《Nature communications》新突破性研究(全译文): CD39 PD-1 CD8 ...

Naturecommunications》新突破性研究(全译文): CD39 PD-1 CD8  T细胞介导乳腺癌转移性休眠

CD39 PD-1 CD8  Tcells mediate metastatic dormancy in breast cancer

Paulino Tallón de Lara, Maries van denBroek

Published: 03 February 2021

摘要

一些乳腺肿瘤侵袭性转移,而另一些则保持休眠多年。控制转移性休眠的机制尚不清楚。通过对小鼠进行高参数单细胞映射,本文在原发性肿瘤和隐匿性转移中发现了离散的CD39 PD-1 CD8  T细胞群,而在侵袭性转移瘤中几乎未发现此细胞群。通过阻断抗体,发现了休眠依赖于TNFαIFNγ免疫疗法减少了肺中休眠癌细胞的数量。纯化CD39 PD-1 CD8  T细胞的过继性转移防止了转移性生长。人乳腺癌中,CD39 PD-1 CD8  ,但不是全部的CD8  T细胞的频率与手术切除后的延迟转移性复发(无病生存)相关,因此强调了CD39 PD-1 CD8 T细胞在控制实验和人乳腺癌方面的生物学相关性。因此,本研究认为,原发性乳腺肿瘤可引发系统性CD39 PD-1 CD8  T细胞反应,该反应有利于肺中转移性休眠

引言

转移是乳腺癌患者死亡的主要原因。这种疾病的转移行为在受影响的器官和时间上不同【1,2】。有些病人在确诊后不久甚至在确诊前就发生转移,而有些病人在原发肿瘤切除后数年或数十年才出现转移【1,3】。事实上,约20%的无病患者在手术切除后7年内会复发,而那些似乎已治愈的乳腺癌患者甚至在术后20年的死亡率也高于其他人群【4,5】。认为这种晚期复发是由于播散性癌细胞(DCCs)造成的,此类细胞到达不同的器官,但休眠数年【6】。事实上,乳腺癌患者骨髓中DCCs的存在与转移复发的发展不太相关【7,8】。

目前,人们对控制转移性休眠的癌细胞的内在和外在机制还知之甚少。一些癌细胞内在因子与休眠有关,如抑制PI3K-AKT通路【9】、激活p38或内质网应激反应的触发【10】。此外,转移(前)器官的微环境可能通过不同信号诱导休眠【11】,例如内皮细胞衍生的血小板反应蛋白-1【12】或骨髓中产生的TGF-β2【13】。只有很少的研究涉及癌细胞外在因子的影响,如先天免疫【14,15】和适应性免疫【16,17】对休眠的影响。虽然原发性肿瘤中T细胞的浸润与良好的预后相关【18】,但目前尚不清楚这种相关性是由清除癌细胞的细胞毒性T细胞,还是由阻止癌细胞生长和诱导休眠的T细胞解释的【19,20】。

本文试图了解控制DCCs转移性休眠的机制,并发现原发性肿瘤引发了阻止肺中DCCs转移性生长的系统性CD8  T细胞反应。保护性T细胞表达了通常与活化和衰竭有关的标志物。促进这种反应可能为未来防止转移复发的免疫疗法的发展提供了理论基础。

研究发现CD39 PD-1 CD8  T细胞介导了乳腺癌临床前模型的转移性休眠,并与乳腺癌患者术后无病生存率的增加相关。

结果

播散性4T07乳腺癌细胞处于休眠状态

本文使用了两种乳腺癌的临床前模型,即4T1和4T07,它们最初来源于BALB/c小鼠的同一自发性肿瘤【21】(图1a)。4T1原位乳腺癌在不同器官中产生大转移【22,23】,而4T07基本上没有转移。然而,一些研究表明,在远端器官中存在播散性4T07细胞【21,24】。

1 4T07而非4T1乳腺癌细胞播散到肺部后处于休眠状态

a Diagramof the origin of 4T1 and 4T07 cell lines21. b Experimentaldesign. 4T07-mCh cells (105) or PBS were injected into the mammaryfat pad of female BALB/c mice. Analysis was performed 30dlater. c Quantification of disseminated 4T07-mCh cells byquantitative pathology, ****p<0.0001. d 4T07-mCh cells (105)were injected into the mammary fat pad of female BALB/c mice. Analysis wasperformed 35d later.Representative section of the primary tumor (upper panels) and lung (lowerpanels) at the endpoint. Disseminated 4T07-mCh cells were detected as single,Ki67-negative cells in the lungs of 5 out of 5 mice. Disseminated 4T07-mChcells are shown in red, Ki67 in green, DAPI in blue. Scale bar indicates 50µm. e Experimentaldesign. 4T07-mCh cells (105) were injected into the mammary fat padof female BALB/c mice. The tumor was resected 20d after injection and analysis was performed 21d afterresection (endpoint). f Quantification of disseminated4T07-mCh cells by quantitative pathology. ****p<0.0001. g Representative sectionof the lung at the endpoint. Disseminated 4T07-mCh cells were detected assingle, Ki67-negative cells in the lungs of 5 out of 5 mice. Disseminated4T07-mCh cells are shown in red, Ki67 in green, DAPI in blue. Scale barindicates 20µm. Eachsymbol represents an individual mouse. Five mice per group. 2-tailedStudent’s t-test. The bar represents the mean±SD.Results are representative of three independent experiments.

虽然4T1肿瘤在晚期生长较快,但原位注射的4T1和4T07细胞在免疫活性BALB/c小鼠中具有相似的乳腺癌生长和发病率(附加图1a和b)。但是,4T1乳腺癌容易诱导侵袭性生长和明显的肺转移,而4T07则没有(附加图1c和d)。尽管没有大转移,但在所有小鼠的肺部均发现了播散性4T07细胞(附加图1e),表明4T07 DCCs在肺部播散,但无法生长。通过免疫荧光确认并量化了所有携带4T07-mCh原发性肿瘤的小鼠肺部存在的休眠DCCs(图1b~d和附加图1f)。播散性4T07细胞以单细胞形式出现,呈Ki67阴性,正如之前对休眠DCCs所描述的那样【15,17】。我们在原发性肿瘤手术切除后21天发现4T07未增殖、播散(图1e~g),进一步强调了4T07 DCCs的休眠状态。

乳腺癌诱导了CD8  T细胞导致的休眠

首先通过静脉注射4T1或4T07乳腺癌细胞到幼鼠体内,研究了癌细胞的内在特征是否影响了肺中DCCs的生长(图2a)。在这些实验条件下,两种细胞系均可形成大转移(图2b),表明两种细胞系均可以在肺部逐步生长。

2 原发性肿瘤诱导的保护性免疫介导了转移性休眠

a Experimentaldesign. 4T1 or 4T07 cells (3×105) were injected i.v. into femaleBALB/c mice. Analysis 25d later. b Number of lungmetastatic nodules. 4T1, n=5; 4T07, n=4. c Experimental design. 4T1 or4T07 cells (105) were injected into the mammary fat pad of femaleBALB/c or BALB/c Foxn1nu/nu mice. Analysis 30d later.Lungs of “WT Late” mice were analyzed when the 4T07 tumor size reached the sizeof the BALB/c Foxn1nu/nu group (d 35). d Weightof primary tumors. Left panel with closed symbols, 4T1. Both groups n=5. Rightpanel with open symbols, 4T07. Foxn1nu/nun=7;WT, n=5; WT late, n=4, **p=0.006 (Foxn1nu/nu vs. WT),**p=0.0027 (Foxn1nu/nu vsWT Late). e Number of lung metastatic nodules. Left panel withclosed symbols, 4T1; right panel with open symbols, 4T07, **p=0.071for both comparisons. f Experimental design. Female BALB/cmice received 0.5mg anti-CD8 or isotype control i.p. at days −1 and 5relative to injection of 105 4T07-mCh cells in the mammary fatpad. Analysis 25d later. g Lung metastatic loadmeasured by bioluminescence, ****p<0.0001. Anti-CD8, n=9;isotype n = 10. h Bioluminescence images.Anti-CD8, n=9; isotype n = 10. i Experimentaldesign. 4T07 cells (105) or PBS were injected into the mammary fatpad of female BALB/c mice on d 0. On d 11, 3×105 4T07-LZ cells were injectedi.v.; analysis of lung metastatic load on d 25. j Quantificationof lung metastatic load by bioluminescence. 4T1, n=5;4T07, n=4, ***p=0.0021. k Bioluminescenceimages. l 4T07 cells (105) or PBS were injectedinto the mammary fat pad of female BALB/c mice on d 0. On day 8, mice wereinjected i.p. with 500µg anti-CD8 or isotype control. On d 15, 3×105 4T07-LZcells were i.v. injected and lung metastatic load was quantified on d 25. m Quantificationof the lung metastatic load by bioluminescence on d 25. All groups n=5, ****p<0.0001.Each symbol represents an individual mouse. ns=not significant. 2-tailed Student’s t-testwith Welch’s correction (bd left panel, e leftpanel, gj); 2-sided ANOVA with Bonferroni correction(d right panel, e right panel, m). Thebar represents the mean±SD. Results are representative of 3 (fgh)or 2 (all other) independent experiments.

