (2 dpi).
• For HSV_d4 rows: Condition: skin organoids infected with Human alphaherpesvirus 1 (HSV-1) for 4 days (4 dpi).
• Continue the same pattern for 6 dpi and 8 dpi.

5. Filename / filename2 Since this dataset is single-end, only the Filename column is required.
   • Filetype: fastq
   • Filename: must match the exact uploaded file name, e.g.:

untreated_rep1 → untreated_rep1.fastq.gz  
untreated_rep2 → untreated_rep2.fastq.gz  
HSV_d2_rep1   → HSV_d2_rep1.fastq.gz  
…

- **filename2:** leave blank (used only for paired-end libraries).

Example: Completed Row (untreated_rep1)

After completing this first row, copy and adjust the fields (Library ID, Title, Design description condition, and Filename) for the other nine samples accordingly.

Varicella-Zoster Virus (VZV) in complex skin organoids is about using advanced 3D human skin models to understand how VZV infects, spreads, and interacts with host immunity in its most relevant organ: the skin. 这些类器官尽量重建了人皮肤的层次结构和细胞多样性，使得对VZV的研究比传统二维细胞培养更接近真实感染情境。²³⁴⁵

为什么皮肤和类器官重要

VZV 的致病性高度依赖其在表皮角质形成细胞中的复制，这最终形成充满高滴度游离病毒的水疱，是人际传播和建立潜伏感染的关键步骤。 传统小鼠模型难以完全模拟人特异性VZV感染，因此发展成人皮肤组织模型、皮肤器官培养以及皮肤类器官对于研究病毒复制、皮损形成和疫苗株减毒机制至关重要。⁶⁷⁴⁸²

复杂皮肤类器官是什么

复杂皮肤类器官通常来源于人多能干细胞或皮肤前体细胞，经过定向分化形成多层表皮、真皮样结构，有时还包括毛囊、皮脂腺甚至神经和免疫细胞成分。 这些3D结构提供了立体组织环境和细胞间相互作用，可以更真实地研究VZV在不同表皮层的扩散、跨层传播以及向神经末梢的进入过程。³⁴⁹⁵²

在类器官中研究VZV复制与传播

类器官允许模拟病毒从基底层角质形成细胞开始复制，再向上层分化细胞扩展的过程，这与患者皮损中观察到的VZV复制模式高度一致。 通过在类器官中接种标记VZV，研究者可以动态监测病毒如何在组织内扩散、形成局灶性病灶，以及是否存在类似体内那样的“离散水疱”样病变。^7410862

免疫控制与固有限制因子

皮肤类器官为研究VZV如何与局部固有免疫和限制因子（如干扰素通路、自噬、PML核小体等）互作提供平台。 通过在类器官中操纵这些信号（例如阻断或增强I型干扰素、改变角质形成细胞分化相关通路），可以评估这些因素如何影响病毒复制强度、病灶大小以及病毒向神经末梢的传播能力。^4111052

向神经系统的连接与潜伏相关模型

VZV最终在感觉神经节建立潜伏，因此将皮肤类器官与人源神经元或神经类器官连接，是当前研究中的一个重要发展方向。 一些工作利用人神经元或神经类器官模型，已经显示可以在体外建立VZV潜伏和再激活系统，未来与皮肤类器官耦合可更好模拟“皮肤–神经–神经节”轴上的完整生命周期。^1213514

与其他模型相比的优势和局限

与单层细胞相比，皮肤类器官在细胞多样性、分化梯度和屏障特性方面更接近真实皮肤，因此对药物筛选、毒力基因鉴定以及疫苗株特性分析更具生理相关性。 不过，当前类器官通常缺乏完整血管系统和成熟适应性免疫成分，这限制了对全身免疫反应和T细胞介导控制的模拟，需要与动物模型或临床数据结合解读。¹⁵⁵⁶²