髓细胞室经历了癌症诱导的变化,这些变化通过在4T1模型中创造转移前生态位促进了转移【25,26】。具体来说,4T1原发性肿瘤诱导了系统性中性粒细胞【27】和脾肿大【26】,并在转移前的肺中积聚了炎性单核细胞【28】、嗜酸性粒细胞、中性粒细胞【27】和肺泡巨噬细胞【29】。因此,比较了4T1-和4T07诱导的血液和肺部髓细胞室的变化,在脾肿大(附加图2a)或中性粒细胞(附加图2b)方面未观察到差异。4T1或4T07乳腺癌小鼠的肺部在流式细胞术(附加图2g)测定的炎性单核细胞(附加图2c)、中性粒细胞(附加图2d)、嗜酸性粒细胞(附加图2e)和肺泡巨噬细胞(附加图2f)的存在上有相似性。在4T1和4T07中观察到了之前报道的乳腺癌进展期间【27】肺中髓细胞的积聚(附加图2)。因此,4T1和4T07乳腺癌转移行为的差异不能通过髓细胞室的系统性改变来解释。

原位4T07肿瘤在T细胞缺乏的BALB/c Foxn1nu/nu 小鼠中诱导逐步生长的肺转移,而T细胞缺乏几乎不影响4T1肿瘤的转移行为(图2c~e)。为了比较两种乳腺癌细胞株的肺转移情况,在注射后的同一时间点分析了野生型和Foxn1nu/nu小鼠,并添加了另外一组野生型小鼠,在这组小鼠中,乳腺肿瘤生长到与Foxn1nu/nu 小鼠(WT晚期)中相同大小(图2d和e)。因此,播散性4T07乳腺癌细胞的转移休眠完全依赖于T细胞。在免疫缺陷小鼠中,4T1细胞本质上比4T07细胞更易转移,这一事实反映了4T1细胞和4T07细胞之间不同(结合)特征,其中一些特征与T细胞无关。事实上,我们认为4T1和4T07细胞之间的许多差异妨碍了在体内进行适当的决定性比较。

为了研究CD8  T细胞是否与转移性休眠有关,本研究从小鼠体内去除CD8  T细胞,然后原位注射4T07细胞,并通过IVIS分析肺转移负荷(图2f)。虽然原发性肿瘤的生长不受CD8缺失的影响,但在缺乏CD8  T细胞的情况下,播散性4T07-LZ细胞生长为大转移(图2g和h),表明原发性4T07乳腺癌诱导CD8  T细胞依赖性免疫。

为了验证CD8 T细胞对转移性休眠至关重要的假设,本文原位注射了未标记的4T07细胞(或PBS,作为对照),然后在11天后静脉注射荧光素酶标记的4T07-LZ细胞(图2i)。如果原发性肿瘤诱导了保护性免疫,预计肺转移负荷会减少。因为通过生物发光来测量肺转移负荷,所以本文特别量化了静脉注射的荧光素酶标记的4T07细胞。原发性4T07乳腺肿瘤的存在阻止了静脉注射4T07-LZ细胞的转移性生长(图2i~k),但不影响静脉注射后0.5和3小时测得的接种量(附加图3a和b)。终点时,静脉注射的细胞在肺部呈播散性、非循环的单个4T07细胞(附加图3c和d)。在静脉注射前手术切除原发性肿瘤并不影响休眠。具体而言,静脉注射的4T07-mCh细胞容易在对照小鼠中诱导明显的肺转移(乳腺和模拟手术中的PBS;附加图3e,f),而在4T07荷瘤小鼠的肺中仅检测到单个非增殖性4T07细胞,与手术切除无关(附加图3e,g)。使用4T07-LZ细胞和读出的生物发光在类似的实验装置中证实了这些结果(附加图4a,b)。静脉注射前CD8 细胞的耗尽导致了转移性生长(图2l和m),证实了休眠对CD8 T细胞的依赖性。相反,尽管4T07诱导的系统性免疫降低了静脉注射4T1-LZ细胞引起的转移负荷,但4T07和4T1乳腺癌均未对实验性4T1肺转移有保护作用(附加图4c和d)。尽管静脉注射4T1细胞导致的肺转移负荷在4T07荷瘤小鼠中降低,但病变仍在进行(即多细胞和Ki67 )且不休眠(附加图4a~d)。此外,我们认为过继转移的CD39 PD-1 CD8  T细胞数量不足以控制具有较高转移潜能的4T1细胞。

总之,这些数据表明4T07乳腺癌诱导了一种系统性的保护性免疫反应,这种反应介导了肺中DCCs的休眠。

CD39 PD-1 CD8  T细胞介导转移性休眠

首先,本文分析了原发性4T1和4T07肿瘤中不同免疫细胞的存在情况(附加图5a)。研究发现嗜酸性粒细胞(附加图5b)、中性粒细胞(附加图5c)、NK细胞(附加图5d)或炎性单核细胞(附加图5e)的数量没有差异,但4T1肿瘤包含的巨噬细胞数量高于4T07肿瘤(附加图5f)。

由于4T07转移性休眠依赖CD8 T细胞,本文用高维流式细胞术对4T1和4T07原发性肿瘤中的这些细胞进行了表征。首先,选通T细胞(附加图6a),使用二维t-随机邻域嵌入(tSNE)投影【30】(图3a)可视化23个参数面板中的所有标记物,并使用FlowSOM【31,32】对细胞进行聚类(图3a和b,上和左下图)。基于在聚类中观察到的中位数标记物强度(图3b,右下图),本文注释了主要细胞群(CD8 T细胞、CD4 T细胞和CD4  CD25 调节性T细胞),并确定了它们的频率和每个群体在tSNE投影中的位置。此外,与4T1肿瘤相比,4T07肿瘤中不仅CD39 PD-1 CD8  T细胞数量较大,而且其CD8  T细胞的比例也较高。调节性T细胞在4T1和4T07肿瘤中的频率相似,而传统的CD4 细胞在4T1肿瘤中比4T07更丰富。相反,4T07原发性肿瘤含有的CD8 T细胞的比例更高(图3b,左下图)。

3 休眠期乳腺癌中出现CD39 PD-1 CD8  T细胞

Female BALB/c mice were injected with 105 4T1or 4T07 cells in the mammary fat pad and primary tumors were analyzed by flowcytometry 20dlater. a t-SNE visualization of markers after gating onsingle, live, CD45  TCRβ  CD44  cells. b Upperpanel: t-SNE plots of the main T cell subsets. Lower left panel: frequency ofthe main T cell subsets in both groups. Each symbol represents an individualmouse, *p=0.019 (CD4  Tconv), *p=0.011(CD8  T cells). Lower right panel: Heat map with markerexpression of the main T cell subsets. c t-SNE visualizationof markers on CD8  TCRβ  cells. d Upperpanel: t-SNE plots of the main CD8  T cell subsets identifiedby FlowSOM algorithm. Lower panel: frequency of the main CD8  Tsubsets identified by FlowSOM algorithm in both groups, *p=0.028,**p=0.0069. e Scaledhistograms of arcsinh-transformed marker expression showing the relative markerdistribution of the population identified by CellCnn (red) among all CD8  Tcells (blue). KS indicates the Kolmogorov–Smirnov two-sample test between thewhole-cell population and the selected cell subsets. f Heatmap with marker expression of the main CD8  T cell subsetsidentified by FlowSOM algorithm. g Representative FACS plot of4T1 and 4T07 primary tumors after gating on CD8  TCRβ . h Frequencyof the population of total CD8  identified by CellCnn in bothmodels, ****p<0.0001. i Number of CD39 PD-1 CD8  Tcells in the lungs, *p=0.0317. j RepresentativeCD103-staining of live CD39 PD-1 CD8  Tcells from the lungs. Each symbol represents an individual mouse. n=4 pergroup for all panels, except panel h (n=3) and i (n=5). Two-tailed Student’s t-test withWelch’s correction. The bar represents the mean±SD.

值得注意的是,在CD8细胞室(图3c)中,观察到4T07乳腺癌聚集了CD39 PD-1 CD8  T细胞(图3d)。

为了证实CD39 PD-1 CD8  T细胞的富集是4T07和4T1乳腺癌之间的主要免疫差异,本文使用无偏表征学习算法CellCnn分析了高维流式细胞术数据【33,34】。与FlowSOM分析一致,CellCnn检测到CD39、PD-1、LAG3和Tim-3高表达的细胞群(图3e和附加图6a和b)。事实上,在全细胞群和所选细胞亚群(KS值较高)之间的Kolmogorov–Smirnov双样本试验中,最能确定群体并显示最大差异丰度的两个标记物是PD-1和CD39(图3e)。此外,CD39 PD-1 CD8  T细胞表达更多的LAG-3和Tim-3(图3f),将该群体表征为经历最近T细胞受体参与【35,36】和可能表明衰竭【37,38】的效应细胞。4T07肿瘤中该种群的丰度大约是4T1肿瘤的三倍(图3g和h,p < 0.001)。因此,分析了从4T07肿瘤中分离的CD39 PD-1 CD8  T细胞的功能,并证实了IFNγ和TNFα的产生(附加图6c)。此外,CD39 PD-1 CD8  T细胞在体外对4T07和4T1细胞显示出细胞毒性(附加图6d)。研究发现,在原发性4T07(而非4T1)乳腺癌中,这种CD39 PD-1 CD8  T细胞在第10天已经增加(附加图7a~d)。总之,这些结果表明,4T07肿瘤中的CD39 PD-1 CD8  T细胞具有保护效应性细胞功能。

重要的是,在患有原发性4T07乳腺癌的小鼠的肺中发现了CD39 PD-1 CD8  T细胞(图3i)。一部分群体表达了CD103,因此类似于组织驻留记忆T细胞【39,40】(图3j)。观察到CD39 PD-1 CD8  T细胞在原发性肿瘤中不表达CD103(图3f),这与最近的数据一致,这些数据表明人乳腺癌中的组织驻留样记忆细胞呈CD103阴性【41】。因此,CD39 PD-1 CD8  T细胞的积聚是休眠性肿瘤和转移性肿瘤的主要免疫差异。

因为在4T07中特异性出现的CD8 细胞表达了PD-1,我们想知道阻断PD-1是否可以改善其假定的保护性作用。事实上,抗PD-1免疫治疗可使肺中播散性4T07细胞数量减少约两倍(图4a和b),表明CD8  T细胞的PD-1 亚群包含了保护性效应细胞。为了研究CD39 PD-1 CD8  T细胞亚群是否确实控制了转移性生长,从而介导了转移性休眠,本研究从4T07乳腺癌中筛选了这一群体,并将其过继性转移到BALB/c幼鼠中,然后静脉注射4T07-LZ细胞(图4c和d)。CD39 PD-1 CD8  T细胞的过继性转移防止了4T07肺转移,而注射PBS或缺乏CD39 PD-1 细胞的CD8 T细胞(称为其他CD8 T细胞)则不能起到阻止作用(图4e和f)。在这些实验中,没有观察到休眠的细胞,因此,当CD39 PD-1 CD8  T细胞转移时,为静脉注射的细胞实际上没有到达肺部的可能性留下了一些空间。因此,进行了同样的实验,并观察了肺中播散性4T07 mCh细胞。在未接受CD39 PD-1 CD8  T细胞的小鼠所有肺中观察了转移性生长;相反,在接受CD39 PD-1 CD8 T细胞的小鼠中,在肺中发现了单一、不增殖、因此处于休眠状态的4T07 mCh细胞(附加图8)。因此,研究结果表明CD39 PD-1 CD8  T细胞对于控制散播性4T07乳腺癌细胞是必要且充分条件。

4:肿瘤相关CD39 PD-1 CD8  T细胞阻止转移性生长

a Experimentaldesign. 4T07-mCh cells (105) were injected into the mammary fat padof female BALB/c mice. The breast tumor was resected on d 20 and histologicalanalysis of lungs was performed on d 32. Mice received an i.p. injection with250µganti-PD-1 or isotype control on d 15, 18, 23, 26, and 29. b Enumerationof disseminated 4T07 cells by quantitative pathology (right panel). Each symbolrepresents an individual mouse. Isotype, n=6; anti-PD-1, n=7, *p=0.0152, (two-tailed Student’s t-testwith Welch’s correction). The bar represents the mean±SD. c Gating used for sorting ofCD44 CD39 PD-1 CD8  and therest of the CD44 CD8  population (termed other CD8  Tcells) from established 4T07 orthotopic breast tumors. d Experimentaldesign. Two-hundred-thousand sorted CD44 CD39 PD-1 CD8  orother CD44 CD8  T cells were transferred afterinjection of 105 4T07-LZ cells into female BALB/c mice. Allinjections were given intravenously. Lung metastatic load was determined bybioluminescence 14 d later. e Bioluminescence of lungs atendpoint. f Quantification of lung metastatic load bybioluminescence, **p=0.0058, ***p=0.0002. Each symbol represents an individual mouse.Five mice per group. **p<0.01, ***p<0.001 (ANOVA with Bonferroni’s correction). The barrepresents the mean±SD. g Experimental design.4T07-mCh cells (105) were injected into the mammary fat pad offemale BALB/c mice. Mice received an i.p. injection with 500µg anti-IFNγplus 500µganti-TNFα every 3rd day. Control mice received isotype control antibody. On d20, 4T07-mCh cells were visualized in lungs by immunofluorescence. h Tworepresentative examples showing clusters of proliferating 4T07-mCh cells in thelungs of mice treated with anti-IFNγ plus anti-TNFα. Proliferating 4T0-mCh7cells were detected in the lungs from 4 out of 8 mice.

肺中存在非增殖播散性4T07细胞,表明CD8 T细胞通过细胞周期停止介导休眠。因此,检测了CD39 PD-1 CD8  T细胞衍生的IFNγ和TNFα是否可通过将4T07细胞体外暴露于这些因子而诱导衰老【42】。根据衰老相关β-半乳糖苷酶活性的表达,观察到衰老细胞的数量显著增加(附加图9a和b)。为了证实IFNα/TNFα诱导体内衰老的相关性,本研究阻断了4T07 mCh乳腺癌小鼠体内的两种细胞因子,并通过免疫荧光分析了肺中播散性4T07细胞的状态(图4g)。阻断IFNγ和TNFα可使播散性细胞转移爆发,如增殖4T07细胞集群的存在所示(图4h)。在8只接受抗IFNγ和抗TNFα治疗的小鼠中观察到4只发生肺转移性爆发。在同型治疗小鼠的肺中,没有发现转移性爆发;相反,在所有分析小鼠的肺中观察到了单一、不增殖的播散性细胞。因此,阻断IFNγ和TNFα后,肺中仅存在多细胞转移性4T07结节,说明休眠的诱导依赖于这两种细胞因子。

因此,4T07原发性肿瘤激发系统性CD8 T细胞,这些细胞通过IFNγ和TNFα诱导DCC的细胞周期停止,从而预防明显的转移性疾病。

CD39 PD-1 CD8  T细胞中效应细胞的功能与衰竭

为了了解该细胞群体控制DCCs的原因,本文从4T07乳腺癌中分离了CD39 PD-1 CD8 和其他CD8 T细胞,并比较了它们的转录组(图5a和b以及附加图10a)。CD39 PD-1 CD8  细胞显示与效应细胞功能相关的转录物的过度表达,包括Prf1FaslGzmfGzmeGzmdGzmcGzmbGzmkIfng(图5c)。基因集富集分析(GSEA)【43】证实了这一点,表明分类后的CD39 PD-1 CD8  细胞的转录组与人类组织驻留的CD8 T细胞有很强的相似性(图5d),最近发现这与三阴性乳腺癌的生存率提高相关【19】。这与它们独特的细胞阻止转移进展的能力一致。此外,在4T1乳腺癌中从未观察到休眠的事实是因为4T1携带宿主中此类保护性T细胞的数量较少(图3),而不是因为其转录特征,因为4T1和4T07肿瘤中的CD39 PD-1 CD8 细胞没有太大差异(附加图10b~d)。CD39 PD-1 CD8 细胞的转录组也富含活化和衰竭的标记物,如Vsir (VISTA)、Tnfrsf18 (GITR)、TigitCd224 (2B4)、Tnfrsf9 (4-1BB)、Tnfrsf4 (OX-40)、IcosLag3Ctla4Havcr2 (Tim-3)。这与转录因子TOX的过度表达一致(图5c),该转录因子导致在慢性抗原刺激期间耗竭标记物的表达和效应细胞功能的持久性【44】。

5 CD39 PD-1 CD8  T细胞具有独特的转录组特征

a Hierarchicalclustering of significantly differentially expressed genes between sorted CD44 CD39 PD-1 CD8  andCD44 CD8  T cells. b Volcano plotcomparing transcripts in CD44 CD39 PD-1 CD8  Tcells with those in other CD44 CD8  T cells sortedfrom 4T07 breast tumors. The red symbols represent transcripts that aresignificantly over-expressed in CD44 CD39 PD-1 CD8  Tcells, whereas the blue symbols represent significantly under-expressedtranscripts. c Heat map showing relative expression ofselected transcripts in CD44 CD39 PD-1 CD8  (threetop rows) and other CD44 CD8  T cells (three bottomrows) identified by differential gene expression analysis. d Geneset enrichment analysis comparing the transcriptional profile to the datapublished by Goldrath et al.80 (left panel) andSavas et al.19 (right panel).