如果你愿意，可以继续围绕这个主题细化几个环节，例如：

• “如何在皮肤类器官中设计VZV基因功能筛选实验？”
• “如何把类器官数据嵌入到系统层面的病毒–宿主互作网络分析？” ^161718192021

模型系统与方法学

• 您所使用的复杂皮肤类器官在多大程度上重建了体内人皮肤的关键特征，例如分层结构、屏障形成、神经支配以及驻留免疫细胞？这些模型在解释VZV传播和免疫逃逸数据时目前主要的局限在哪里？
• 您是如何将经典病毒学方法与对受感染皮肤和神经类器官的多组学分析相结合，用于刻画VZV–宿主相互作用的时间动态？在区分“早期限制”与“成功免疫逃逸”方面，哪些读出指标被证明最具信息量？
• 在稳定感染复杂皮肤和周围神经系统类器官时，使用报告病毒时遇到的主要技术挑战是什么？您如何验证类器官中观察到的传播模式以及潜伏/再激活特征能够真实反映人皮肤和感觉神经节中的情况？

皮肤与神经中的传播

• 在您的比较分析中，皮肤类器官与神经类器官之间是否存在不同的细胞间传播方式或动力学？是否有特定的病毒基因或宿主通路在这两个解剖部位对传播发挥了差异化的调控作用？
• 在类器官不同表皮层中，病毒基因表达谱（例如基底层的立即早期基因与上层角化细胞的晚期糖蛋白表达）与体外或体内人皮肤模型中的描述有多接近？这对确定最有效的干预靶点有何启示？
• 与以往的皮肤器官培养或SCID-hu小鼠模型相比，类器官模型是否帮助澄清了VZV何时以及通过何种方式进入皮肤中的感觉神经末梢并建立潜伏感染？

先天限制因子与固有免疫

• 您提到已有工具用于研究POL III和PML核小体等限制因子；在皮肤微环境的类器官感染中，这些因子与VZV之间的相互作用有何新发现？VZV采用了哪些机制在局部对这些固有防御进行反制？
• 在您的筛选中，是否发现了特异于表皮或特异于神经元的关键抗病毒蛋白？它们的表达模式是否能解释体内水痘皮损呈离散分布且单个皮损大小受限的现象？
• 在类器官中阻断I型干扰素信号后，皮损大小及病毒产生量与既往异种移植实验相比有何差异？这一结果是否提示了VZV针对某些新的候选通路进行免疫逃逸？

免疫逃逸与免疫调控

• 在皮肤和神经类器官中，VZV是否同样能重现hiPSC来源神经球模型中报道的那种对干扰素信号和抗原呈递的深度抑制，例如ISG诱导减弱以及MHC II相关基因（如CD74）下调？
• 有研究显示VZV可以重塑T细胞表面标志促进其向皮肤归巢；在您的类器官微环境中，是否观察到类似的趋化因子、整合素或黏附分子改变，从而有利于组织内传播？
• 与传统二维角质形成细胞培养相比，在三维皮肤类器官中，VZV对表皮分化通路（例如K10降解以及对kallikrein家族的调控）的影响有何不同？这是否改变了我们对水疱形成及病毒排出的认识？

病毒基因与分子机制

• 通过类器官感染，对于特定病毒蛋白（如与PML核小体互作的ORF61）在决定皮肤趋向性、皮损形态以及溶菌性感染与免疫识别之间平衡中的作用，有哪些新的机制性见解？
• 近期多蛋白组学研究绘制了大量VZV–宿主蛋白互作网络；在类器官中，您验证了其中哪些互作具有功能意义？是否有出乎意料的宿主靶点被发现是调控传播或免疫逃逸的关键节点？
• 是否存在这样的病毒功能：在二维细胞培养中似乎可有可无，但在复杂皮肤类器官中却被证明对有效复制或传播是必需的，从而提示存在皮肤特异的毒力因子？

转化应用与未来方向

• 您如何展望利用皮肤和神经类器官作为VZV下一代疫苗或抗病毒药物的前临床评估平台，特别是那些旨在增强局部先天免疫应答或阻断特定免疫逃逸机制的策略？
• 鉴于人群中VZV相关疾病存在显著个体差异，您是否计划构建患者来源或基因分型分层的类器官面板，用以研究宿主遗传背景如何塑造限制因子表达以及对免疫逃逸的易感性？
• 在您看来，就VZV皮肤发病机制和免疫控制的若干悬而未决问题而言，类器官模型在未来几年最有优势回答的是哪些问题？哪些问题仍然必须依赖体内或临床研究来解决？

如果你希望，我也可以把这些问题进一步压缩成适合现场快速提问的精简版要点。 ^211345678910

Here are example seminar questions you could ask, focused on newly advanced work using complex skin (and neuronal) organoids for VZV spread, immune control, and evasion.