综上所述,CD39 PD-1 CD8  细胞具有由免疫检查点分子和效应细胞蛋白共同表达决定的独特的转录组特征。

CD39 PD-1 CD8 T细胞与人乳腺癌的生存相关

为了研究临床前研究结果的临床相关性,本文分析了54名患者的原发性乳腺癌组织,使用5倍免疫荧光检测了CD8、CD39、PD-1和上皮细胞的存在(图6a),发现这些患者在原发性肿瘤手术切除后有转移复发(附加表1)。队列研究显示了个体亚型的典型生存曲线,表明其具有代表性(附加图11a)。高密度肿瘤内CD39 PD-1 CD8 T细胞的患者术后无病生存期明显长于低密度肿瘤内CD39 PD-1 CD8 T细胞的患者(图6b)。肿瘤外CD39 PD-1 CD8  T细胞的密度与无病生存率无关(附加图11b),而CD39 PD-1 CD8  T 细胞的密度与位置无关(附加图11c)。此外,在管腔A和B患者中观察到了类似数据(附加图11d和e)。根据多变量Cox回归分析,肿瘤内CD39 PD-1 CD8  T细胞的密度不是独立变量(附加表1)。重要的是,肿瘤内CD8 T细胞的密度与无病生存率无相关性(图6c),有力地支持了临床前观察结果,即播散性癌细胞受CD39 PD-1 CD8  T细胞这一特定亚群的控制。

6 肿瘤内CD39 PD-1 CD8  而不是总CD8 T细胞的高密度与乳腺癌的无病生存相关

a Representativeimages of 5-color multiplex immunofluorescence on human breast cancer. Stainingshows epithelial cells (PanCK EpCAM, yellow), CD8  T cells (CD8,magenta), PD-1 (green), CD39 (red) and nuclear staining (DAPI, blue). Scale baris 50µm. b Disease-freesurvival of 54 patients with high or low number of intra-tumoral CD39 PD-1 CD8  Tcells. c Disease-free survival of 54 patients with high or lownumber of intra-tumoral CD8  T cells. The threshold for separatingpatients with high and low CD39 PD-1 CD8  Tcell densities was defined using ROC curve analysis. Survival was compared forpatients with high and low density of cells to be compared as indicated for theindividual graphs by Kaplan–Meier curves. Confidence intervals are indicated asshaded areas surrounding survival curves. Significance was calculated bylog-rank test. Patient numbers at risk are displayed for each 15 years offollow-up.

讨论

研究发现乳腺癌可以引发系统性CD8 T细胞反应,该反应介导肺部转移性休眠,从而防止临床转移性生长。因此,除了众所周知的促癌转移能力外,本文还描述了原发性肿的一个意外特点【25,26,45,46】。利用原发性肿瘤的高维单细胞谱,本文确定了保护性CD8 T细胞群体为PD-1 和CD39 细胞,表明最近的同源相互作用和肿瘤特异性【20,47】。与临床前数据一致,本文发现肿瘤内CD39 PD-1 CD8 T细胞的高密度与乳腺癌患者手术切除后无病生存显著相关。即使在管腔A或B亚型患者中,其中肿瘤组织CD8 T细胞的浸润通常较少【48】,CD39 PD-1 CD8  T细胞的存在与原发性肿瘤手术切除后的无病生存相关。如临床前模型所示,包含所有CD8 T细胞的异质性群体与无病生存不相关,有力地支持了CD39 PD-1 CD8  T细胞组成了独特的细胞群体来控制播散性癌细胞的观点。

CD39在癌症中的作用相当复杂。一方面,有证据表明CD39的表达标志着肿瘤特异性T细胞的存在,这种细胞的存在与更好的预后或免疫介导的控制相关【47】。另一方面,CD39由肿瘤微环境中的某些癌细胞和抑制细胞(Tregs和髓细胞)组成性表达,并催化促炎细胞外ATP,导致CD8 T细胞依赖性肿瘤控制受损【49,50】。同样的研究描述了在阻断CD39的酶活性后,肿瘤相关巨噬细胞和单核细胞减少,而T细胞功能(和肿瘤控制)更佳。

观察到 CD39 PD-1 CD8  T细胞表达了效应细胞分子,但也表达与衰竭相关的标志物。最近在人乳腺癌中也发现了类似群体【19,51-53】。认为耗尽的T细胞效应细胞功能降低且丧失了增殖能力【54,55】。尽管如此,乳腺癌中耗尽的T细胞可能比黑色素瘤更具功能性【56】。

除局部免疫外,免疫治疗诱导系统性免疫应答对其疗效至关重要【57】,说明保护性CD8 T细胞被募集到肿瘤中。循环的CD8 T细胞也可以进入组织,在那里发展成CD103 组织驻留记忆细胞,并可能通过诱导细胞死亡或细胞周期停止来保护组织免受肿瘤细胞的扩散【58】。事实上,在原发性肿瘤和肺部检测到CD39 PD-1 CD8 T细胞,并发现有一部分表达的CD103可作为组织驻留的标志。最近的研究表明,肿瘤内CD103表达的CD8 T细胞与人乳腺癌的生存相关【41】。

如上所述,转移性休眠可能由两种不同的状态引起【59】:静止的播散性单细胞【60】,或通过增殖和死亡之间的平衡保持稳定的微转移【61】。研究发现播散性4T07细胞以单个非循环细胞的形式存在于肺中,表明通过细胞周期停止进入休眠状态。观察到休眠本质上依赖于CD8 T细胞,揭示了一种新的休眠机制。我们发现保护性CD39 PD-1 CD8  T细胞产生TNFα和IFNγ,在Th1介导的抗肿瘤保护的背景下,它们被描述为诱导不可逆衰老【42】;同样的细胞因子阻止了播散性4T07乳腺癌细胞在肺部的增殖。

可能是环境因素诱导播散性癌细胞具有癌干细胞样的特性【62】,这与低循环频率相一致。这种癌干细胞代表着对环境变化作出反应的进行性转移的有效储库【63】。为什么休眠癌细胞有时几十年后才苏醒【3】,这仍然是个谜,不能排除不同的途径会聚在一起,以启动休眠病变的生长。假设CD8 T细胞是维持休眠所必需的,可以想象,组织驻留记忆CD8 T细胞的消耗可能导致休眠细胞的觉醒。肿瘤切除和随后的抗原丢失可能导致免疫记忆减弱;组织驻留记忆细胞在缺乏抗原的情况下特别不稳定【64,65】。另外,休眠细胞本身也可能发生改变,例如失去阻止免疫介导清除的分子的表达【66】。最后,最广义的环境因素可能唤醒休眠的癌细胞。这种微环境的变化包括纤维化【67】、组织重塑【68】、肥胖【69】、炎症【70】、血管内稳态紊乱【71】、糖皮质激素【72】或香烟烟气【73】。目前尚不清楚这些因素如何诱导休眠细胞的循环,以及免疫细胞是否能在这一阶段进行干预。

本文发现CD39 PD-1 CD8  T细胞介导了播散性癌细胞的休眠,但它们本身不能完全清除所有癌细胞。这与临床高度相关,因为休眠癌细胞很可能是未来转移性疾病的源头。由于决定觉醒的因素尚不清楚,甚至可能无法控制,因此更好地了解休眠癌细胞的性质以及必须调动哪些途径才能完全消除它们很重要

Methods

Mice

BALB/cJRj and BALB/cAnNRj-Foxn1nu/nu werepurchased from Janvier labs (Roubaix, FR). NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJmice were originally obtained from the Jackson Laboratory and provided byChristian Münz (University of Zurich, Switzerland). Mice were kept underspecific pathogen-free conditions in individually ventilated cages at theLaboratory Animal Services Center at the University of Zurich. Mice had accessto food and water ad libitum and were maintained on a 12-h light/dark cyclewith environmental enrichment. All experiments were performed with8–14-weeks-old female mice in accordance with the Swiss federal and cantonalregulations on animal protection and were approved by The Cantonal VeterinaryOffice Zurich (156/2018).

Cell lines

4T07 and 4T1 cells were a gift from Fred Miller(Karmanos Cancer Institute, Detroit, USA). Cells were cultured in Dulbecco’sModified Eagle’s Medium (DMEM, Gibco) supplemented with 10% fetal bovine serum(FBS, ThermoFisher Scientific), 2mM L-glutamine and 2% penicillin/streptomycin(ThermoFisher Scientific). Cells were cultured at 37°C in a humidified atmosphere with 5% CO2.4T1 and 4T07 were lentivirally transduced to express firefly luciferase andZsGreen (pHIV-Luc-ZsGreen, addgene plasmid #39196) or mCherry(pCDH-CMV-mCherry-T2A-Puro, addgene plasmid #72264). Transduced cells weresorted based on expression of GFP or mCherry, respectively. Luciferase-ZsGreenis referred to as LZ, mCherry as mCh.

Only cells of early passages were used forexperiments. Cells were regularly tested negative for mycoplasma by PCRanalysis. Cells were also tested negative for 18 additional mouse pathogens byPCR (IMPACT II Test, IDEXX Bioanalytics).

In vivo tumor experiments and treatments

Hundred-thousand 4T1 or 4T07 cells in 50µl PBSwere injected into the fourth mammary fat pad. Alternatively, 3×105 cellsin 50µl PBSwere injected into the lateral tail vein.

For resection of primary tumors, mice wereanesthetized with 2.5% isoflurane and given 0.04mg/kg fentanyl (Kantonsapotheke Zurich) i.p. aspre-emptive analgesia. Primary tumors were resected, and wounds were closedusing Autoclip wound clips (BD Biosciences). For post-operative analgesia 0.1mg/kgbuprenorphine (Temgesic, Schering-Plough) was given i.p. immediately aftersurgery and in the drinking water at 10µg/ml for 48h ad libitum.

For depletion of CD8  T cells, 500µg ratanti-mouse CD8 (clone YTS 169.4, hybridoma originally obtained from H.Waldmann, Oxford, United Kingdom) was injected intraperitoneally (i.p.) in PBSas described in each experiment. Control mice were injected with 500µg ratanti-Trinitrophenol (clone 2A3, BioXCell). Antibodies were purified fromhybridoma culture supernatant using protein G Sepharose 4 Fast Flow (Sigma).Injection of the respective antibody resulted in depletion of >90% of CD4  orCD8  T cells for at least 14 days, as determined by flowcytometry. Antibodies were administered i.p. in 200µl PBS. Full depletion of the targeted populationwas confirmed by flow cytometry on blood 2 d after injection of the antibody inevery experiment.

For blockade of PD-1, mice were injected with 250µganti-PD-1 (clone RMP1-14, made in-house by H. Yagita) as indicated. Controlmice received 250µg rat-anti-trinitrophenol (clone 2A3, BioXCell).

For blockade of INFγ and TNFα, mice were injectedevery 3rd day starting on day 3 until the endpoint with 500µganti-INFγ (clone R4-6A2, BioXCell) plus 500µg anti-TNFα (clone XT3.11, BioXCell) in 150µl PBS.Control mice received 1000µg rat-anti-trinitrophenol (clone 2A3, BioXCell).

Lung metastasis from luciferase- ormCherry-expressing tumors was quantified using an IVIS200 imaging system(PerkinElmer) as previously described23. Briefly, forluciferase-expressing tumors, mice were injected i.p. with 150mg/kgD-luciferin (Promega) and photon flux was measured 20min later in vivo as well as from dissected lungs.Lung metastasis from parental tumors was quantified by India Ink as previouslydescribed23. Briefly, mice wereeuthanized and India Ink (Pelikan, 15% in PBS) was injected intratracheally,lungs were harvested, washed in PBS and fixed in Fekete’s solution (62%ethanol, 3.3% formaldehyde, 0.25M acetic acid). Metastatic foci were countedblinded using a dissection microscope.

Disseminated cancer cells were detected in thelungs using a colony-forming assay as previously described21. Briefly, lungs wereresected from euthanized mice, cut into small pieces and digested for 45min at37°C inDMEM with 1mg/mlcollagenase IV and 2.6µg/ml DNase I (both Sigma) on a rotating device.Subsequently, samples were washed with PBS by centrifugation at 350×g. For detection of circulating cancer cells, bloodwas collected by heart puncture in a 25-gauge syringe containing 100μlheparin (5000IU/ml,Braun) and cells were washed with PBS by centrifugation at 350×g. Blood and lung samples were suspended in completeDMEM containing 6µM of 6-thioguanine and cultured in a T175 flask.Medium was exchanged after 5–7 days and presence of colonies of tumor cells wasevaluated after 14 days by light microscopy and crystal violet staining.

Immunofluorescence of mouse samples

Organs were fixed with 4% paraformaldehyde(Roti-Histofix 4%, Roth) for 10min at RT and cryoembedded in Optimal CuttingTemperature (O.C.T.) Compound (O.C.T.TM Compound, Tissue-Tek)using dry ice/100% ethanol slurry. Ten-µm thick sections cut using a cryotome(Leica), mounted on SuperFrost glass slides, dried for 1h at 37°C andpreserved at −80°C untilimmunostained. To prevent unspecific binding of antibodies, slides wereincubated with 4% donkey serum in PBS (ANAWA BioWorld) for 10min atRT. Subsequently, slides were incubated overnight at 4°C with primary antibody diluted in PBS 1% donkeyserum. Following primary antibodies were used: Anti-mCherry (goat polyclonalantibody, AB8181-200 SICGEN, 1:400), Ki67 (rabbit monoclonal antibody, ab16667abcam, 1:100). After incubation, slides were washed three times with PBS Tween20 0.05% and incubated for 1h at room temperature with secondary antibodiesdiluted in PBS 1% donkey serum. Following secondary antibodies were used:AF594-conjugated anti-goat IgG, AF488-conjugated anti-rabbit IgG (both JacksonImmunoResearch, 1:400). Finally, slides were washed, incubated with 0.5µg/ml4′,6 diamidine-2-phenylindole (DAPI; Invitrogen, 1:5000) for 5min,washed again and mounted with ProlongDiamond medium (Invitrogen). The slideswere scanned using the automated multispectral microscopy system Vectra 3.0(PerkinElmer). An unstained slide was used to generate the spectral profile ofautofluorescence in studied tissues. The Inform software (PerkinElmer) was usedfor spectral unmixing of individual fluorophores and autofluorescence.

Flow cytometry

Animals were euthanized by isoflurane overdose. Theright heart ventricle was perfused with 10ml PBS to eliminate the blood from the lungvessels. Primary tumors and lungs were collected in DMEM, cut into small piecesand digested for 45min at 37°C in DMEM containing 1mg/ml collagenase IV and 2.6µg/mlDNase I (both Sigma) on a rotating device. Samples were washed with PBS bycentrifugation for 5min at 350×g, the pelletwas suspended in PBS and filtered through a 70-μm filter (BD Biosciences) toobtain a single cell suspension. For lymphocyte analysis, cells were furtherpurified by centrifugation over a Percoll gradient (GE Healthcare, 17-0891-01,Sigma Aldrich).

Single cells were stained according to standardprotocols. Briefly, cells were surface-stained in 50µl antibody-mix in PBS. For intracellular cytokinestaining, cells were stimulated with 100ng/ml phorbol 12-myristate 13-acetate (PMA) plus 1µg/mlionomycin for 4h at 37°C in the presence of GolgiPlug/GolgiStop (BDPharmigen). Cells were stained for surface molecules as described above, washedwith PBS, and fixed for 30min on ice using IC Fixation Buffer fromFoxp3/Transcription Factor Staining Buffer Set (eBioscience). Subsequently,cells were stained for intracellular cytokines in permeabilization buffer fromthe Foxp3/Transcription Factor Staining Buffer Set overnight at 4°C.After washing with permeabilization buffer, samples were suspended in FACSbuffer (PBS, 20mM EDTA pH 8.0, 2% FCS) and acquired using a CyAnADP9 flow cytometer (Beckman Coulter), FACS LSRII Fortessa or FACSymphony (bothBD Biosciences). For quantitative analysis, CountBright absolute counting beadswere used (ThermoFisher Scientific). In all staining, dead cells were excludedusing Live/Dead fixable staining reagents (Invitrogen), and doublets wereexcluded by FSC-A versus FSC-H and SSC-A versus SSC-H gating. Followingdirectly labeled anti-mouse primary antibodies were used: Anti-CD8a in BUV 805 (clone53-6.7, rat IgG2a, BD Pharmigen, 1:100), anti-CD11b in BUV 661 (clone M1/70,rat IgG2b, BD Pharmigen, 1:400), anti-CD45.2 in BUV 653 (clone 30-F11, ratIgG2b, BD Pharmigen, 1:400), anti-VISTA in AF488 (clone MH5A, armenian hamsterIgG1, BioLegend, 1:100), anti-CD39 in PerCP-eFluor710 (clone 24DMS1, rat IgG2b,ThermoFisher Scientific, 1:100), anti-LAG3 in BV 421 (clone C9B7W, rat IgG1,BioLegend, 1:100), anti-CD44 in BV 570 (clone IM7, rat IgG2b, BioLegend,1:100), anti-CD73 in BV 605 (clone TY/11.8, rat IgG1, BioLegend, 1:100),anti-CD25 in BV 650 (clone PC61, rat IgG1, BioLegend, 1:100), anti-PD-1 in BV785 (clone 29F.1A12, rat IgG2a, BioLegend, 1:100), anti-TCRβ in PE-Cy5 (cloneH57-597, armenian hamster IgG1, BioLegend, 1:400), anti-KLRG1 in APC-Cy7 (clone2F1/KLRG1, armenian hamster IgG1, BioLegend, 1:100), anti-TIM3 in AF647, cloneB8.2C12, rat IgG1, BioLegend, 1:100), anti-CD103 in Biotin, clone 2E7, armenianhamster IgG1, BioLegend, 1:100), anti-Streptavidin in BUV 395 (BD, 1:200),anti-Ki67 in BV 480 (clone B56, mouse IgG1, BD, 1:100), anti-TNFα in BV711(clone MP6-XT22, rat IgG1, BioLegend, 1:400), anti-INFγ BUV 737 (clone XMG1.2,rat IgG1, BD, 1:100), anti-CD4 in BUV 496 (clone GK1.5, rat IgG2b, BD, 1:200),anti-FOXP3 in PE (clone FJK-16s, rat IgG2a, ThermoFisher Scientific, 1:200),anti-EOMES in PE-eFluor610 (clone Dan11mag, rat IgG2a, ThermoFisher Scientific,1:200), anti-T-bet in PE-Cy7 (clone eBio4B10, rat IgG1, ThermoFisherScientific, 1:100), anti-CTLA4 in APC-AR700 (clone UC10-4F10-11, armenianhamster IgG1, BD, 1:100) and anti-CD24 in FITC (clone M1/69, rat IgG2b,BioLegend, 1:200).

Flow cytometry data were compensated and exportedusing FlowJo software (version 10, TreeStar Inc.). The exported FCS files werenormalized using Cyt3 MATLAB (version 2017b) and uploaded into Rstudio (Rsoftware environment, version 3.4.0). tSNE and FlowSOM algorithm mapping livecells from a pooled sample were performed as described74. CellCNN was run usingdefault parameters, dividing data into training and validation steps asdescribed33.