Model system and methodology

• How do the complex skin organoids used in your project recapitulate key in vivo features of human skin, such as stratification, barrier formation, innervation and resident immune cells, and what are the main limitations that still affect interpretation of VZV spread and immune evasion data?²⁴
• Could you elaborate on how you combined classical virology with multi‑omics in infected skin and neuronal organoids to map VZV–host interactions over time, and which readouts turned out to be most informative for distinguishing early restriction from successful immune evasion?¹¹²
• What were the main technical challenges in stably infecting complex skin and PNS organoids with reporter VZV, and how did you validate that the observed spread patterns and latency/reactivation features reflect the situation in human skin and sensory ganglia?⁴²

Viral spread in skin and neurons

• In your comparative analysis, did you observe distinct modes or kinetics of cell‑to‑cell spread in skin versus neuronal organoids, and are there specific viral genes or host pathways that differentially control spread in these two compartments?²⁴
• How closely does the pattern of viral gene expression across epidermal layers in organoids (immediate‑early in basal cells versus late glycoproteins in suprabasal layers) mirror what has been described in ex vivo or in vivo human skin, and what does that tell us about where intervention would be most effective?⁹⁴
• Have organoid models helped to clarify when and how VZV gains access to sensory nerve endings in skin and then establishes latency, compared with earlier skin organ culture or SCID‑hu mouse data?⁵⁴⁹

Innate restriction factors and intrinsic immunity

• You mentioned established tools to study POL III and PML nuclear bodies as restriction factors; what have organoid infections revealed about how VZV counters these particular intrinsic defenses in the skin microenvironment?⁷²
• Are there epidermis‑specific or neuron‑specific antiviral proteins that emerged as key restriction factors from your screens, and how do their expression patterns correlate with discrete lesion formation and limited lesion size seen in vivo?⁴²
• How does blocking type I interferon signaling in organoids affect lesion size and viral yield compared with earlier xenograft experiments, and does this reveal any new candidate pathways that VZV targets for immune evasion?⁹⁴

Immune evasion and modulation

• To what extent do skin and neuronal organoids reproduce the profound suppression of interferon signaling and antigen presentation that has been reported in hiPSC‑derived neurospheroids, for example reduced ISG induction and MHC class II–associated genes such as CD74?⁶⁸²
• Some studies show that VZV remodels T‑cell surface markers to promote skin homing; have similar VZV‑driven changes in chemokines, integrins or adhesion molecules been observed within the organoid microenvironment that might facilitate intra‑tissue spread?²⁹
• How does VZV manipulation of epidermal differentiation pathways (e.g., degradation of K10 and modulation of kallikreins) look in your 3D organoids compared with traditional monolayer keratinocyte cultures, and does this change how we think about blister formation and virus shedding?⁴

Viral genes and molecular mechanisms

• What new insights have organoid infections given into the role of specific viral proteins such as ORF61 (and its interaction with PML nuclear bodies) in determining skin tropism, lesion morphology, and the balance between lytic replication and immune detection?⁷⁹
• Recent multi‑proteomic surveys mapped dozens of VZV–host protein interactions; which of these interactions have you been able to validate functionally in organoids, and did any unexpected host targets emerge as central nodes for controlling spread or immune evasion?¹¹²
• Have you identified viral functions that appear dispensable in 2D cell culture but are clearly required for efficient replication or spread in complex skin organoids, suggesting skin‑specific virulence determinants?⁷²

Translational and future directions

• How do you envision using skin and neuronal organoids as preclinical platforms to test next‑generation VZV vaccines or antivirals, particularly those aimed at enhancing local innate responses or blocking specific immune‑evasive mechanisms?⁵⁹²
• Given the inter‑individual variability in human VZV disease, do you plan to develop patient‑derived or genetically stratified organoid panels to study how host genetics shapes restriction factor expression and susceptibility to immune evasion?¹¹²
• Which open questions about VZV skin pathogenesis and immune control do you think organoid models are best positioned to answer in the next few years, and where will in vivo or clinical studies still be indispensable?⁹²⁴

³¹⁰