Purification of CD39 PD-1 CD8  Tcells

Primary tumors were processed as described under“Flow Cytometry”. After the Percoll gradient, the leukocyte fraction wasenriched for CD8  T cells using anti-mouse CD8a microbeads(Miltenyi Biotec) and autoMACS equipment (Miltenyi Biotec) according to themanufacturer’s instructions. Subsequently, cells were stained for CD44, CD39,PD-1, and CD8 as described under “Flow cytometry” and live, single CD44 CD39 PD-1 CD8  Tcells, as well as other CD44 CD8  T cells, weresorted using an ARIA III Sorter (BD Biosciences).

Adoptive transfer of CD39 PD-1 CD8  Tcells

Sorted cells were counted using trypan blue (TrypanBlue solution, Sigma Aldrich) to exclude dead cells. Two hundred-thousand CD44 CD39 PD-1 CD8  Tcells were injected immediately after 105 4T07-LZ cells.Injections were given intravenously.

In vitro induction of senescence

Twenty-thousand 4T07 cells were plated in a 6-wellplate. Subsequently, 75ng/ml mouse IFNγ (R&D Systems) plus 5ng/mlmouse TNFα (PeproTech) were added and cells were incubated at 37°C in ahumidified atmosphere with 5% CO2 for 6 days as described. Theproportion of senescent cells was determined by staining for β-galactosidaseactivity using Senescence β-Galactosidase Staining Kit (9860, Cell SignalingTechnology) as described42.

In vitro cytotoxicity assay

CD44 CD39 PD-1 CD8  Tcells were sorted from 4T07 tumors as described above. T cells were culturedovernight at 37°C in a humidified atmosphere with 5% CO2 withmouse T-Activator CD3/CD28 beads (Dynabeads, ThermoFisher) plus recombinantmouse IL-2 (100U/ml, eBioscience). Cultures were performed in96-well round-bottom plates with 50,000 T cells and (if applicable) 50,000activator beads per well. Subsequently, 10,000 live 4T1 or 4T07 cells wereincubated with 50,000 live T cells (or medium as control) per well of a 96-wellround-bottom plate for 6h at 37°C in a humidified atmosphere with 5% CO2.Co-cultures were collected and stained with anti-mouse CD24-FITC (clone M1/69,1:200), anti-mouse CD45.2-APC (clone 104, 1:200), anti-mouse CD8a-BV421 (clone53-6.7, 1:200) and Zombie-NIR fixable dye (all BioLegend) and samples wereacquired on a LSRII Fortessa flow cytometer (BD). T cells were identified asCD45.2 CD8 CD24, 4T1 and 4T07 target cells as CD45.2-CD8-CD24 .Data were analyzed using FlowJo v10 software (Tree Star). The percentage dead(Zombie-NIR ) target cells was determined after gating on CD45.2-CD8-CD24  cells.

RNA-sequencing

Library preparation

The libraries were prepared following theSmart-seq2 protocol75. Five-hundred pg of cDNAfrom each sample were tagmented and amplified using Illumina Nextera XT kit.The resulting libraries were pooled, double-sided size selected (0.5× followedby 0.8× ratio using Beckman Ampure XP beads) and quantified using an Agilent4200 TapeStation System. The pool of libraries was sequenced in an IlluminaNovaSeq6000 sequencer (single-end 100bp) with a depth of around 20 Mio reads per sampleas a service by the Functional Genomics Center Zurich (FGCZ).

Data evaluation

The raw data generated by Illumina NovaSeq6000sequencer were analyzed using SUSHI, a framework for analysis of NGS datadeveloped by the FGCZ76,77. Briefly, after qualitycontrol, the reads were aligned to a reference genome (Ensembl genomeMus_musculus. GRCm38 dated 2018.02.26) with STAR 2.7.3a. The software packageEdgeR was used to detect differentially expressed genes. We applied a thresholdof p<0.01, FDR<1% and a log fold-change of >1.0 for upregulatedgenes and< −1.0 for downregulated genes. Unsupervised hierarchical clustering was doneusing the Ward2 method. Heatmaps were generated with the software FunRichV3.1.3.

Gene set enrichment analysis43,78 was performed on alist of genes ranked from high to low estimated fold-change using the GSEA4.0.3 Software (Broad Institute) with enrichment statistic classic and 1000permutations.

Patient material

Tumor tissues from 54 patients with breast cancer(Supplementary Table 1) were collected at theUniversity Hospital Zurich, Switzerland. The cohort was established with theintention to eventually compare the immune infiltrate among primary breastcancer tissue and intrapatient matched distant metastatic sites. To this end, wesearched breast cancer patients suffering from either invasive-ductal orinvasive-lobular breast cancer with hematogenous metastases in the archives ofthe Department of Pathology and Molecular Pathology, University HospitalZurich. Our cohort is, therefore, biased for patients with advanced metastaticdisease and differs from an average breast cancer cohort. While we cannotformally exclude any bias by this cohort selection we are not aware of anystudies systematically comparing the tumor immune microenvironment in primarybreast cancer between metastatic and non-metastatic cases. Donors providedwritten, informed consent to tissue collection, analysis and data publicationaccording to the Declaration of Helsinki. Law abidance was reviewed andapproved by the ethics commission of the Canton Zurich (BASEC-Nr. 2018-02282and KEK-ZH-2013-0584). Samples were numerically coded to protect donors’ rightsto confidentiality and privacy.

Immunofluorescence of human samples

Formalin-fixed paraffin-embedded samples were cutinto 2µm-thicksections and dried overnight at 55°C. Antigen retrieval and deparaffinization wereperformed in 1× Trilogy Solution (Cell Marque, 920P-06) for 15min at115°C.Staining was performed employing tyramide signal amplification (TSA) and theOpalTM 7-color Manual IHC Kit (PerkinElmer). To preventunspecific binding of antibodies, slides were incubated with 4% donkey serum inPBS (ANAWA BioWorld) in PBS/0.1% Triton X-100 for 15min at 37°C. TSA staining protocol was performed as described79. Briefly, slides wereincubated overnight at 4°C with primary antibody diluted in PBS/1% donkeyserum/0.1% Triton X-100. Subsequently, slides were washed in PBS Triton X-1000.1% and incubated with corresponding HRP-conjugated secondary antibodies.Following primary antibodies were used: Anti-CD39 (mouse monoclonal antibody,clone A1, BioLegend, 1:500), anti-PD-1 (rabbit monoclonal antibody, clone D4W2JBioConcept, 1:4000), anti-CD8 (mouse monoclonal antibody, clone RPA-T8, CellSignaling, 1:2000), anti-PAN-Cytokeratin (rabbit polyclonal antibody, cat. no.H-240, Santa Cruz, 1:2000) and anti-EpCAM (rabbit monoclonal antibody, cloneEPR20532-225, Abcam, 1:200). Following HRP-conjugated secondary antibodies wereused: anti-mouse IgG (H L) (donkey polyclonal antibody, 715-035-151 JacksonImmunoResearch 1:500) and anti-rabbit IgG (H L) (donkey polyclonal antibody, 715-035-152 JacksonImmunoResearch 1:1000). Fluorescent signal was developed using the OpalTM 7-colorManual IHC Kit. Finally, slides were washed, incubated with 0.5µg/ml4′,6-diamidine-2-phenylindole (DAPI; Invitrogen) for 5min, washed again and mounted with ProlongDiamondmedium (Invitrogen). The slides were scanned using the automated multispectralmicroscopy system Vectra 3.0 (PerkinElmer). An unstained slide was used togenerate the spectral profile of autofluorescence in studied tissues. TheInform software (PerkinElmer) was used for (i) unmixing the spectra ofindividual fluorophores and autofluorescence, (ii) performing automated tissueand cell segmentation into tumor and non-tumor cells, and (iii) exportingsingle-cell intensity values for all fluorescent signals. Single-cell parameterfiles were transformed into.fcs files using R script (version 3.6.1) involvingFlowCore and Biobase packages and were analyzed using FlowJo software to gateon CD39 PD-1 CD8  T cell populations intumor tissue category (Supplementary Fig. 11f). Gates were set based onthe intensity values identified in the fluorescent images. Cell density valueswere measured as number of cells/mm2 obtained as cell countsfrom the gated populations divided by the tumor tissue area measured in thecorresponding patient. Samples were analyzed blinded.

Statistical analysis

Data were analyzed using GraphPad Prism 8.0 forWindows or Mac, and RStudio (version 1.2.5019). For comparison of 2 experimentalgroups, 2-tailed Student’s t test with Welch’s correction wasperformed. More than 2 groups were compared using an ANOVA test withBonferroni’s correction. Data following a logarithmic distribution (i.e., IVISsignal) were log transformed prior to analysis. The number of mice perexperimental group was determined based on our previous experience with similarmodels. Every point represents one mouse. Unless stated otherwise, data areshown as mean±SD. p<0.05 isconsidered significant throughout. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. CellCnn data were analyzed using theKolmogorov–Smirnov two-sample statistical test. In addition, the populationidentified by the algorithm was compared in the two groups using the two-tailedStudent’s t test with Welch’s correction. Survival wascompared for patients with high and low density of cells to be compared asindicated for the individual graphs by Kaplan–Meier curves. Confidenceintervals are indicated as shaded areas surrounding survival curves.Significance was calculated by log-rank test. Patient numbers at risk aredisplayed for each 15 years of follow-up. The threshold of cell type densitywas identified using the Receiver Operating Characteristics (ROC) curveanalysis.

Density of CD39 PD-1 CD8  Tcells was correlated with the clinicopathological information of the patientsusing corresponding R packages. Disease-free survival was analyzed byKaplan–Meier curve and log-rank test (Survminer and Survival R packages), aswell as univariate and multivariate Cox regression (pROC, ROCR, and Survival Rpackages). The threshold of CD39 PD-1 CD8  orCD8  T cell density was identified by the receiver operatingcharacteristics (ROC) curve analysis using 3-year disease-free survival fordefining long- and short-term survivors. Following values apply: Intra-tumoralCD39 PD-1 CD8  T cells, 4.20;extra-tumoral CD39 PD-1 CD8  T cells,8.43; total CD39 PD-1 CD8  T cells, 5.71;intra-tumoral CD8  T cells, 249.20; extra-tumoral CD8  Tcells, 1218.60; total CD8  T cells, 562.29.

Reporting summary

Further information on research design is availablein the Nature Research Reporting Summary linked to this article.

Data availability

RNA-sequencing data that support the findings ofthis study have been deposited in NCBI with the accession code PRJNA609233. The remainingdata are available within the Article, Supplementary Information or available from the authors upon request.

References

1. Kennecke, H. et al. Metastatic behavior of breastcancer subtypes. J. Clin. Oncol. https://doi.org/10.1200/JCO.2009.25.9820 (2010).

2. Colzani, E. et al. Time-dependent risk ofdeveloping distant metastasis in breast cancer patients according to treatment,age and tumour characteristics. Br. J. Cancer https://doi.org/10.1038/bjc.2014.5 (2014).

3. Pan, H. et al. 20-year risks of breast-cancerrecurrence after stopping endocrine therapy at 5 years. N. Engl. J.Med. https://doi.org/10.1056/NEJMoa1701830 (2017).

4. Karrison, T. G., Ferguson, D. J. & Meier, P.Dormancy of mammary carcinoma after mastectomy. J. Natl Cancer Inst. https://doi.org/10.1093/jnci/91.1.80 (1999).

5. Demicheli, R., Abbattista, A., Miceli, R.,Valagussa, P. & Bonadonna, G. Time distribution of the recurrence risk forbreast cancer patients undergoing mastectomy: Further support about the conceptof tumor dormancy. Breast Cancer Res. Treat. https://doi.org/10.1007/BF01807163 (1996).

6. Sosa, M. S., Bragado, P. & Aguirre-Ghiso, J.A. Mechanisms of disseminated cancer cell dormancy: an awakening field. Nat.Rev. Cancer 14, 611–622 (2014).

CAS Article Google Scholar 

7. Braun, S. et al. A pooled analysis of bone marrowmicrometastasis in breast cancer. N. Engl. J. Med. https://doi.org/10.1056/NEJMoa050434 (2005).

8. Janni, W. et al. Persistence of disseminated tumorcells in the bone marrow of breast cancer patients predicts increased risk forrelapse—A European pooled analysis. Clin. Cancer Res. https://doi.org/10.1158/1078-0432.CCR-10-2515 (2011).

9. Balz, L. M. et al. The interplay of HER2/HER3/PI3Kand EGFR/HER2/PLC-γ1 signalling in breast cancer cell migration anddissemination. J. Pathol. https://doi.org/10.1002/path.3991 (2012).

10. Ranganathan, A. C., Zhang, L., Adam, A. P. &Aguirre-Ghiso, J. A. Functional coupling of p38-induced up-regulation of BiPand activation of RNA-dependent protein kinase-like endoplasmic reticulumkinase to drug resistance of dormant carcinoma cells. Cancer Res. https://doi.org/10.1158/0008-5472.CAN-05-3092. (2006).

11. Ghajar, C. M. Metastasis prevention by targetingthe dormant niche. Nat. Rev. Cancer https://doi.org/10.1038/nrc3910 (2015).

12. Ghajar, C. M. et al. The perivascular niche regulatesbreast tumour dormancy. Nat. Cell Biol. 15, 807–817(2013).

CAS Article Google Scholar 

13. Bragado, P. et al. TGF-β2 dictates disseminatedtumour cell fate in target organs through TGF-β-RIII and p38α/βsignalling. Nat. Cell Biol. https://doi.org/10.1038/ncb2861 (2013).

14. Hagerling, C. et al. Immune effectormonocyte–neutrophil cooperation induced by the primary tumor preventsmetastatic progression of breast cancer. Proc. Natl Acad. Sci. USA https://doi.org/10.1073/pnas.1907660116 (2019).

15. Malladi, S. et al. Metastatic latency and immuneevasion through autocrine inhibition of WNT. Cell 165,45–60 (2016).

16. Eyles, J. et al. Tumor cells disseminate early,but immunosurveillance limits metastatic outgrowth突起, in a mouse model of melanoma. J. Clin.Invest 120, 2030–2039 (2010).

17. Pommier, A. et al. Unresolved endoplasmicreticulum stress engenders immune-resistant, latent pancreatic cancermetastases. Science https://doi.org/10.1126/science.aao4908 (2018).

18. Ali, H. R. et al. Association between CD8 T-cellinfiltration and breast cancer survival in 12,439 patients. Ann. Oncol. https://doi.org/10.1093/annonc/mdu191 (2014).

19. Savas, P. et al. Single-cell profiling of breastcancer T cells reveals a tissue-resident memory subset associated with improvedprognosis. Nat. Med. https://doi.org/10.1038/s41591-018-0078-7 (2018).

20. Simoni, Y. et al. Bystander CD8 T cells areabundant and phenotypically distinct in human tumour infiltrates. Nature https://doi.org/10.1038/s41586-018-0130-2 (2018).

21. Aslakson, C. J. & Miller, F. R. Selectiveevents in the metastatic process defined by analysis of the sequentialdissemination of subpopulations of a mouse mammary tumor. Cancer Res52,1399–1405 (1992).

22. Pulaski, B. A. & Ostrand-Rosenberg, S. Mouse4T1 breast tumor model. Curr. Protocol. Immunol. https://doi.org/10.1002/0471142735.im2002s39 (2001).

23. Beffinger, M. et al. CSF1R-dependent myeloid cellsare required for NKmediated control of metastasis. JCI Insight 3,e97792 (2018).

24. Carlson, P. et al. Targeting the perivascularniche sensitizes disseminated tumour cells to chemotherapy. Nat. CellBiol. https://doi.org/10.1038/s41556-018-0267-0 (2019).

25. Peinado, H. et al. Pre-metastatic niches:organ-specific homes for metastases. Nat. Rev. Cancer 17,302–317 (2017).

26. Spiegel, A. et al. Neutrophils suppressintraluminal NK cell-mediated tumor cell clearance and enhance extravasation ofdisseminated carcinoma cells. Cancer Disco. 6, 630–649(2016).

27. Kowanetz, M. et al. Granulocyte-colony stimulatingfactor promotes lung metastasis through mobilization of Ly6G Ly6C granulocytes. Proc. Natl Acad. Sci. USA 107,21248–21255 (2010).

28. Ouzounova, M. et al. Monocytic and granulocyticmyeloid derived suppressor cells differentially regulate spatiotemporal tumourplasticity during metastatic cascade. Nat. Commun. https://doi.org/10.1038/ncomms14979 (2017).

29. Sharma, S. K. et al. Pulmonary alveolarmacrophages contribute to the premetastatic niche by suppressing antitumor Tcell responses in the lungs. J. Immunol. https://doi.org/10.4049/jimmunol.1403215 (2015).

30. Van Der Maaten, L. & Hinton, G. Visualizingdata using t-SNE. J. Mach. Learn. Res. 9, 2579–2605(2008).

31. Van Gassen, S. et al. FlowSOM: Usingself-organizing maps for visualization and interpretation of cytometrydata. Cytometry A. https://doi.org/10.1002/cyto.a.22625 (2015).

32. Weber, L. M. & Robinson, M. D. Comparison ofclustering methods for high-dimensional single-cell flow and mass cytometrydata. Cytometry A. https://doi.org/10.1002/cyto.a.23030 (2016).

33. Arvaniti, E. & Claassen, M. Sensitivedetection of rare disease-Associated cell subsets via representationlearning. Nat. Commun. https://doi.org/10.1038/ncomms14825 (2017)

34. Krieg, C. et al. High-dimensional single-cellanalysis predicts response to anti-PD-1 immunotherapy. Nat. Med. 24,144–153 (2018).

35. Anderson, A. C., Joller, N. & Kuchroo, V. K.Lag-3, Tim-3, and TIGIT: co-inhibitory receptors with specialized functions inimmune regulation. Immunity 44, 989–1004 (2016).

36. Pandiyan, P., Hegel, J. K. E., Krueger, M.,Quandt, D. & Brunner-Weinzierl, M. C. High IFN-γ production of individualCD8 T lymphocytes is controlled by CD152 (CTLA-4). J. Immunol. 178,2132–2140 (2007).

37. Workman, C. J. et al. Lymphocyte activation gene-3(CD223) Regulates the size of the expanding T cell population following antigenactivation in vivo. J. Immunol. 172, 5450–5455 (2004).

38. Koyama, S. et al. Adaptive resistance totherapeutic PD-1 blockade is associated with upregulation of alternative immunecheckpoints. Nat. Commun. 7, 1–9 (2016).

39. Clarke, J. et al. Single-cell transcriptomicanalysis of tissue-resident memory T cells in human lung cancer. J.Exp. Med. https://doi.org/10.1084/jem.20190249 (2019).

40. Park, C. O. & Kupper, T. S. The emerging roleof resident memory T cells in protective immunity and inflammatorydisease. Nat. Med. https://doi.org/10.1038/nm.3883 (2015).

41. Egelston, C. A. et al. Resident memory CD8 Tcells within cancer islands mediate survival in breast cancer patients. JCIInsight https://doi.org/10.1172/jci.insight.130000 (2019).

42. Braumuller, H. et al. T-helper-1-cell cytokinesdrive cancer into senescence. Nature 494, 361–365(2013).

43. Subramanian, A. et al. Gene set enrichmentanalysis: a knowledge-based approach for interpreting genome-wide expressionprofiles. Proc. Natl Acad. Sci. USA https://doi.org/10.1073/pnas.0506580102 (2005).

44. Scott, A. C. et al. TOX is a critical regulator oftumour-specific T cell differentiation. Nature https://doi.org/10.1038/s41586-019-1324-y (2019).

45. Costa-Silva, B. et al. Pancreatic cancer exosomesinitiate pre-metastatic niche formation in the liver. Nat. Cell Biol. 17,816–826 (2015).

46. Wculek, S. K. & Malanchi, I. Neutrophilssupport lung colonization of metastasis-initiating breast cancer cells. Nature 528,413–417 (2015).

47. Duhen, T. et al. Co-expression of CD39 and CD103identifies tumor-reactive CD8 T cells in human solid tumors. Nat.Commun. https://doi.org/10.1038/s41467-018-05072-0 (2018).

48. Wagner, J. et al. A single-cell atlas of the tumorand immune ecosystem of human breast cancer. Cell https://doi.org/10.1016/j.cell.2019.03.005 (2019).

49. Yan, J. et al. Control of metastases via myeloidCD39 and NK cell effector function. Cancer Immunol. Res. https://doi.org/10.1158/2326-6066.CIR-19-0749. (2020).

50. Li, X. Y. et al. Targeting CD39 in cancer revealsan extracellular ATP-and inflammasome-driven tumor immunity. CancerDiscov. https://doi.org/10.1158/2159-8290.CD-19-0541 (2019).

51. Canale, F. P. et al. CD39 expression defines cellexhaustion in tumor-infiltrating CD8 T cells. Cancer Res. https://doi.org/10.1158/0008-5472.CAN-16-2684 (2018).

52. Egelston, C. et al. Complex phenotyping of PD-1 CD39 exhausted CD8 T cells in human carcinomas. J. Immunol. 200,(2018).

53. Lee, Y., Park, J., Park, S.-H. & Shin, E.-C.CD39 CD8  T cells exhibit a distinct phenotype amongtumor-infiltrating tumor-antigenspecific CD8  T cells. J.Immunol202, (2019).

54. Blank, C. U. et al. Defining 'T cellexhaustion’. Nat. Rev. Immunol. https://doi.org/10.1038/s41577-019-0221-9 (2019).

55. McLane, L. M., Abdel-Hakeem, M. S. & Wherry,E. J. CD8 T cell exhaustion during chronic viral infection and cancer. Annu.Rev. Immunol. https://doi.org/10.1146/annurev-immunol-041015-055318 (2019).

56. Egelston, C. A. et al. Human breast tumor-infiltratingCD8 T cells retain polyfunctionality despite PD-1 expression. Nat.Commun. https://doi.org/10.1038/s41467-018-06653-9. (2018)

57. Spitzer, M. H. et al. systemic系统性 immunity is required for effective cancerimmunotherapy. Cell https://doi.org/10.1016/j.cell.2016.12.022 (2017).

58. Menares, E. et al. Tissue-resident memory CD8 Tcells amplify anti-tumor immunity by triggering antigen spreading throughdendritic cells. Nat. Commun. https://doi.org/10.1038/s41467-019-12319-x (2019)

59. Goddard, E. T., Bozic, I., Riddell, S. R. &Ghajar, C. M. Dormant tumour cells, their niches and the influence ofimmunity. Nat. Cell Biol. https://doi.org/10.1038/s41556-018-0214-0 (2018).

60. Aguirre-Ghiso, J. A. Models, mechanisms andclinical evidence for cancer dormancy. Nat. Rev. Cancer https://doi.org/10.1038/nrc2256 (2007).

61. Koebel, C. M. et al. Adaptive immunity maintainsoccult cancer in an equilibrium state. Nature 450,903–907 (2007).

62. Kim, S. Y. et al. Role of theIL-6-JAK1-STAT3-Oct-4 pathway in the conversion of non-stem cancer cells intocancer stem-like cells. Cell. Signal. https://doi.org/10.1016/j.cellsig.2013.01.007 (2013).

63. Baccelli, I. et al. Identification of a populationof blood circulating tumor cells from breast cancer patients that initiatesmetastasis in a xenograft assay. Nat. Biotechnol. https://doi.org/10.1038/nbt.2576 (2013).

64. Slütter, B. et al. Dynamics of influenza-inducedlung-resident memory T cells underlie waning heterosubtypic immunity. Sci.Immunol. https://doi.org/10.1126/sciimmunol.aag2031 (2017).

65. Welten, S. P. M., Sandu, I., Baumann, N. S. &Oxenius, A. Memory CD8 T cell inflation vs tissue-resident memory T cells: samepatrollers, same controllers? Immunol. Rev. https://doi.org/10.1111/imr.12649 (2018).

66. Pereira, B. I. et al. Senescent cells evade immuneclearance via HLA-E-mediated NK and CD8 T cell inhibition. Nat.Commun. https://doi.org/10.1038/s41467-019-10335-5 (2019).

67. Cox, T. R. et al. LOX-mediated collagencrosslinking is responsible for fibrosis-enhanced metastasis. CancerRes. https://doi.org/10.1158/0008-5472.CAN-12-2233 (2013).

68. Goddard, E. T. et al. The rodent liver undergoesweaning-induced involution and supports breast cancer metastasis. CancerDiscov. https://doi.org/10.1158/2159-8290.CD-16-0822 (2017).

69. Quail, D. F. et al. Obesity alters the lungmyeloid cell landscape to enhance breast cancer metastasis through IL5 andGM-CSF. Nat. Cell Biol. https://doi.org/10.1038/ncb3578 (2017).

70. Rachman-Tzemah, C. et al. Blocking surgicallyinduced lysyl oxidase activity reduces the risk of lung metastases. CellRep. https://doi.org/10.1016/j.celrep.2017.04.005 (2017).

71. O’Reilly, M. S. et al. Angiostatin: A novelangiogenesis inhibitor that mediates the suppression of metastases by a lewislung carcinoma. Cell https://doi.org/10.1016/0092-8674(94)90200-3 (1994).

72. Obradović, M. M. S. et al. Glucocorticoids promotebreast cancer metastasis. Nature https://doi.org/10.1038/s41586-019-1019-4 (2019).

73. Albrengues, J. et al. Neutrophil extracellulartraps produced during inflammation awaken dormant cancer cells in mice. Science https://doi.org/10.1126/science.aao4227 (2018).

74. Brummelman, J. et al. Development, application andcomputational analysis of high-dimensional fluorescent antibody panels forsingle-cell flow cytometry. Nat. Protoc. https://doi.org/10.1038/s41596-019-0166-2 (2019).

75. Picelli, S. et al. Full-length RNA-seq from singlecells using Smart-seq2. Nat. Protoc. https://doi.org/10.1038/nprot.2014.006 (2014).

76. Hatakeyama, M. et al. SUSHI: an exquisite recipefor fully documented, reproducible and reusable NGS data analysis. BMCBioinformatics 17, 228 (2016).

77. Qi, W., Schlapbach, R. & Rehrauer, H.In: Plant Germline Development. (ed. Schmidt, A.) pp 295–307(Humana Press, New York, 2017).

78. Reimand, J. et al. Pathway enrichment analysis andvisualization of omics data using g:Profiler, GSEA, Cytoscape andEnrichmentMap. Nat. Protoc. https://doi.org/10.1038/s41596-018-0103-9 (2019).

79. Siliņa, K., Burkhardt, C., Casanova, R.,Solterman, A. & van den Broek, M. A Quantitative pathology approach toanalyze the development of human cancer-associated tertiary lymphoidstructures. Methods Mol. Biol. https://doi.org/10.1007/978-1-4939-8709-2_5 (2018).

80. Goldrath, A. W., Luckey, C. J., Park, R., Benoist,C. & Mathis, D. The molecular program induced in T cells undergoinghomeostatic proliferation. Proc. Natl Acad. Sci. USA https://doi.org/10.1073/pnas.0407417101 (2004).

(0)

相关推荐